apps/web/app/lib · flagon-io/g1t

about.server.tsThe repository sidebar reads like it should. TSX counts as TypeScript, as JSX already counted as JavaScript, so one language is not split in two (stats version 3, so every repository is read again); the languages list every one down to a tenth of a percent, the largest eight, rather than folding MDX, CSS and Shell into Other; up to five contributors are listed by name, each with their commits and card, more as avatars; and with no release published, Releases says how many tags there are, linked to them.5d ago

about.test.tsThe Branches page does things to branches now. It has tabs (Overview, Yours, Active, Stale, All, kept in the address as ?tab=) and search; Overview shows the default branch, then the first five of your branches, the active ones and the stale ones, each linking to the rest. Each branch is a row with its name in a code chip with a copy button, who made its latest commit and when, that commit's checks, behind and ahead counts with bars drawn against the largest shown, and its pull request (the open one, else the newest merged or closed) with its state; on a phone the columns fold under the name. Yours is branches whose latest commit is the viewer's; stale is no commit in 90 days.5d ago

about.tsThe Branches page does things to branches now. It has tabs (Overview, Yours, Active, Stale, All, kept in the address as ?tab=) and search; Overview shows the default branch, then the first five of your branches, the active ones and the stale ones, each linking to the rest. Each branch is a row with its name in a code chip with a copy button, who made its latest commit and when, that commit's checks, behind and ahead counts with bars drawn against the largest shown, and its pull request (the open one, else the newest merged or closed) with its state; on a phone the columns fold under the name. Yours is branches whose latest commit is the viewer's; stale is no commit in 90 days.5d ago

access-tokens.server.tsOne kind of access token; presence and status; usernames keep their case; the tour is a miniature of the real app; icons for password managers1w ago

access-tokens.test.tsMerge main into Artifacts Phase 21w ago

access-tokens.tsThe Agent preset creates repositories, with a new repo:create scope (#52 )2d ago

access.server.tsMerge branch 'worktree-agent-a4babc5b5815f6f3a'4d ago

access.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

account-deletion.test.tsMerge sudo: delete an account with the workspaces it alone owns, purge each1w ago

account-deletion.tsMerge account deletion: soft delete for 30 days, staff restore and purge, ghost for what remains (identity 0037)1w ago

account-settings.test.tsMerge branch 'worktree-agent-a6da31375506fc440'4d ago

account-settings.tsMerge branch 'worktree-agent-a6da31375506fc440'4d ago

add-ci.server.tsFast pages, required checks on the branch, self-hosted runners, honest incidents1w ago

addresses.server.tsMerge g1tusercontent.com: registry answers run nothing in a browser, the site's pages run only their own scripts, repository files and avatars on their own origin, raw files rate limited per address1w ago

addresses.test.tsMerge g1tusercontent.com: registry answers run nothing in a browser, the site's pages run only their own scripts, repository files and avatars on their own origin, raw files rate limited per address1w ago

addresses.tsMerge g1tusercontent.com: registry answers run nothing in a browser, the site's pages run only their own scripts, repository files and avatars on their own origin, raw files rate limited per address1w ago

advisories.test.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

advisories.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

agent-setup.test.tsg1t's Claude Code plugin installs from its own repository, published by a workspace token, and a token for chosen repositories no longer creates others by pushing. `claude plugin marketplace add https://g1t.sh/flagon-io/g1t-plugins.git` then `claude plugin install g1t@g1t` clones only the plugin, not this monorepo, which stays its source (.claude-plugin/marketplace.json, plugins/claude-code/). scripts/plugins-release.mjs builds the published tree (the plugin and a marketplace that points at it), checks that both manifests agree and that nothing else is in it, and pushes it with a tag `g1t--v<version>`; it refuses changed plugin files under a version already published, and says "nothing to publish" when that version is there. .g1t/workflows/plugins-release.yml runs it on pushes to main with the flagon-io workspace token PLUGINS_PUBLISH_TOKEN, sent as Basic auth, never a person's token; CI runs its tests. The README, the plugin guide, the docs' agent list, the MCP reference and its instructions, bring-your-own-agent, why-blame, the site's agent setup and the Live now empty state, llms.txt, install/claude.sh and bench/ask all give the new address, and the earlier ~/.g1t/hook.mjs install path is gone. Automation is pointed at a workspace's own tokens: the API reference's Authentication section, Actions' note on changing workflows from a job (it said a personal token) and llms.txt (never ask your person for a personal token for that) now say so, as Settings → Access tokens, the authentication and workspaces guides and git's deploy keys already did. Pushing to an address that does not exist made a repository for any token holding code:write and a member's or the workspace's identity, so a workspace token for only g1t-plugins could create flagon-io/<anything> and push to it; repos' git_token_refusal now refuses push-to-create for a token whose reach is chosen repositories, public ones only, or another workspace, and the git guide says so. Checked: the plugin's hook tests (49), plugins-release tests (8) and its check (g1t@g1t 0.2.2), cargo test -p g1t-repos --lib (314, new: chosen repositories never create one, all of a workspace creates there only, a token without a reach as before) and its wasm32 check, the site's agent-setup and llms.txt tests (20), the docs build (513 pages).3d ago

agent-setup.tsg1t's Claude Code plugin installs from its own repository, published by a workspace token, and a token for chosen repositories no longer creates others by pushing. `claude plugin marketplace add https://g1t.sh/flagon-io/g1t-plugins.git` then `claude plugin install g1t@g1t` clones only the plugin, not this monorepo, which stays its source (.claude-plugin/marketplace.json, plugins/claude-code/). scripts/plugins-release.mjs builds the published tree (the plugin and a marketplace that points at it), checks that both manifests agree and that nothing else is in it, and pushes it with a tag `g1t--v<version>`; it refuses changed plugin files under a version already published, and says "nothing to publish" when that version is there. .g1t/workflows/plugins-release.yml runs it on pushes to main with the flagon-io workspace token PLUGINS_PUBLISH_TOKEN, sent as Basic auth, never a person's token; CI runs its tests. The README, the plugin guide, the docs' agent list, the MCP reference and its instructions, bring-your-own-agent, why-blame, the site's agent setup and the Live now empty state, llms.txt, install/claude.sh and bench/ask all give the new address, and the earlier ~/.g1t/hook.mjs install path is gone. Automation is pointed at a workspace's own tokens: the API reference's Authentication section, Actions' note on changing workflows from a job (it said a personal token) and llms.txt (never ask your person for a personal token for that) now say so, as Settings → Access tokens, the authentication and workspaces guides and git's deploy keys already did. Pushing to an address that does not exist made a repository for any token holding code:write and a member's or the workspace's identity, so a workspace token for only g1t-plugins could create flagon-io/<anything> and push to it; repos' git_token_refusal now refuses push-to-create for a token whose reach is chosen repositories, public ones only, or another workspace, and the git guide says so. Checked: the plugin's hook tests (49), plugins-release tests (8) and its check (g1t@g1t 0.2.2), cargo test -p g1t-repos --lib (314, new: chosen repositories never create one, all of a workspace creates there only, a token without a reach as before) and its wasm32 check, the site's agent-setup and llms.txt tests (20), the docs build (513 pages).3d ago

ai-providers.test.tsAdding an AI provider now says where to get its key, and how. The hint under each key links to the provider's key page (Anthropic's is now platform.claude.com, where the Console lives) and says what to choose there, and whether the account needs credit first: a Claude.ai or ChatGPT subscription does not include the API. Azure OpenAI links to the portal and says where Keys and Endpoint are. The AI guide has the same in a table under "Get a key", and a test asks every provider that takes a key for both. Checked: the AI providers' tests (3 passing) and the site's type-check of these files.2d ago

ai-providers.tsAdding an AI provider now says where to get its key, and how. The hint under each key links to the provider's key page (Anthropic's is now platform.claude.com, where the Console lives) and says what to choose there, and whether the account needs credit first: a Claude.ai or ChatGPT subscription does not include the API. Azure OpenAI links to the portal and says where Keys and Endpoint are. The AI guide has the same in a table under "Get a key", and a test asks every provider that takes a key for both. Checked: the AI providers' tests (3 passing) and the site's type-check of these files.2d ago

analytics-consent.test.tsMerge site-audit: SEO routes, analytics with consent first in Europe, security page, founder and X account, docs header; the tour opens on the pull request and leads with Code, on the real shell, with Agents and Docs working today1w ago

analytics-consent.tsMerge site-audit: SEO routes, analytics with consent first in Europe, security page, founder and X account, docs header; the tour opens on the pull request and leads with Code, on the real shell, with Agents and Docs working today1w ago

analytics-scrub.test.tsMerge site-audit: SEO routes, analytics with consent first in Europe, security page, founder and X account, docs header; the tour opens on the pull request and leads with Code, on the real shell, with Agents and Docs working today1w ago

analytics-scrub.tsMerge branch 'worktree-agent-a4babc5b5815f6f3a'4d ago

analytics.client.tsMerge site-audit: SEO routes, analytics with consent first in Europe, security page, founder and X account, docs header; the tour opens on the pull request and leads with Code, on the real shell, with Agents and Docs working today1w ago

ansi.test.tsA running job's log is live, as on GitHub. A step's output reaches its run's page within about two seconds: the runner sends what is printed every second, and the page asks every second while a step runs, every three while none does; an answer that fails (a 429, a 5xx, the network) is asked again after a wait that doubles to half a minute, or as long as Retry-After says, where before one such answer stopped the log for good and the step said "No output yet." until the page was reloaded, and a signed-out viewer's run page, refreshing beside its log, ran into the 30-a-minute limit within a minute; the live log now counts as a page, and only log downloads as costly. A hidden tab stops asking and catches up when shown, and the page says when it is reconnecting. The running step opens by itself, its timer counts, its log follows the newest line while its end is on screen and offers Jump to latest once the viewer scrolls up, a step that has printed nothing for a while says "Running · no output for 2m 10s", and the job's header names the step it is on. Colour that tools print (bold, dim, italic, underline, the 16 colours as theme tokens for light and dark, 256 colours and 24-bit colour, text and background) is drawn as printed, other escape codes are dropped and links keep their text, and search reads the plain text and keeps the colours around what it marks. Every line is stamped with the time it was printed, as GitHub's raw logs are, so log.txt and the run's zip have a time at the start of each line, and a clock button beside Search logs shows the times, remembered by the browser. Set up job now gives the runner's version and name, the operating system, the runner image, the machine's architecture, CPUs and memory, the token's permissions under G1T_TOKEN Permissions, the secret source and the actions it downloads; each run step opens with GitHub's folded Run group, the script in cyan, the shell's whole path and the step's env, each uses step with its with inputs (an action's defaults among them) and env; cache and artifact transfers say how far they have got every five seconds; and a last Complete job step sets the job's outputs and stops what the steps left running, found by RUNNER_TRACKING_ID as GitHub's runner finds it. Two batches of a job's log written at once can no longer reach for the same seq and lose one. The Actions guide and the rate limits page say so.5d ago

ansi.tsA running job's log is live, as on GitHub. A step's output reaches its run's page within about two seconds: the runner sends what is printed every second, and the page asks every second while a step runs, every three while none does; an answer that fails (a 429, a 5xx, the network) is asked again after a wait that doubles to half a minute, or as long as Retry-After says, where before one such answer stopped the log for good and the step said "No output yet." until the page was reloaded, and a signed-out viewer's run page, refreshing beside its log, ran into the 30-a-minute limit within a minute; the live log now counts as a page, and only log downloads as costly. A hidden tab stops asking and catches up when shown, and the page says when it is reconnecting. The running step opens by itself, its timer counts, its log follows the newest line while its end is on screen and offers Jump to latest once the viewer scrolls up, a step that has printed nothing for a while says "Running · no output for 2m 10s", and the job's header names the step it is on. Colour that tools print (bold, dim, italic, underline, the 16 colours as theme tokens for light and dark, 256 colours and 24-bit colour, text and background) is drawn as printed, other escape codes are dropped and links keep their text, and search reads the plain text and keeps the colours around what it marks. Every line is stamped with the time it was printed, as GitHub's raw logs are, so log.txt and the run's zip have a time at the start of each line, and a clock button beside Search logs shows the times, remembered by the browser. Set up job now gives the runner's version and name, the operating system, the runner image, the machine's architecture, CPUs and memory, the token's permissions under G1T_TOKEN Permissions, the secret source and the actions it downloads; each run step opens with GitHub's folded Run group, the script in cyan, the shell's whole path and the step's env, each uses step with its with inputs (an action's defaults among them) and env; cache and artifact transfers say how far they have got every five seconds; and a last Complete job step sets the job's outputs and stops what the steps left running, found by RUNNER_TRACKING_ID as GitHub's runner finds it. Two batches of a job's log written at once can no longer reach for the same seq and lose one. The Actions guide and the rate limits page say so.5d ago

apps.test.tsA workspace's AI is its own setting, not an integration. AI providers and what Ask answers with were connections like Sentry's or Linear's: rows of `connections`, cards in the Integrations directory, `provider` values of the integration operations. Integrations are becoming a directory others publish to; a workspace's AI is a base setting g1t looks after itself. Contracts: `g1t_contracts::ai` (and packages/contracts ai.ts) has AiProvider (the 14), AiGroup (labs, platforms, self_hosted), AiConfig, AiConnection and their methods' arguments, and the Ask types moved there; AskSetup names the provider by `provider_id` and `provider_name` (were connection_id, connection_name). `integrations::Provider` is sentry, datadog, webhook, jira and linear; ProviderKind loses Models, ConnectionConfig its base_url, auth_header and model, Connection its models. Storage: integrations migration 0011 makes `ai_providers` and `ai_routes`, moves the AI rows out of `connections` with their ids (sealed keys are bound to the id, and managed_agents and agent_runs keep finding them), keeps Ask's route and drops `model_routes`; renaming and deleting a workspace cover both tables. The service keeps them in src/ai.rs with methods of their own (ai_list, ai_add, ai_update, ai_remove, ai_test, ask_settings, set_ask_settings, ask); a provider's key is checked when added and on each test; removing one turns Ask off when it answered with it and forgets its managed agent. Sessions on issues find the workspace's Anthropic provider there. API: list_ai_providers, add_ai_provider, update_ai_provider, remove_ai_provider, test_ai_provider (/workspaces/{workspace}/ai/providers[/{id}[/test]]), get_ask_settings and set_ask_settings (/workspaces/{workspace}/ai/ask), under the OpenAPI tag AI; reading takes workspace:read, changing workspace:admin and an owner's own token; the integration operations take only integration providers. MCP: a tool `ai` (list_providers, add_provider, update_provider, remove_provider, test_provider, get_ask, set_ask). The CLI's operation table and command reference are regenerated. Site: AI at /<workspace>/-/ai (the sidebar's AI → Models, the command palette's AI models) holds Ask g1t, the providers and Add a provider, grouped Frontier labs, Model platforms, Self-hosted or your own gateway; AI providers are no longer in the Integrations directory or Apps, and the integrations pages are alerts and trackers only; -/integrations/models and ?add=<an AI provider> lead to -/ai. Ask's settings link, and @g1t's "not set up yet" in Chat, point there. Docs: the AI providers guide (with its API), and the integrations, Ask, sessions, CLI, plugin and MCP pages say so. Deploy the integrations migration first, then integrations, then api, chat and web. Checked: cargo test for g1t-contracts (309), g1t-integrations (70), g1t-api (204, the OpenAPI document regenerated) and the CLI (105 and 44, the command reference current); cargo check --workspace; wasm32 checks of api and integrations; every migration applied in order to SQLite with an AI provider, a Sentry connection and Ask's route; the site's tsc -b --force, 927 tests and a production build; contracts' 48; chat's 76 and typecheck; the deploy tool's 92; the docs build (521 pages).3d ago

apps.tsA workspace's AI is its own setting, not an integration. AI providers and what Ask answers with were connections like Sentry's or Linear's: rows of `connections`, cards in the Integrations directory, `provider` values of the integration operations. Integrations are becoming a directory others publish to; a workspace's AI is a base setting g1t looks after itself. Contracts: `g1t_contracts::ai` (and packages/contracts ai.ts) has AiProvider (the 14), AiGroup (labs, platforms, self_hosted), AiConfig, AiConnection and their methods' arguments, and the Ask types moved there; AskSetup names the provider by `provider_id` and `provider_name` (were connection_id, connection_name). `integrations::Provider` is sentry, datadog, webhook, jira and linear; ProviderKind loses Models, ConnectionConfig its base_url, auth_header and model, Connection its models. Storage: integrations migration 0011 makes `ai_providers` and `ai_routes`, moves the AI rows out of `connections` with their ids (sealed keys are bound to the id, and managed_agents and agent_runs keep finding them), keeps Ask's route and drops `model_routes`; renaming and deleting a workspace cover both tables. The service keeps them in src/ai.rs with methods of their own (ai_list, ai_add, ai_update, ai_remove, ai_test, ask_settings, set_ask_settings, ask); a provider's key is checked when added and on each test; removing one turns Ask off when it answered with it and forgets its managed agent. Sessions on issues find the workspace's Anthropic provider there. API: list_ai_providers, add_ai_provider, update_ai_provider, remove_ai_provider, test_ai_provider (/workspaces/{workspace}/ai/providers[/{id}[/test]]), get_ask_settings and set_ask_settings (/workspaces/{workspace}/ai/ask), under the OpenAPI tag AI; reading takes workspace:read, changing workspace:admin and an owner's own token; the integration operations take only integration providers. MCP: a tool `ai` (list_providers, add_provider, update_provider, remove_provider, test_provider, get_ask, set_ask). The CLI's operation table and command reference are regenerated. Site: AI at /<workspace>/-/ai (the sidebar's AI → Models, the command palette's AI models) holds Ask g1t, the providers and Add a provider, grouped Frontier labs, Model platforms, Self-hosted or your own gateway; AI providers are no longer in the Integrations directory or Apps, and the integrations pages are alerts and trackers only; -/integrations/models and ?add=<an AI provider> lead to -/ai. Ask's settings link, and @g1t's "not set up yet" in Chat, point there. Docs: the AI providers guide (with its API), and the integrations, Ask, sessions, CLI, plugin and MCP pages say so. Deploy the integrations migration first, then integrations, then api, chat and web. Checked: cargo test for g1t-contracts (309), g1t-integrations (70), g1t-api (204, the OpenAPI document regenerated) and the CLI (105 and 44, the command reference current); cargo check --workspace; wasm32 checks of api and integrations; every migration applied in order to SQLite with an AI provider, a Sentry connection and Ask's route; the site's tsc -b --force, 927 tests and a production build; contracts' 48; chat's 76 and typecheck; the deploy tool's 92; the docs build (521 pages).3d ago

art.test.tsInvite-only launch: sign in with GitHub, repository access and lifecycle, many emails, a new look1w ago

art.tsInvite-only launch: sign in with GitHub, repository access and lifecycle, many emails, a new look1w ago

ask-model.test.tsA workspace's AI is its own setting, not an integration. AI providers and what Ask answers with were connections like Sentry's or Linear's: rows of `connections`, cards in the Integrations directory, `provider` values of the integration operations. Integrations are becoming a directory others publish to; a workspace's AI is a base setting g1t looks after itself. Contracts: `g1t_contracts::ai` (and packages/contracts ai.ts) has AiProvider (the 14), AiGroup (labs, platforms, self_hosted), AiConfig, AiConnection and their methods' arguments, and the Ask types moved there; AskSetup names the provider by `provider_id` and `provider_name` (were connection_id, connection_name). `integrations::Provider` is sentry, datadog, webhook, jira and linear; ProviderKind loses Models, ConnectionConfig its base_url, auth_header and model, Connection its models. Storage: integrations migration 0011 makes `ai_providers` and `ai_routes`, moves the AI rows out of `connections` with their ids (sealed keys are bound to the id, and managed_agents and agent_runs keep finding them), keeps Ask's route and drops `model_routes`; renaming and deleting a workspace cover both tables. The service keeps them in src/ai.rs with methods of their own (ai_list, ai_add, ai_update, ai_remove, ai_test, ask_settings, set_ask_settings, ask); a provider's key is checked when added and on each test; removing one turns Ask off when it answered with it and forgets its managed agent. Sessions on issues find the workspace's Anthropic provider there. API: list_ai_providers, add_ai_provider, update_ai_provider, remove_ai_provider, test_ai_provider (/workspaces/{workspace}/ai/providers[/{id}[/test]]), get_ask_settings and set_ask_settings (/workspaces/{workspace}/ai/ask), under the OpenAPI tag AI; reading takes workspace:read, changing workspace:admin and an owner's own token; the integration operations take only integration providers. MCP: a tool `ai` (list_providers, add_provider, update_provider, remove_provider, test_provider, get_ask, set_ask). The CLI's operation table and command reference are regenerated. Site: AI at /<workspace>/-/ai (the sidebar's AI → Models, the command palette's AI models) holds Ask g1t, the providers and Add a provider, grouped Frontier labs, Model platforms, Self-hosted or your own gateway; AI providers are no longer in the Integrations directory or Apps, and the integrations pages are alerts and trackers only; -/integrations/models and ?add=<an AI provider> lead to -/ai. Ask's settings link, and @g1t's "not set up yet" in Chat, point there. Docs: the AI providers guide (with its API), and the integrations, Ask, sessions, CLI, plugin and MCP pages say so. Deploy the integrations migration first, then integrations, then api, chat and web. Checked: cargo test for g1t-contracts (309), g1t-integrations (70), g1t-api (204, the OpenAPI document regenerated) and the CLI (105 and 44, the command reference current); cargo check --workspace; wasm32 checks of api and integrations; every migration applied in order to SQLite with an AI provider, a Sentry connection and Ask's route; the site's tsc -b --force, 927 tests and a production build; contracts' 48; chat's 76 and typecheck; the deploy tool's 92; the docs build (521 pages).3d ago

ask-model.tsA workspace's AI is its own setting, not an integration. AI providers and what Ask answers with were connections like Sentry's or Linear's: rows of `connections`, cards in the Integrations directory, `provider` values of the integration operations. Integrations are becoming a directory others publish to; a workspace's AI is a base setting g1t looks after itself. Contracts: `g1t_contracts::ai` (and packages/contracts ai.ts) has AiProvider (the 14), AiGroup (labs, platforms, self_hosted), AiConfig, AiConnection and their methods' arguments, and the Ask types moved there; AskSetup names the provider by `provider_id` and `provider_name` (were connection_id, connection_name). `integrations::Provider` is sentry, datadog, webhook, jira and linear; ProviderKind loses Models, ConnectionConfig its base_url, auth_header and model, Connection its models. Storage: integrations migration 0011 makes `ai_providers` and `ai_routes`, moves the AI rows out of `connections` with their ids (sealed keys are bound to the id, and managed_agents and agent_runs keep finding them), keeps Ask's route and drops `model_routes`; renaming and deleting a workspace cover both tables. The service keeps them in src/ai.rs with methods of their own (ai_list, ai_add, ai_update, ai_remove, ai_test, ask_settings, set_ask_settings, ask); a provider's key is checked when added and on each test; removing one turns Ask off when it answered with it and forgets its managed agent. Sessions on issues find the workspace's Anthropic provider there. API: list_ai_providers, add_ai_provider, update_ai_provider, remove_ai_provider, test_ai_provider (/workspaces/{workspace}/ai/providers[/{id}[/test]]), get_ask_settings and set_ask_settings (/workspaces/{workspace}/ai/ask), under the OpenAPI tag AI; reading takes workspace:read, changing workspace:admin and an owner's own token; the integration operations take only integration providers. MCP: a tool `ai` (list_providers, add_provider, update_provider, remove_provider, test_provider, get_ask, set_ask). The CLI's operation table and command reference are regenerated. Site: AI at /<workspace>/-/ai (the sidebar's AI → Models, the command palette's AI models) holds Ask g1t, the providers and Add a provider, grouped Frontier labs, Model platforms, Self-hosted or your own gateway; AI providers are no longer in the Integrations directory or Apps, and the integrations pages are alerts and trackers only; -/integrations/models and ?add=<an AI provider> lead to -/ai. Ask's settings link, and @g1t's "not set up yet" in Chat, point there. Docs: the AI providers guide (with its API), and the integrations, Ask, sessions, CLI, plugin and MCP pages say so. Deploy the integrations migration first, then integrations, then api, chat and web. Checked: cargo test for g1t-contracts (309), g1t-integrations (70), g1t-api (204, the OpenAPI document regenerated) and the CLI (105 and 44, the command reference current); cargo check --workspace; wasm32 checks of api and integrations; every migration applied in order to SQLite with an AI provider, a Sentry connection and Ask's route; the site's tsc -b --force, 927 tests and a production build; contracts' 48; chat's 76 and typecheck; the deploy tool's 92; the docs build (521 pages).3d ago

ask.test.tsAsk's card remembers what it did, step by step, on the answer that proposed it (#53 ): the card kept its progress only in the page that clicked it, so back in Chat the same answer offered Start session again, and a second click on the new-repository card opened a second issue and started a second session on the same empty repository (flagon-io/ask-hello-test got #1 and #2 , each with a session racing to push the first version). What a click does is now kept on the @g1t answer as it happens, in a proposal_state step beside the proposal: chat's new g1t_proposal merges it in (the viewer's own direct message with g1t only, known fields only) and tells the room, so every open copy of the answer updates. The card starts from it wherever the answer is read: Created with the repository, Live once the first version is pushed at with the address or Not deployed: with why, and Started with its links. Another click carries on from there rather than over: a session already started is shown, not started again; a repository already made, its deployments and the issue already opened are used; and while a click is still starting it (90 seconds at most), another says it's already starting. A failed step clears that hold so a retry works. Ask's steps also stop marking a read that found nothing as not read: checking that a new repository's name is free answers "not found", which showed as 1 not read on every new-repository answer.2d ago

ask.tsAsk's card remembers what it did, step by step, on the answer that proposed it (#53 ): the card kept its progress only in the page that clicked it, so back in Chat the same answer offered Start session again, and a second click on the new-repository card opened a second issue and started a second session on the same empty repository (flagon-io/ask-hello-test got #1 and #2 , each with a session racing to push the first version). What a click does is now kept on the @g1t answer as it happens, in a proposal_state step beside the proposal: chat's new g1t_proposal merges it in (the viewer's own direct message with g1t only, known fields only) and tells the room, so every open copy of the answer updates. The card starts from it wherever the answer is read: Created with the repository, Live once the first version is pushed at with the address or Not deployed: with why, and Started with its links. Another click carries on from there rather than over: a session already started is shown, not started again; a repository already made, its deployments and the issue already opened are used; and while a click is still starting it (90 seconds at most), another says it's already starting. A failed step clears that hold so a retry works. Ask's steps also stop marking a read that found nothing as not read: checking that a new repository's name is free answers "not found", which showed as 1 not read on every new-repository answer.2d ago

attempts.test.tsCards you act on in chat; agents comment and review as themselves; names shown cleanly; commits on the calendar1w ago

attempts.tsCards you act on in chat; agents comment and review as themselves; names shown cleanly; commits on the calendar1w ago

audit.server.tsAudit logs are kept by plan: a week on free, 90 days on the plan, and what staff set for an account in sudo1w ago

audit.test.tsThe audit log records a push refused at git's first request, tells deploy keys from workspace tokens, and the activity digest names a workspace as itself. git asks for the refs it may push to (GET info/refs?service=git-receive-pack) before it sends a pack, and that is where a credential that does not reach the repository or may not push is refused, so such a push left no entry: only the pack's POST was recorded. Repos now starts an entry for that first request too and records it only when it is refused (any 4xx or 5xx), as a denied git.push with the refusal's own message, so an allowed push is still one entry. A push allowed through is recorded under the rule for what pushed: person, workspace-token, deploy-key (new; a deploy key resolves as its workspace, and was recorded as workspace-token) or agent-token; the audit page labels deploy-key. The digest counted any wsp_ actor as g1t's agent (user:usr_g1t_agent), so a workspace token's pushes and changes were g1t's; it now keeps them the workspace's (user:wsp_<id>), which the site already never counts among the people. The audit log guide says what a refused push records, who the actor is and which rules a push carries. Checked: cargo test -p g1t-repos --lib (317, new: a workspace token's push is the workspace's with its token id and no person, a deploy key has its own rule, a push and one refused at its first request are recorded and an allowed first request is not, fetches are not), cargo test -p g1t-events --lib (55) and wasm32 checks of both; the site's audit and home tests (27) and typecheck; the docs build.3d ago

audit.tsThe audit log records a push refused at git's first request, tells deploy keys from workspace tokens, and the activity digest names a workspace as itself. git asks for the refs it may push to (GET info/refs?service=git-receive-pack) before it sends a pack, and that is where a credential that does not reach the repository or may not push is refused, so such a push left no entry: only the pack's POST was recorded. Repos now starts an entry for that first request too and records it only when it is refused (any 4xx or 5xx), as a denied git.push with the refusal's own message, so an allowed push is still one entry. A push allowed through is recorded under the rule for what pushed: person, workspace-token, deploy-key (new; a deploy key resolves as its workspace, and was recorded as workspace-token) or agent-token; the audit page labels deploy-key. The digest counted any wsp_ actor as g1t's agent (user:usr_g1t_agent), so a workspace token's pushes and changes were g1t's; it now keeps them the workspace's (user:wsp_<id>), which the site already never counts among the people. The audit log guide says what a refused push records, who the actor is and which rules a push carries. Checked: cargo test -p g1t-repos --lib (317, new: a workspace token's push is the workspace's with its token id and no person, a deploy key has its own rule, a push and one refused at its first request are recorded and an allowed first request is not, fetches are not), cargo test -p g1t-events --lib (55) and wasm32 checks of both; the site's audit and home tests (27) and typecheck; the docs build.3d ago

avatar-upload.test.tsWorkspace names and icons, and a component kit for every control1w ago

avatar-upload.tsWorkspace names and icons, and a component kit for every control1w ago

badge.test.tsMerge Actions runs: summaries, attempts and re-runs, graceful cancel, log downloads, badges (actions 0009)1w ago

badge.tsMerge Actions runs: summaries, attempts and re-runs, graceful cancel, log downloads, badges (actions 0009)1w ago

billing-return.test.tsBilling is ready to take real money, and stays on Stripe's test key until the switch. The g1t plan is $21 a month, all-in, with $10 of usage included: the card processing fee is part of the price, and no payment carries a fee line any more (old fee rows still read as they were). Card, billing address, tax IDs, invoices and cancelling are on Stripe's billing page (Manage billing), whose configuration billing keeps in step; Checkout asks for the billing address and a tax ID; g1t's Billing page keeps usage, the spend limit, budget alerts and prepaying. Members other than owners and billing managers no longer see billing details or invoices, from the page or the API.4d ago

billing-return.tsBilling is ready to take real money, and stays on Stripe's test key until the switch. The g1t plan is $21 a month, all-in, with $10 of usage included: the card processing fee is part of the price, and no payment carries a fee line any more (old fee rows still read as they were). Card, billing address, tax IDs, invoices and cancelling are on Stripe's billing page (Manage billing), whose configuration billing keeps in step; Checkout asks for the billing address and a tax ID; g1t's Billing page keeps usage, the spend limit, budget alerts and prepaying. Members other than owners and billing managers no longer see billing details or invoices, from the page or the API.4d ago

billing.test.tsBilling is ready to take real money, and stays on Stripe's test key until the switch. The g1t plan is $21 a month, all-in, with $10 of usage included: the card processing fee is part of the price, and no payment carries a fee line any more (old fee rows still read as they were). Card, billing address, tax IDs, invoices and cancelling are on Stripe's billing page (Manage billing), whose configuration billing keeps in step; Checkout asks for the billing address and a tax ID; g1t's Billing page keeps usage, the spend limit, budget alerts and prepaying. Members other than owners and billing managers no longer see billing details or invoices, from the page or the API.4d ago

billing.tsMerge main into project-descriptions2d ago

branch-redirect.server.tsComposer from the workspace's own repositories, and go get from g1t.sh1w ago

branches.server.tsMerge branch 'worktree-agent-a7823ad540eb3433b'4d ago

branches.test.tsThe Branches page does things to branches now. It has tabs (Overview, Yours, Active, Stale, All, kept in the address as ?tab=) and search; Overview shows the default branch, then the first five of your branches, the active ones and the stale ones, each linking to the rest. Each branch is a row with its name in a code chip with a copy button, who made its latest commit and when, that commit's checks, behind and ahead counts with bars drawn against the largest shown, and its pull request (the open one, else the newest merged or closed) with its state; on a phone the columns fold under the name. Yours is branches whose latest commit is the viewer's; stale is no commit in 90 days.5d ago

branches.tsThe Branches page does things to branches now. It has tabs (Overview, Yours, Active, Stale, All, kept in the address as ?tab=) and search; Overview shows the default branch, then the first five of your branches, the active ones and the stale ones, each linking to the rest. Each branch is a row with its name in a code chip with a copy button, who made its latest commit and when, that commit's checks, behind and ahead counts with bars drawn against the largest shown, and its pull request (the open one, else the newest merged or closed) with its state; on a phone the columns fold under the name. Yours is branches whose latest commit is the viewer's; stale is no commit in 90 days.5d ago

brand-icons.tsMerge branch 'worktree-agent-a11ea05ad7db44298'4d ago

busy.test.tsMerge branch 'main' into rm-agents4d ago

busy.tsMerge branch 'main' into rm-agents4d ago

cache.server.tsUsage, Billing settings and prepaid AI credit; fixes from the UX audit1w ago

card-actions.test.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

card-actions.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

chat-cache.test.tsMerge the workspace shell: navigation and phone shell, g1t as orchestrator, agents in roles with audience-checked reads, reactions and custom emoji, live notifications and browser push, the homepage tour (agents 0002, chat 0002)1w ago

chat-cache.tsMerge the workspace shell: navigation and phone shell, g1t as orchestrator, agents in roles with audience-checked reads, reactions and custom emoji, live notifications and browser push, the homepage tour (agents 0002, chat 0002)1w ago

chat-compose.test.tsThe top of every page is one surface: the breadcrumb, search, spend, Ask g1t, bell and + share one height, corners and spacing, Ask g1t is a plain tinted button with the g1t mark drawn on whole pixels, a page's own header sits directly under the bar at the same left edge, and every section, conversations and open artifacts included, draws the same rounded panel. Tables written in chat render as tables, and the site's file, checkbox, radio, number and date controls are the design system's. Button, Separator, Breadcrumb, Table and FileDrop join the shared components; the chat guide says how tables work.1w ago

chat-compose.tsThe top of every page is one surface: the breadcrumb, search, spend, Ask g1t, bell and + share one height, corners and spacing, Ask g1t is a plain tinted button with the g1t mark drawn on whole pixels, a page's own header sits directly under the bar at the same left edge, and every section, conversations and open artifacts included, draws the same rounded panel. Tables written in chat render as tables, and the site's file, checkbox, radio, number and date controls are the design system's. Button, Separator, Breadcrumb, Table and FileDrop join the shared components; the chat guide says how tables work.1w ago

chat-sounds.test.tsOne tab pings, once, and only for what is for you. Every open tab of g1t heard the same arrival, so each played its sound and showed its desktop toast; now a tab claims the arrival across the browser before pinging (lib/one-tab.ts: a Web Lock held while a ten-minute mark is written to the localStorage every tab shares, so a tab that comes second finds the mark), a tab in front claims at once and the others wait 400 ms, so the one you are looking at is the one that pings, and the sound and the toast for an arrival go together. A message heard over both the conversation's socket and the feed is claimed once, by its id. g1t's answer no longer pings as it starts: it arrives empty and thinking the moment a question is asked, which pinged right after your own message; it now pings once, when it is written. Where Web Locks or storage are missing, every tab pings as before. Checked: the site's typecheck and its 893 tests (new: the claim's rule and the answer's silence until it is written), and a production build whose chunks load first without error.4d ago

chat-sounds.tsOne tab pings, once, and only for what is for you. Every open tab of g1t heard the same arrival, so each played its sound and showed its desktop toast; now a tab claims the arrival across the browser before pinging (lib/one-tab.ts: a Web Lock held while a ten-minute mark is written to the localStorage every tab shares, so a tab that comes second finds the mark), a tab in front claims at once and the others wait 400 ms, so the one you are looking at is the one that pings, and the sound and the toast for an arrival go together. A message heard over both the conversation's socket and the feed is claimed once, by its id. g1t's answer no longer pings as it starts: it arrives empty and thinking the moment a question is asked, which pinged right after your own message; it now pings once, when it is written. Where Web Locks or storage are missing, every tab pings as before. Checked: the site's typecheck and its 893 tests (new: the claim's rule and the answer's silence until it is written), and a production build whose chunks load first without error.4d ago

chat.server.tsMerge main into project-descriptions2d ago

chat.test.tsMerge branch 'worktree-agent-ace9d6458cc8f30c0'2d ago

chat.tsMerge branch 'worktree-agent-ace9d6458cc8f30c0'2d ago

chrome.test.tsMerge branch 'worktree-agent-a4babc5b5815f6f3a'4d ago

chrome.tsMerge branch 'worktree-agent-a4babc5b5815f6f3a'4d ago

cn.tsAccount dropdown, llms.txt onboarding, hosted agent runner (not yet deployed)2w ago

comments.test.tsCards you act on in chat; agents comment and review as themselves; names shown cleanly; commits on the calendar1w ago

comments.tsCards you act on in chat; agents comment and review as themselves; names shown cleanly; commits on the calendar1w ago

commit-checks.server.tsMerge checks: statuses and check runs on every commit1w ago

commit-checks.test.tsMerge checks: statuses and check runs on every commit1w ago

commit-checks.tsMerge checks: statuses and check runs on every commit1w ago

commit-comments.tsMerge branch 'forking': repository forking, pull request merge methods and every repository setting for pull requests, commits, pushes and issues3d ago

commit-message.test.tsAn issue says what closed it, linked: on flagon-io/demo #1 the session that built the page closed the issue through the issue tool's close, and its timeline said only "syntaqx closed this as completed", with nothing to follow to the commit that did the work or the session that made it. close_issue (REST, the MCP issue tool's close, `g1t issue close --commit`) now takes an optional commit: one on the default branch (repos' log and branch_drift) whose message names the issue after a closing keyword (closing::references). The issue then closes as completed in that commit exactly as its push would (close_by_commit: resolved_commit, "closed this in <commit>", issue.closed with commit), by the caller; a commit not on the default branch, or one not naming the issue, is refused saying why, and an already-closed issue says what closed it. The work service's CloseIssueArgs flattens IssueActionArgs, so integrations and security keep calling close_issue as they do. When an issue is closed in a session, the note ends "via session ses_…" and issue.closed carries session: the session the closing commit names in a G1t-Session trailer (on a push too), or, for a close made with a token, the closer's own active session on the issue (attached or linked; the person it acts for when it acts for one). The timeline renders that id as a session chip linking /<owner>/<repo>/work/sessions/<id>, next to the existing commit chip and #N links (messageParts gains a session part). The pull requests guide's Closing issues, the sessions guide, webhooks, the API reference notes, openapi.json, the CLI reference and ops table say so. Deploy: work and api together, then the site; the CLI flag ships with the next CLI release. Checked: cargo test -p g1t-work (202), -p g1t-api (209, openapi regenerated), -p g1t-contracts, -p g1t (ops_table and cli_reference regenerated); wasm32 check of g1t-work and g1t-api; apps/web npm run typecheck; node --test app/lib/commit-message.test.ts.2d ago

commit-message.tsAn issue says what closed it, linked: on flagon-io/demo #1 the session that built the page closed the issue through the issue tool's close, and its timeline said only "syntaqx closed this as completed", with nothing to follow to the commit that did the work or the session that made it. close_issue (REST, the MCP issue tool's close, `g1t issue close --commit`) now takes an optional commit: one on the default branch (repos' log and branch_drift) whose message names the issue after a closing keyword (closing::references). The issue then closes as completed in that commit exactly as its push would (close_by_commit: resolved_commit, "closed this in <commit>", issue.closed with commit), by the caller; a commit not on the default branch, or one not naming the issue, is refused saying why, and an already-closed issue says what closed it. The work service's CloseIssueArgs flattens IssueActionArgs, so integrations and security keep calling close_issue as they do. When an issue is closed in a session, the note ends "via session ses_…" and issue.closed carries session: the session the closing commit names in a G1t-Session trailer (on a push too), or, for a close made with a token, the closer's own active session on the issue (attached or linked; the person it acts for when it acts for one). The timeline renders that id as a session chip linking /<owner>/<repo>/work/sessions/<id>, next to the existing commit chip and #N links (messageParts gains a session part). The pull requests guide's Closing issues, the sessions guide, webhooks, the API reference notes, openapi.json, the CLI reference and ops table say so. Deploy: work and api together, then the site; the CLI flag ships with the next CLI release. Checked: cargo test -p g1t-work (202), -p g1t-api (209, openapi regenerated), -p g1t-contracts, -p g1t (ops_table and cli_reference regenerated); wasm32 check of g1t-work and g1t-api; apps/web npm run typecheck; node --test app/lib/commit-message.test.ts.2d ago

commit-people.server.tsMerge commit authors and hovercards: commits show their accounts, every name opens a card (repos 0016)1w ago

commit-people.test.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

commit-people.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

compute.server.tsInvite-only launch: sign in with GitHub, repository access and lifecycle, many emails, a new look1w ago

compute.test.tsBilling is ready to take real money, and stays on Stripe's test key until the switch. The g1t plan is $21 a month, all-in, with $10 of usage included: the card processing fee is part of the price, and no payment carries a fee line any more (old fee rows still read as they were). Card, billing address, tax IDs, invoices and cancelling are on Stripe's billing page (Manage billing), whose configuration billing keeps in step; Checkout asks for the billing address and a tax ID; g1t's Billing page keeps usage, the spend limit, budget alerts and prepaying. Members other than owners and billing managers no longer see billing details or invoices, from the page or the API.4d ago

compute.tsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

confirm-gate.test.tsMerge workspace invitations: nobody joins a workspace without saying yes, people are found by username, your own invites can bring someone in, and nobody is left without a workspace (identity 0040)1w ago

confirm-gate.tsMerge workspace invitations: nobody joins a workspace without saying yes, people are found by username, your own invites can bring someone in, and nobody is left without a workspace (identity 0040)1w ago

connected.server.tsA workspace's AI is its own setting, not an integration. AI providers and what Ask answers with were connections like Sentry's or Linear's: rows of `connections`, cards in the Integrations directory, `provider` values of the integration operations. Integrations are becoming a directory others publish to; a workspace's AI is a base setting g1t looks after itself. Contracts: `g1t_contracts::ai` (and packages/contracts ai.ts) has AiProvider (the 14), AiGroup (labs, platforms, self_hosted), AiConfig, AiConnection and their methods' arguments, and the Ask types moved there; AskSetup names the provider by `provider_id` and `provider_name` (were connection_id, connection_name). `integrations::Provider` is sentry, datadog, webhook, jira and linear; ProviderKind loses Models, ConnectionConfig its base_url, auth_header and model, Connection its models. Storage: integrations migration 0011 makes `ai_providers` and `ai_routes`, moves the AI rows out of `connections` with their ids (sealed keys are bound to the id, and managed_agents and agent_runs keep finding them), keeps Ask's route and drops `model_routes`; renaming and deleting a workspace cover both tables. The service keeps them in src/ai.rs with methods of their own (ai_list, ai_add, ai_update, ai_remove, ai_test, ask_settings, set_ask_settings, ask); a provider's key is checked when added and on each test; removing one turns Ask off when it answered with it and forgets its managed agent. Sessions on issues find the workspace's Anthropic provider there. API: list_ai_providers, add_ai_provider, update_ai_provider, remove_ai_provider, test_ai_provider (/workspaces/{workspace}/ai/providers[/{id}[/test]]), get_ask_settings and set_ask_settings (/workspaces/{workspace}/ai/ask), under the OpenAPI tag AI; reading takes workspace:read, changing workspace:admin and an owner's own token; the integration operations take only integration providers. MCP: a tool `ai` (list_providers, add_provider, update_provider, remove_provider, test_provider, get_ask, set_ask). The CLI's operation table and command reference are regenerated. Site: AI at /<workspace>/-/ai (the sidebar's AI → Models, the command palette's AI models) holds Ask g1t, the providers and Add a provider, grouped Frontier labs, Model platforms, Self-hosted or your own gateway; AI providers are no longer in the Integrations directory or Apps, and the integrations pages are alerts and trackers only; -/integrations/models and ?add=<an AI provider> lead to -/ai. Ask's settings link, and @g1t's "not set up yet" in Chat, point there. Docs: the AI providers guide (with its API), and the integrations, Ask, sessions, CLI, plugin and MCP pages say so. Deploy the integrations migration first, then integrations, then api, chat and web. Checked: cargo test for g1t-contracts (309), g1t-integrations (70), g1t-api (204, the OpenAPI document regenerated) and the CLI (105 and 44, the command reference current); cargo check --workspace; wasm32 checks of api and integrations; every migration applied in order to SQLite with an AI provider, a Sentry connection and Ask's route; the site's tsc -b --force, 927 tests and a production build; contracts' 48; chat's 76 and typecheck; the deploy tool's 92; the docs build (521 pages).3d ago

connectors.test.tsA workspace's AI is its own setting, not an integration. AI providers and what Ask answers with were connections like Sentry's or Linear's: rows of `connections`, cards in the Integrations directory, `provider` values of the integration operations. Integrations are becoming a directory others publish to; a workspace's AI is a base setting g1t looks after itself. Contracts: `g1t_contracts::ai` (and packages/contracts ai.ts) has AiProvider (the 14), AiGroup (labs, platforms, self_hosted), AiConfig, AiConnection and their methods' arguments, and the Ask types moved there; AskSetup names the provider by `provider_id` and `provider_name` (were connection_id, connection_name). `integrations::Provider` is sentry, datadog, webhook, jira and linear; ProviderKind loses Models, ConnectionConfig its base_url, auth_header and model, Connection its models. Storage: integrations migration 0011 makes `ai_providers` and `ai_routes`, moves the AI rows out of `connections` with their ids (sealed keys are bound to the id, and managed_agents and agent_runs keep finding them), keeps Ask's route and drops `model_routes`; renaming and deleting a workspace cover both tables. The service keeps them in src/ai.rs with methods of their own (ai_list, ai_add, ai_update, ai_remove, ai_test, ask_settings, set_ask_settings, ask); a provider's key is checked when added and on each test; removing one turns Ask off when it answered with it and forgets its managed agent. Sessions on issues find the workspace's Anthropic provider there. API: list_ai_providers, add_ai_provider, update_ai_provider, remove_ai_provider, test_ai_provider (/workspaces/{workspace}/ai/providers[/{id}[/test]]), get_ask_settings and set_ask_settings (/workspaces/{workspace}/ai/ask), under the OpenAPI tag AI; reading takes workspace:read, changing workspace:admin and an owner's own token; the integration operations take only integration providers. MCP: a tool `ai` (list_providers, add_provider, update_provider, remove_provider, test_provider, get_ask, set_ask). The CLI's operation table and command reference are regenerated. Site: AI at /<workspace>/-/ai (the sidebar's AI → Models, the command palette's AI models) holds Ask g1t, the providers and Add a provider, grouped Frontier labs, Model platforms, Self-hosted or your own gateway; AI providers are no longer in the Integrations directory or Apps, and the integrations pages are alerts and trackers only; -/integrations/models and ?add=<an AI provider> lead to -/ai. Ask's settings link, and @g1t's "not set up yet" in Chat, point there. Docs: the AI providers guide (with its API), and the integrations, Ask, sessions, CLI, plugin and MCP pages say so. Deploy the integrations migration first, then integrations, then api, chat and web. Checked: cargo test for g1t-contracts (309), g1t-integrations (70), g1t-api (204, the OpenAPI document regenerated) and the CLI (105 and 44, the command reference current); cargo check --workspace; wasm32 checks of api and integrations; every migration applied in order to SQLite with an AI provider, a Sentry connection and Ask's route; the site's tsc -b --force, 927 tests and a production build; contracts' 48; chat's 76 and typecheck; the deploy tool's 92; the docs build (521 pages).3d ago

connectors.tsMerge branch 'worktree-agent-a11ea05ad7db44298'4d ago

content-cache.test.tsMerge the site's CPU: code highlighted once per data centre by content, markdown parsed once per text, and the request handler built once per isolate1w ago

content-cache.tsMerge the site's CPU: code highlighted once per data centre by content, markdown parsed once per text, and the request handler built once per isolate1w ago

content-safety.test.tsMerge g1tusercontent.com: registry answers run nothing in a browser, the site's pages run only their own scripts, repository files and avatars on their own origin, raw files rate limited per address1w ago

content-safety.tsMerge g1tusercontent.com: registry answers run nothing in a browser, the site's pages run only their own scripts, repository files and avatars on their own origin, raw files rate limited per address1w ago

contribution-calendar.test.tsCards you act on in chat; agents comment and review as themselves; names shown cleanly; commits on the calendar1w ago

contribution-calendar.tsCards you act on in chat; agents comment and review as themselves; names shown cleanly; commits on the calendar1w ago

coordination.test.tsLive now lists the sessions working now: on a repository's overview and Work, a session not heard from for 15 minutes (idle) is folded into one row under the list, "N idle sessions", last seen when, which opens to show them; each moves back up when it is heard from and drops off when it times out. A session that has done nothing since it started (no title, issue or pull request, never heard from again: a Claude Code tab opened and not typed in) is not listed at all (startedOnly). The panel's counts are of the sessions working now, with "N idle" beside them, and its icon is green only for those. Sessions on the default branch are headed "On main" (the branch's name), not "No branch yet", which is kept for sessions git has named no branch for, and their rows leave the branch out. An untitled session's title, read from its first prompt, skips what VS Code puts ahead of what was typed (<ide_opened_file>, <ide_selection>): such a prompt read as a turn Claude Code started itself, so a session started from VS Code took its second prompt as its title. Checked: apps/web tsc -b; npm test (986; new: awake and idle split, startedOnly, an IDE-prefixed prompt's intent).2d ago

coordination.tsLive now lists the sessions working now: on a repository's overview and Work, a session not heard from for 15 minutes (idle) is folded into one row under the list, "N idle sessions", last seen when, which opens to show them; each moves back up when it is heard from and drops off when it times out. A session that has done nothing since it started (no title, issue or pull request, never heard from again: a Claude Code tab opened and not typed in) is not listed at all (startedOnly). The panel's counts are of the sessions working now, with "N idle" beside them, and its icon is green only for those. Sessions on the default branch are headed "On main" (the branch's name), not "No branch yet", which is kept for sessions git has named no branch for, and their rows leave the branch out. An untitled session's title, read from its first prompt, skips what VS Code puts ahead of what was typed (<ide_opened_file>, <ide_selection>): such a prompt read as a turn Claude Code started itself, so a session started from VS Code took its second prompt as its title. Checked: apps/web tsc -b; npm test (986; new: awake and idle split, startedOnly, an IDE-prefixed prompt's intent).2d ago

current-workspace.server.tsNew project and Import from GitHub work in the workspace you're in, as the sidebar shows it, with no workspace to pick on the page; a link that names another of your workspaces switches you to it first. The projects guide says so.1w ago

deadline.test.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

deadline.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

dependency-updates.test.tsTeams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar1w ago

dependency-updates.tsTeams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar1w ago

deploy-kind.test.tsMerge GitHub pull request #8 (claude/cool-newton-7nm8r9): Docker deployments on g1t.page behind a switch that stays off until Cloudflare's container APIs are checked live, workflows a workspace manages in its repositories, tarball and zipball downloads, refs/pull/* kept to g1t alone, framework detection for g1t.page builds, and a workflow's event payload matching GitHub's.5d ago

deploy-kind.tsMerge branch 'worktree-agent-ae1b1f625cfdcab1b'4d ago

deployments.server.tsMerge branch 'main' into worktree-agent-a69aeabc4b0deeb971w ago

deployments.test.tsMerge branch 'site-fixes' into fast-push1w ago

deployments.tsAn app that deploys itself says so, instead of "Deployed elsewhere". A project whose own workflows (or another pipeline) deploy it, as g1t's do, read as if g1t had nothing to do with it: grey "Deployed elsewhere" in the workspace's Projects, "App, deployed elsewhere" on its badge, and the API said `runs` is `g1t` "when g1t deploys it". What tells the two apart is who hosts it, so they are now named that way: App or site, hosted on g1t.page (badge App on g1t.page) and App or site that deploys itself (badge App, deploys itself), in settings, the badge's menu, the overview's "Where does it run?" (It deploys itself), the Deployments pages, the guides and the API's descriptions; `runs` keeps its values. In Projects an app that deploys itself shows the rocket, Deploys itself, and the Deploys filter counts it. Its production card says which workflow run deployed it when g1t Actions did ("deployed 16m ago by its workflow"), else "by its own pipeline". An app nobody has placed that deploys production itself (a successful production deployment from a workflow or the API, not g1t.page's) now deploys itself everywhere, where before only its overview inferred it: the projects service keeps that it has (migration 0013, `reports_production`), and Deployments being on or a setting still win. Checked: the site's typecheck (tsc -b --force), its 911 tests and a production build; the projects service's typecheck and 49 tests (new: an app that deploys production itself); the API's openapi tests with the docs copy regenerated; the deploy tool's 92 tests.3d ago

desktop-runner.test.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

desktop-runner.tsMerge branch 'main' into rm-agents4d ago

desktop-settings.test.tsMerge branch 'worktree-agent-a6b141eb3a7e9cf01'4d ago

desktop-settings.tsMerge branch 'worktree-agent-a6b141eb3a7e9cf01'4d ago

desktop.test.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

desktop.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

developer-mode.test.tsDeveloper mode: a page that fails can say why. Under Settings, Account, a Developer mode switch (kept per browser in a g1t_dev cookie, like Appearance) makes the error page show the error itself, its message and stack, the address, the time, whether it came from a page the server drew or a navigation in the browser, the routes it was drawn from, the site's build (its stylesheet's hashed name), the browser and the desktop app's version, with Copy details to paste it all into a report; with it off, the page says where to turn it on. A signed-in request with developer mode on is answered by React Router in its development mode (workers/app.ts, a second handler made the first time it is wanted), so an error on the server comes through in full instead of "Unexpected Server Error"; that mode differs from production only in that and in hooks only Vite's dev server sets, and only ever covers the person's own requests. Checked in a production build: a failing repository page shows "counts failed with status 503" and its stack with developer mode on, and the usual page with it off. The workspaces guide covers it under Developer mode.4d ago

developer-mode.tsDeveloper mode: a page that fails can say why. Under Settings, Account, a Developer mode switch (kept per browser in a g1t_dev cookie, like Appearance) makes the error page show the error itself, its message and stack, the address, the time, whether it came from a page the server drew or a navigation in the browser, the routes it was drawn from, the site's build (its stylesheet's hashed name), the browser and the desktop app's version, with Copy details to paste it all into a report; with it off, the page says where to turn it on. A signed-in request with developer mode on is answered by React Router in its development mode (workers/app.ts, a second handler made the first time it is wanted), so an error on the server comes through in full instead of "Unexpected Server Error"; that mode differs from production only in that and in hooks only Vite's dev server sets, and only ever covers the person's own requests. Checked in a production build: a failing repository page shows "counts failed with status 503" and its stack with developer mode on, and the usual page with it off. The workspaces guide covers it under Developer mode.4d ago

diff-files.test.tsMerge branch 'worktree-agent-a432c4ed9deb834b1'4d ago

diff-files.tsMerge branch 'worktree-agent-a432c4ed9deb834b1'4d ago

diff.test.tsMerge branch 'worktree-agent-a432c4ed9deb834b1'4d ago

diff.tsMerge branch 'worktree-agent-a432c4ed9deb834b1'4d ago

dock.server.tsThe apps you pin to your dock are kept with your account, per workspace and in your order, so the dock is the same on every device: identity keeps them in dock_pins and answers dock_pins and set_dock_pins, the dock reads them with the rest of the page, pins kept only on this device carry over with your first change, this device's copy still draws the dock when identity can't be reached, a pin that can't be saved says so, and they go when you or the workspace do; the workspaces guide says how.1w ago

download-platform.test.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

download-platform.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

emails.server.tsstatus.g1t.sh with incident management, invites that land you in the workspace, settings as pages, usage without quotas1w ago

emails.test.tsInvite-only launch: sign in with GitHub, repository access and lifecycle, many emails, a new look1w ago

emails.tsInvite-only launch: sign in with GitHub, repository access and lifecycle, many emails, a new look1w ago

emoji.test.tsMerge the workspace shell: navigation and phone shell, g1t as orchestrator, agents in roles with audience-checked reads, reactions and custom emoji, live notifications and browser push, the homepage tour (agents 0002, chat 0002)1w ago

emoji.tsMerge the workspace shell: navigation and phone shell, g1t as orchestrator, agents in roles with audience-checked reads, reactions and custom emoji, live notifications and browser push, the homepage tour (agents 0002, chat 0002)1w ago

environments.test.tsMerge branch 'worktree-agent-a3abfcce648e87dca'1w ago

environments.tsMerge branch 'worktree-agent-a3abfcce648e87dca'1w ago

folios.test.tsArtifacts is a mode in the rail, the shell and the phone's tab bar in place of Docs: its home lists everything you can open by day or as cards under All, Yours and Shared with you, with search, filters and tiles to start a doc (slides, designs and dashboards say they are coming soon), and its sidebar holds favorites, spaces, Private, Shared, projects' docs and the trash.1w ago

folios.tsThe artifacts service is services/artifacts, the Worker g1t-artifacts, bound as ARTIFACTS by the API, the site and the agents; its live rooms move to it with a Durable Object transfer from g1t-docs-service, and its database, bucket, indexes and queue keep their names. The git store's binding and settings are GITSTORE, its ops scripts gitstore-*, and workflow run artifacts keep their compatible API under run_artifacts modules. The deploy tool puts a Worker that has never deployed before the Workers in its stage that bind to it, and the deploy guide gives the cutover runbook.1w ago

forks.server.tsMerge branch 'forking': repository forking, pull request merge methods and every repository setting for pull requests, commits, pushes and issues3d ago

front-door-limits.test.tsMerge GitHub pull request #8 (claude/cool-newton-7nm8r9): Docker deployments on g1t.page behind a switch that stays off until Cloudflare's container APIs are checked live, workflows a workspace manages in its repositories, tarball and zipball downloads, refs/pull/* kept to g1t alone, framework detection for g1t.page builds, and a workflow's event payload matching GitHub's.5d ago

front-door-limits.tsMerge GitHub pull request #8 (claude/cool-newton-7nm8r9): Docker deployments on g1t.page behind a switch that stays off until Cloudflare's container APIs are checked live, workflows a workspace manages in its repositories, tarball and zipball downloads, refs/pull/* kept to g1t alone, framework detection for g1t.page builds, and a workflow's event payload matching GitHub's.5d ago

g1t-chat.test.tsAsk names people by their g1t username, knows what shipped means, and picks its model by itself. Asked "What shipped today?", Ask made "@chasepierce" out of a commit's git author name, corrected itself mid-answer, and said nothing had reached production though the repository's own Deploy run had: commit listings carried only the git name, the prompt never said which field is a username, and activity counted only g1t.page deployments as production. Commits from list_commits now say `author: { name, username }`, the username from identity's address owners (a confirmed or noreply address; g1t's own are g1t), never the address, and activity's landed commits carry `username` too; the prompt mentions only usernames tools returned, never turns a git name or address into one, and answers once, without "wait" or corrections. Shipping, to Ask, is reaching the default branch and then production: a successful run of a repository's deploy workflow on its default branch (a workflow whose name or file says deploy, not preview) now counts as a production deployment beside g1t.page's, each landed commit says whether it is live (and by which deploy), deploying, failed or not deployed (a deploy carries everything before it on the branch), summary.shipped counts them, and list_runs takes since and until; releases and tags are one more signal, and today is the start of your day in your time zone. Ask's model is Auto by default (migration 0010 turns it on for every Ask route nobody chose a model for: the old picker started on the provider's first model, Haiku for Anthropic). Auto sends each question to a model of the workspace's provider without asking another model: a short lookup to a fast one, a question about why, how, what shipped, changes, failures, a time range, or a long conversation to a standard one; tiers come from the models' names, so it works for every provider, the largest only when nothing else is there, and an Azure deployment or an endpoint without a list keeps its own model. The answer says which ("Anthropic · Auto · claude-sonnet-5-5"); an explicit pick still wins. The Ask guide has Auto and what Ask means by shipped. Checked on this change alone (a worktree of the commit before it): cargo test for g1t-integrations (70, new: Auto's routing for each provider, prompt rules), g1t-api (204, new: commit authors with and without an account, shipped states), g1t-actions-service (108), g1t-contracts (306) and the CLI (149); cargo check --workspace; wasm32 builds of api, integrations and actions; the site's typecheck (tsc -b --force), its Ask tests and a production build; chat's tests and typecheck; the deploy tool's tests.3d ago

g1t-chat.tsAsk names people by their g1t username, knows what shipped means, and picks its model by itself. Asked "What shipped today?", Ask made "@chasepierce" out of a commit's git author name, corrected itself mid-answer, and said nothing had reached production though the repository's own Deploy run had: commit listings carried only the git name, the prompt never said which field is a username, and activity counted only g1t.page deployments as production. Commits from list_commits now say `author: { name, username }`, the username from identity's address owners (a confirmed or noreply address; g1t's own are g1t), never the address, and activity's landed commits carry `username` too; the prompt mentions only usernames tools returned, never turns a git name or address into one, and answers once, without "wait" or corrections. Shipping, to Ask, is reaching the default branch and then production: a successful run of a repository's deploy workflow on its default branch (a workflow whose name or file says deploy, not preview) now counts as a production deployment beside g1t.page's, each landed commit says whether it is live (and by which deploy), deploying, failed or not deployed (a deploy carries everything before it on the branch), summary.shipped counts them, and list_runs takes since and until; releases and tags are one more signal, and today is the start of your day in your time zone. Ask's model is Auto by default (migration 0010 turns it on for every Ask route nobody chose a model for: the old picker started on the provider's first model, Haiku for Anthropic). Auto sends each question to a model of the workspace's provider without asking another model: a short lookup to a fast one, a question about why, how, what shipped, changes, failures, a time range, or a long conversation to a standard one; tiers come from the models' names, so it works for every provider, the largest only when nothing else is there, and an Azure deployment or an endpoint without a list keeps its own model. The answer says which ("Anthropic · Auto · claude-sonnet-5-5"); an explicit pick still wins. The Ask guide has Auto and what Ask means by shipped. Checked on this change alone (a worktree of the commit before it): cargo test for g1t-integrations (70, new: Auto's routing for each provider, prompt rules), g1t-api (204, new: commit authors with and without an account, shipped states), g1t-actions-service (108), g1t-contracts (306) and the CLI (149); cargo check --workspace; wasm32 builds of api, integrations and actions; the site's typecheck (tsc -b --force), its Ask tests and a production build; chat's tests and typecheck; the deploy tool's tests.3d ago

github.server.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

github.test.tsConnecting GitHub to a workspace is one click and one trip to GitHub. A workspace owner's Connect GitHub on Import from GitHub links their GitHub account on the way if it is not linked yet and comes straight back to carry on: an installation they already have on GitHub and the workspace lacks is added rather than installed again (with several, they choose), and only with none are they sent to install the app; GitHub then returns them to the import page with the account connected. Before, linking and installing were separate chores (link, come back, install), Settings -> GitHub stopped at "Linked" with nowhere to go, and an app already on the account was "installed" again through GitHub's settings page. When GitHub returns after an installation that did not start from g1t, an unlinked account is linked on the way instead of showing an error with a button, and someone who owns one workspace has it added there without being asked which. Choose repositories on GitHub opens in the same tab, since GitHub brings you back after saving; Add a GitHub account still always installs on another account; Settings -> GitHub points to Import from GitHub once linked. A link that does not take shows the page instead of going round again. The GitHub guide describes the one-step flow. (The app's Setup URL was empty on GitHub, which left people there after installing; it is now set to /integrations/github/setup.) Checked: the site's typecheck (tsc -b --force), its 911 tests (new: what Connect GitHub does next), a production build and the deploy tool's 92 tests.3d ago

github.tsConnecting GitHub to a workspace is one click and one trip to GitHub. A workspace owner's Connect GitHub on Import from GitHub links their GitHub account on the way if it is not linked yet and comes straight back to carry on: an installation they already have on GitHub and the workspace lacks is added rather than installed again (with several, they choose), and only with none are they sent to install the app; GitHub then returns them to the import page with the account connected. Before, linking and installing were separate chores (link, come back, install), Settings -> GitHub stopped at "Linked" with nowhere to go, and an app already on the account was "installed" again through GitHub's settings page. When GitHub returns after an installation that did not start from g1t, an unlinked account is linked on the way instead of showing an error with a button, and someone who owns one workspace has it added there without being asked which. Choose repositories on GitHub opens in the same tab, since GitHub brings you back after saving; Add a GitHub account still always installs on another account; Settings -> GitHub points to Import from GitHub once linked. A link that does not take shows the page instead of going round again. The GitHub guide describes the one-step flow. (The app's Setup URL was empty on GitHub, which left people there after installing; it is now set to /integrations/github/setup.) Checked: the site's typecheck (tsc -b --force), its 911 tests (new: what Connect GitHub does next), a production build and the deploy tool's 92 tests.3d ago

go-get.test.tsComposer from the workspace's own repositories, and go get from g1t.sh1w ago

go-get.tsComposer from the workspace's own repositories, and go get from g1t.sh1w ago

guardrails.test.tsMerge branch 'worktree-agent-ac950b9c891220ad3'4d ago

guardrails.tsMerge branch 'worktree-agent-ac950b9c891220ad3'4d ago

highlight.server.tsMerge branch 'worktree-agent-a432c4ed9deb834b1'4d ago

home-live.test.tsMerge branch 'main' into worktree-agent-a654d8ba46392bedd2d ago

home-live.tsMerge branch 'main' into worktree-agent-a654d8ba46392bedd2d ago

home.server.tsMerge remote-tracking branch 'g1t/main' into HEAD2d ago

home.test.tsHome reads like a day's start (#54 ): what needs you first with the one to start on at the top, then one card for what happened, and what is running, what it cost and what settled by itself in a side column that stays in view.2d ago

home.tsHome reads like a day's start (#54 ): what needs you first with the one to start on at the top, then one card for what happened, and what is running, what it cost and what settled by itself in a side column that stays in view.2d ago

hovercard.test.tsMerge leftovers: plan activity filtered in the query, pushes counted by account, ghost during the deletion window, profile time zones (identity 0039)1w ago

hovercard.tsOne kind of access token; presence and status; usernames keep their case; the tour is a miniature of the real app; icons for password managers1w ago

immutable.server.tsAgents as a team: lifecycle, merge queue, billing and a new shell2w ago

inbox.test.tsMerge remote-tracking branch 'g1t/main' into HEAD2d ago

inbox.tsMerge remote-tracking branch 'g1t/main' into HEAD2d ago

inline-markdown.test.tsUsage, Billing settings and prepaid AI credit; fixes from the UX audit1w ago

inline-markdown.tsMerge branch 'worktree-agent-a4babc5b5815f6f3a'4d ago

integration-sections.tsA workspace's AI is its own setting, not an integration. AI providers and what Ask answers with were connections like Sentry's or Linear's: rows of `connections`, cards in the Integrations directory, `provider` values of the integration operations. Integrations are becoming a directory others publish to; a workspace's AI is a base setting g1t looks after itself. Contracts: `g1t_contracts::ai` (and packages/contracts ai.ts) has AiProvider (the 14), AiGroup (labs, platforms, self_hosted), AiConfig, AiConnection and their methods' arguments, and the Ask types moved there; AskSetup names the provider by `provider_id` and `provider_name` (were connection_id, connection_name). `integrations::Provider` is sentry, datadog, webhook, jira and linear; ProviderKind loses Models, ConnectionConfig its base_url, auth_header and model, Connection its models. Storage: integrations migration 0011 makes `ai_providers` and `ai_routes`, moves the AI rows out of `connections` with their ids (sealed keys are bound to the id, and managed_agents and agent_runs keep finding them), keeps Ask's route and drops `model_routes`; renaming and deleting a workspace cover both tables. The service keeps them in src/ai.rs with methods of their own (ai_list, ai_add, ai_update, ai_remove, ai_test, ask_settings, set_ask_settings, ask); a provider's key is checked when added and on each test; removing one turns Ask off when it answered with it and forgets its managed agent. Sessions on issues find the workspace's Anthropic provider there. API: list_ai_providers, add_ai_provider, update_ai_provider, remove_ai_provider, test_ai_provider (/workspaces/{workspace}/ai/providers[/{id}[/test]]), get_ask_settings and set_ask_settings (/workspaces/{workspace}/ai/ask), under the OpenAPI tag AI; reading takes workspace:read, changing workspace:admin and an owner's own token; the integration operations take only integration providers. MCP: a tool `ai` (list_providers, add_provider, update_provider, remove_provider, test_provider, get_ask, set_ask). The CLI's operation table and command reference are regenerated. Site: AI at /<workspace>/-/ai (the sidebar's AI → Models, the command palette's AI models) holds Ask g1t, the providers and Add a provider, grouped Frontier labs, Model platforms, Self-hosted or your own gateway; AI providers are no longer in the Integrations directory or Apps, and the integrations pages are alerts and trackers only; -/integrations/models and ?add=<an AI provider> lead to -/ai. Ask's settings link, and @g1t's "not set up yet" in Chat, point there. Docs: the AI providers guide (with its API), and the integrations, Ask, sessions, CLI, plugin and MCP pages say so. Deploy the integrations migration first, then integrations, then api, chat and web. Checked: cargo test for g1t-contracts (309), g1t-integrations (70), g1t-api (204, the OpenAPI document regenerated) and the CLI (105 and 44, the command reference current); cargo check --workspace; wasm32 checks of api and integrations; every migration applied in order to SQLite with an AI provider, a Sentry connection and Ask's route; the site's tsc -b --force, 927 tests and a production build; contracts' 48; chat's 76 and typecheck; the deploy tool's 92; the docs build (521 pages).3d ago

invitations.server.tsMerge workspace invitations: nobody joins a workspace without saying yes, people are found by username, your own invites can bring someone in, and nobody is left without a workspace (identity 0040)1w ago

invitations.tsMerge workspace invitations: nobody joins a workspace without saying yes, people are found by username, your own invites can bring someone in, and nobody is left without a workspace (identity 0040)1w ago

invites.server.tsMerge branch 'worktree-agent-a6da31375506fc440'4d ago

invites.test.tsMerge branch 'worktree-agent-a6da31375506fc440'4d ago

invites.tsMerge branch 'worktree-agent-a6da31375506fc440'4d ago

labels.test.tsTeams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar1w ago

labels.tsTeams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar1w ago

last-commits.server.tsMerge last commits over the whole history: progress kept per ref and path, a push reads only its own commits1w ago

legal.test.tsInvite-only launch: sign in with GitHub, repository access and lifecycle, many emails, a new look1w ago

legal.tsThe footer and the landing page name Flagon, Inc. and stop there: "g1t is designed, built, and backed by Flagon, Inc." and "g1t is made by Flagon, Inc.", without "a small, independent software company founded by Chase Pierce". COMPANY.about is gone; COMPANY.founder stays for the structured data search engines read. Checked: apps/web tsc -b.2d ago

line-anchor.test.tsFast pages, required checks on the branch, self-hosted runners, honest incidents1w ago

line-anchor.tsFast pages, required checks on the branch, self-hosted runners, honest incidents1w ago

live-socket.tsA doc opens at once: the page carries the document's saved state, so the editor mounts without waiting for its live room, which connects in the background and sends only what changed, with a quiet dot that turns green when it's synced; edits made before then are kept and sent once. The editor's code is preloaded from the page and fetched early when you hover a doc, the app's shared code ships in a few chunks instead of a hundred small ones, socket tickets come with the page instead of a round trip, and the artifacts service answers a page, a sidebar and a room in parallel lookups instead of a dozen in a row, reporting its database round trips in Server-Timing. The artifacts guide says how a doc opens, and CONTRIBUTING's Speed section has the numbers.1w ago

log-download.server.tsMerge g1tusercontent.com: registry answers run nothing in a browser, the site's pages run only their own scripts, repository files and avatars on their own origin, raw files rate limited per address1w ago

log-follow.test.tsMerge the run page fixes: Cancel run shows on every view of a run that is still going, including an earlier attempt's and job links, and pages refresh when a run is re-run from anywhere; a job's log follows to its true end, above anything covering the bottom of the window, and Jump to latest lands there.5d ago

log-follow.tsMerge the run page fixes: Cancel run shows on every view of a run that is still going, including an earlier attempt's and job links, and pages refresh when a run is re-run from anywhere; a job's log follows to its true end, above anything covering the bottom of the window, and Jump to latest lands there.5d ago

log-lines.test.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9'5d ago

log-lines.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9'5d ago

log-poll.test.tsA running job's log is live, as on GitHub. A step's output reaches its run's page within about two seconds: the runner sends what is printed every second, and the page asks every second while a step runs, every three while none does; an answer that fails (a 429, a 5xx, the network) is asked again after a wait that doubles to half a minute, or as long as Retry-After says, where before one such answer stopped the log for good and the step said "No output yet." until the page was reloaded, and a signed-out viewer's run page, refreshing beside its log, ran into the 30-a-minute limit within a minute; the live log now counts as a page, and only log downloads as costly. A hidden tab stops asking and catches up when shown, and the page says when it is reconnecting. The running step opens by itself, its timer counts, its log follows the newest line while its end is on screen and offers Jump to latest once the viewer scrolls up, a step that has printed nothing for a while says "Running · no output for 2m 10s", and the job's header names the step it is on. Colour that tools print (bold, dim, italic, underline, the 16 colours as theme tokens for light and dark, 256 colours and 24-bit colour, text and background) is drawn as printed, other escape codes are dropped and links keep their text, and search reads the plain text and keeps the colours around what it marks. Every line is stamped with the time it was printed, as GitHub's raw logs are, so log.txt and the run's zip have a time at the start of each line, and a clock button beside Search logs shows the times, remembered by the browser. Set up job now gives the runner's version and name, the operating system, the runner image, the machine's architecture, CPUs and memory, the token's permissions under G1T_TOKEN Permissions, the secret source and the actions it downloads; each run step opens with GitHub's folded Run group, the script in cyan, the shell's whole path and the step's env, each uses step with its with inputs (an action's defaults among them) and env; cache and artifact transfers say how far they have got every five seconds; and a last Complete job step sets the job's outputs and stops what the steps left running, found by RUNNER_TRACKING_ID as GitHub's runner finds it. Two batches of a job's log written at once can no longer reach for the same seq and lose one. The Actions guide and the rate limits page say so.5d ago

log-poll.tsA running job's log is live, as on GitHub. A step's output reaches its run's page within about two seconds: the runner sends what is printed every second, and the page asks every second while a step runs, every three while none does; an answer that fails (a 429, a 5xx, the network) is asked again after a wait that doubles to half a minute, or as long as Retry-After says, where before one such answer stopped the log for good and the step said "No output yet." until the page was reloaded, and a signed-out viewer's run page, refreshing beside its log, ran into the 30-a-minute limit within a minute; the live log now counts as a page, and only log downloads as costly. A hidden tab stops asking and catches up when shown, and the page says when it is reconnecting. The running step opens by itself, its timer counts, its log follows the newest line while its end is on screen and offers Jump to latest once the viewer scrolls up, a step that has printed nothing for a while says "Running · no output for 2m 10s", and the job's header names the step it is on. Colour that tools print (bold, dim, italic, underline, the 16 colours as theme tokens for light and dark, 256 colours and 24-bit colour, text and background) is drawn as printed, other escape codes are dropped and links keep their text, and search reads the plain text and keeps the colours around what it marks. Every line is stamped with the time it was printed, as GitHub's raw logs are, so log.txt and the run's zip have a time at the start of each line, and a clock button beside Search logs shows the times, remembered by the browser. Set up job now gives the runner's version and name, the operating system, the runner image, the machine's architecture, CPUs and memory, the token's permissions under G1T_TOKEN Permissions, the secret source and the actions it downloads; each run step opens with GitHub's folded Run group, the script in cyan, the shell's whole path and the step's env, each uses step with its with inputs (an action's defaults among them) and env; cache and artifact transfers say how far they have got every five seconds; and a last Complete job step sets the job's outputs and stops what the steps left running, found by RUNNER_TRACKING_ID as GitHub's runner finds it. Two batches of a job's log written at once can no longer reach for the same seq and lose one. The Actions guide and the rate limits page say so.5d ago

log-window.test.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9'5d ago

log-window.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9'5d ago

markdown-plugins.test.tsMerge branch 'worktree-agent-a8385d293d42c913a'1w ago

markdown-plugins.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

markdown-tree.test.tsThe docs folder is gone, and what it held lives where people read it: how a self-hosted g1t runs and how to deploy g1t to Cloudflare are pages on docs.g1t.sh under Run g1t yourself, and speed, rate limits and operating g1t.sh are sections of CONTRIBUTING.md; code that cited a file in docs/ now points to the page or section that covers it, or says what it means itself, and applied migrations and the runner images are left as they were.1w ago

markdown-tree.tsMerge the site's CPU: code highlighted once per data centre by content, markdown parsed once per text, and the request handler built once per isolate1w ago

member-invitations.server.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

mention.test.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

mention.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

merge.test.tsMerge branch 'forking': repository forking, pull request merge methods and every repository setting for pull requests, commits, pushes and issues3d ago

merge.tsMerge branch 'forking': repository forking, pull request merge methods and every repository setting for pull requests, commits, pushes and issues3d ago

meta.test.tsOnly projects have descriptions now; a repository no longer keeps one of its own (#45 ). Every repository already had a project of its own name, which showed the repository's description until it was given another, so the same line lived in two places with two settings pages. Now:2d ago

meta.tsOnly projects have descriptions now; a repository no longer keeps one of its own (#45 ). Every repository already had a project of its own name, which showed the repository's description until it was given another, so the same line lived in two places with two settings pages. Now:2d ago

milestones.server.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

mission-control.server.tsMerge branch 'worktree-agent-ac2922d9eb84a7909'2d ago

mission-control.test.tsMerge branch 'worktree-agent-aa8485640dce3cc93' into ship-batch2d ago

mission-control.tsMerge branch 'worktree-agent-aa8485640dce3cc93' into ship-batch2d ago

mission.test.tsMerge branch 'worktree-agent-a7823ad540eb3433b'4d ago

mission.tsMerge branch 'worktree-agent-a7823ad540eb3433b'4d ago

money.test.tsMoney is written one way. A single formatter turns millionths of a dollar into dollars, rounding half up on whole micros rather than on a float, so the same sum reads the same on every page: under a cent reads <$0.01 on a total and exactly nothing is $0.00, while the statement's lines, a session's receipt, the price book and an agent's effort costs carry up to four places where the fraction of a cent is the point; the six formatters that each rounded their own way are gone. This month is the calendar month in UTC from its first day to today everywhere, and billing counts the month's not-yet-closed usage in any range that reaches into the current month, so the top bar's pill, Spend, Home and Usage ask for the same days and get the same figure; Home now reads the usage report the others read instead of adding up statements. Usage's pending sentence says what of that usage the close will charge after the discount and included usage, which the billing API returns as pending_charged_micros. A reconciliation test holds the pill, Spend, Home and Usage to one number for one month. The usage and billing guide says how amounts are written and what this month means.6d ago

money.tsMoney is written one way. A single formatter turns millionths of a dollar into dollars, rounding half up on whole micros rather than on a float, so the same sum reads the same on every page: under a cent reads <$0.01 on a total and exactly nothing is $0.00, while the statement's lines, a session's receipt, the price book and an agent's effort costs carry up to four places where the fraction of a cent is the point; the six formatters that each rounded their own way are gone. This month is the calendar month in UTC from its first day to today everywhere, and billing counts the month's not-yet-closed usage in any range that reaches into the current month, so the top bar's pill, Spend, Home and Usage ask for the same days and get the same figure; Home now reads the usage report the others read instead of adding up statements. Usage's pending sentence says what of that usage the close will charge after the discount and included usage, which the billing API returns as pending_charged_micros. A reconciliation test holds the pill, Spend, Home and Usage to one number for one month. The usage and billing guide says how amounts are written and what this month means.6d ago

next.test.tsTeam plan, an open-source pool, monthly trials and honest metering; the sidebar for everyone; a workspace that stays put1w ago

next.tsTeam plan, an open-source pool, monthly trials and honest metering; the sidebar for everyone; a workspace that stays put1w ago

no-native-title.test.tsMerge branch 'worktree-agent-ae1299e92e4462012'1w ago

nonce.tsMerge g1tusercontent.com: registry answers run nothing in a browser, the site's pages run only their own scripts, repository files and avatars on their own origin, raw files rate limited per address1w ago

not-found.server.tsTeam plan, an open-source pool, monthly trials and honest metering; the sidebar for everyone; a workspace that stays put1w ago

not-found.test.tsstatus.g1t.sh with incident management, invites that land you in the workspace, settings as pages, usage without quotas1w ago

not-found.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

notify-client.tsMerge remote-tracking branch 'g1t/main' into HEAD2d ago

notify-store.test.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

notify-store.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

observability-live.tsProjects have Observability: their requests, logs, traces, metrics and errors, from apps on g1t.page with nothing to set up, and from software anywhere else over OpenTelemetry.5d ago

observability-trace.tsProjects have Observability: their requests, logs, traces, metrics and errors, from apps on g1t.page with nothing to set up, and from software anywhere else over OpenTelemetry.5d ago

observability-words.tsProjects have Observability: their requests, logs, traces, metrics and errors, from apps on g1t.page with nothing to set up, and from software anywhere else over OpenTelemetry.5d ago

observability.server.tsProjects have Observability: their requests, logs, traces, metrics and errors, from apps on g1t.page with nothing to set up, and from software anywhere else over OpenTelemetry.5d ago

observability.test.tsProjects have Observability: their requests, logs, traces, metrics and errors, from apps on g1t.page with nothing to set up, and from software anywhere else over OpenTelemetry.5d ago

observability.tsProjects have Observability: their requests, logs, traces, metrics and errors, from apps on g1t.page with nothing to set up, and from software anywhere else over OpenTelemetry.5d ago

onboarding.server.tsMerge branch 'worktree-agent-ac7d6e0b40efe9fe8' into ship-batch2d ago

onboarding.test.tsMerge branch 'worktree-agent-ac7d6e0b40efe9fe8' into ship-batch2d ago

onboarding.tsMerge branch 'worktree-agent-ac7d6e0b40efe9fe8' into ship-batch2d ago

one-tab.test.tsOne tab pings, once, and only for what is for you. Every open tab of g1t heard the same arrival, so each played its sound and showed its desktop toast; now a tab claims the arrival across the browser before pinging (lib/one-tab.ts: a Web Lock held while a ten-minute mark is written to the localStorage every tab shares, so a tab that comes second finds the mark), a tab in front claims at once and the others wait 400 ms, so the one you are looking at is the one that pings, and the sound and the toast for an arrival go together. A message heard over both the conversation's socket and the feed is claimed once, by its id. g1t's answer no longer pings as it starts: it arrives empty and thinking the moment a question is asked, which pinged right after your own message; it now pings once, when it is written. Where Web Locks or storage are missing, every tab pings as before. Checked: the site's typecheck and its 893 tests (new: the claim's rule and the answer's silence until it is written), and a production build whose chunks load first without error.4d ago

one-tab.tsOne tab pings, once, and only for what is for you. Every open tab of g1t heard the same arrival, so each played its sound and showed its desktop toast; now a tab claims the arrival across the browser before pinging (lib/one-tab.ts: a Web Lock held while a ten-minute mark is written to the localStorage every tab shares, so a tab that comes second finds the mark), a tab in front claims at once and the others wait 400 ms, so the one you are looking at is the one that pings, and the sound and the toast for an arrival go together. A message heard over both the conversation's socket and the feed is claimed once, by its id. g1t's answer no longer pings as it starts: it arrives empty and thinking the moment a question is asked, which pinged right after your own message; it now pings once, when it is written. Where Web Locks or storage are missing, every tab pings as before. Checked: the site's typecheck and its 893 tests (new: the claim's rule and the answer's silence until it is written), and a production build whose chunks load first without error.4d ago

opened-by.test.tsMerge branch 'worktree-agent-a7823ad540eb3433b'4d ago

opened-by.tsMerge branch 'worktree-agent-a7823ad540eb3433b'4d ago

otp.test.tsPersonal-data erasure has an intake, a staff case with its trail, and cleanup that identity commits with the purge and retries until each service confirms it; one-time-code fields work with password managers and take recovery codes; the Ask panel closes when you follow a link from it. Settings → Account records an erasure request (identity request_erasure / my_erasure, migration 0049), even when a workspace you own alone or billing stands in the way of closing the account; it deletes nothing, keeps the first deadline when asked again, and the deadline is the same day next calendar month, clamped to its last day. sudo → Customers → Privacy lists the cases (the sidebar counts open ones), records a request received elsewhere with its original receipt time (an intake for an open case joins its trail and can only move the deadline earlier), shows each case's trail of intake, reviews, refusal or completion, and refuses completion while the account or any of its cleanup jobs remain or any of the seven evidence areas is empty. Purging an account or a workspace now writes five purge jobs (avatar, Chat, Notify, Artifacts, the lifecycle event) in the same D1 batch as the deletion, guarded like every statement in it, so a restore that won leaves neither; the 15-minute tick runs them, keeps a failure pending with backoff to 16 hours, never stores a response body, and clears a done job's payload. Publishing user.deleted or workspace.deleted is delivery to the bus, not proof its consumers finished. A workspace's jobs carry its members at purge time and its former slugs no live workspace holds. Chat (migration 0006): the person's direct message with g1t goes with its answers and pending job; in shared conversations only the person is shut out and their messages show as ghost, where before the room's clear told every member's tab the conversation was emptied and dropped others' pending answers; a workspace's channels are deleted 500 messages at a time within a budget, answering pending until done; the channels to visit are listed once with the tombstone; tombstones, looked up by key with new indexes, refuse late messages, members, reactions, channels, settings, emoji and meter rows. Notify: an erased feed drops later deliveries quietly instead of throwing; a purged workspace's slug can be made again by the person whose username it is, so its marks are times, not bans: notifications from before the purge are dropped, counts, seeds and visits refused for a day, and its presence room refuses only words said before the purge; the purge reaches identity's members and the room's people under every slug, and clears complete:, forgot: and the workspace's notification level. Artifacts (migration 0010): unshared private folios and their files, versions, search rows and vectors go; authorship (now also cleared_by and repo_spaces.added_by) shows as ghost and mention lists drop the username; a workspace is swept a page at a time and resumes after a failure; full-text rows of a document are deleted by document; an index run that races an erasure writes no full-text row without a chunk row and takes back the vectors it wrote; a room's destroy waits for its index run; ordinary page deletion now destroys the page's live document. Not covered automatically, and listed in docs/operations/privacy-erasure.md as manual review: comment threads inside shared documents' live state and version snapshots, private-space pages, rooms of pages deleted before this, folio trash past 200 rooms, earlier folios' version and upload blobs, former members' and outsiders' notifications, backups. One-time codes: the sign-in's second step, two-factor settings and email confirmation share otpInputProps (lib/otp.ts): a stable id and label, autocomplete one-time-code, no password-manager ignore attributes; the sign-in field takes either kind of code at recovery-code length whichever is chosen, and Use a recovery code switches the keyboard. The Ask panel closes on every completed navigation (location.key), including its direct-message link and citations to the page you are on and back or forward; links opened in another tab and revalidation leave it open. The Privacy Policy distinguishes closure from erasure; the authentication guide says how to ask and how recovery codes are entered; the services page lists identity's new CHAT, NOTIFY and ARTIFACTS bindings. Deploy migrations first, then chat, notify and artifacts before identity, then web and sudo. Checked: cargo test -p g1t-identity --lib (175) and its wasm32 check; node --test conformance/erasure.test.mjs (11, new: a shared channel's room shuts out only the erased member, a large channel deleted in chunks with late writes refused, a reclaimed slug keeps new notifications and presence, vectors taken back after a racing erasure, an upload sweep resuming after a failure, late deliveries dropped quietly); artifacts' 126 tests and typecheck, notify's 49 and typecheck, chat's typecheck; the site's tsc -b --force, 926 tests (new: otp) and a production build; sudo's typecheck, 156 tests and build; contracts' 48; the deploy tool's tests (the one failure is services/runner/base.json, from separate uncommitted runner work); the docs build.3d ago

otp.tsPersonal-data erasure has an intake, a staff case with its trail, and cleanup that identity commits with the purge and retries until each service confirms it; one-time-code fields work with password managers and take recovery codes; the Ask panel closes when you follow a link from it. Settings → Account records an erasure request (identity request_erasure / my_erasure, migration 0049), even when a workspace you own alone or billing stands in the way of closing the account; it deletes nothing, keeps the first deadline when asked again, and the deadline is the same day next calendar month, clamped to its last day. sudo → Customers → Privacy lists the cases (the sidebar counts open ones), records a request received elsewhere with its original receipt time (an intake for an open case joins its trail and can only move the deadline earlier), shows each case's trail of intake, reviews, refusal or completion, and refuses completion while the account or any of its cleanup jobs remain or any of the seven evidence areas is empty. Purging an account or a workspace now writes five purge jobs (avatar, Chat, Notify, Artifacts, the lifecycle event) in the same D1 batch as the deletion, guarded like every statement in it, so a restore that won leaves neither; the 15-minute tick runs them, keeps a failure pending with backoff to 16 hours, never stores a response body, and clears a done job's payload. Publishing user.deleted or workspace.deleted is delivery to the bus, not proof its consumers finished. A workspace's jobs carry its members at purge time and its former slugs no live workspace holds. Chat (migration 0006): the person's direct message with g1t goes with its answers and pending job; in shared conversations only the person is shut out and their messages show as ghost, where before the room's clear told every member's tab the conversation was emptied and dropped others' pending answers; a workspace's channels are deleted 500 messages at a time within a budget, answering pending until done; the channels to visit are listed once with the tombstone; tombstones, looked up by key with new indexes, refuse late messages, members, reactions, channels, settings, emoji and meter rows. Notify: an erased feed drops later deliveries quietly instead of throwing; a purged workspace's slug can be made again by the person whose username it is, so its marks are times, not bans: notifications from before the purge are dropped, counts, seeds and visits refused for a day, and its presence room refuses only words said before the purge; the purge reaches identity's members and the room's people under every slug, and clears complete:, forgot: and the workspace's notification level. Artifacts (migration 0010): unshared private folios and their files, versions, search rows and vectors go; authorship (now also cleared_by and repo_spaces.added_by) shows as ghost and mention lists drop the username; a workspace is swept a page at a time and resumes after a failure; full-text rows of a document are deleted by document; an index run that races an erasure writes no full-text row without a chunk row and takes back the vectors it wrote; a room's destroy waits for its index run; ordinary page deletion now destroys the page's live document. Not covered automatically, and listed in docs/operations/privacy-erasure.md as manual review: comment threads inside shared documents' live state and version snapshots, private-space pages, rooms of pages deleted before this, folio trash past 200 rooms, earlier folios' version and upload blobs, former members' and outsiders' notifications, backups. One-time codes: the sign-in's second step, two-factor settings and email confirmation share otpInputProps (lib/otp.ts): a stable id and label, autocomplete one-time-code, no password-manager ignore attributes; the sign-in field takes either kind of code at recovery-code length whichever is chosen, and Use a recovery code switches the keyboard. The Ask panel closes on every completed navigation (location.key), including its direct-message link and citations to the page you are on and back or forward; links opened in another tab and revalidation leave it open. The Privacy Policy distinguishes closure from erasure; the authentication guide says how to ask and how recovery codes are entered; the services page lists identity's new CHAT, NOTIFY and ARTIFACTS bindings. Deploy migrations first, then chat, notify and artifacts before identity, then web and sudo. Checked: cargo test -p g1t-identity --lib (175) and its wasm32 check; node --test conformance/erasure.test.mjs (11, new: a shared channel's room shuts out only the erased member, a large channel deleted in chunks with late writes refused, a reclaimed slug keeps new notifications and presence, vectors taken back after a racing erasure, an upload sweep resuming after a failure, late deliveries dropped quietly); artifacts' 126 tests and typecheck, notify's 49 and typecheck, chat's typecheck; the site's tsc -b --force, 926 tests (new: otp) and a production build; sudo's typecheck, 156 tests and build; contracts' 48; the deploy tool's tests (the one failure is services/runner/base.json, from separate uncommitted runner work); the docs build.3d ago

packages.test.tsMerge branch 'worktree-agent-a6a121745e81f639f'1w ago

packages.tsMerge branch 'worktree-agent-a6a121745e81f639f'1w ago

page-headers.test.tsMerge g1tusercontent.com: registry answers run nothing in a browser, the site's pages run only their own scripts, repository files and avatars on their own origin, raw files rate limited per address1w ago

page-headers.tsThe policies are published, and g1t is written to at support@g1t.sh. Terms, Privacy, Refunds, Acceptable use and Subprocessors lose the draft banner (POLICIES_DRAFT is false, updated 2026-10-14, with the change in the policies' history): the terms name Delaware as governing law, say plainly that the plan renews each month until cancelled, and leave statutory consumer rights (an EU or UK consumer's 14 days to withdraw, how to ask, refunds within the legal deadline) above the no-refund and no-proration rules; writing to us before disputing a charge is a request, never a condition. Privacy clarifies analytics consent and what credential redaction can and cannot catch, and drops assurances about provider contracts and transfers that were not verified. Every contact (support, security, privacy, billing, abuse, legal) is support@g1t.sh, an alias for support@flagon.io, in the policies, SECURITY.md, the security page, pricing's enterprise link, the spike banner, a suspended account's sign-in message and the guides that named hey@flagon.io. Email sign-up is protected by Cloudflare Turnstile: the site's public key is set (the challenge still turns on only with both keys present), and Privacy and Subprocessors say so. Cloudflare Web Analytics' beacon is out of the site's script policy, as it no longer loads.3d ago

pending.test.tsMerge branch 'worktree-agent-a985c01f43538af01'1w ago

pending.tsMerge branch 'worktree-agent-a985c01f43538af01'1w ago

people-search.test.tsMerge workspace invitations: nobody joins a workspace without saying yes, people are found by username, your own invites can bring someone in, and nobody is left without a workspace (identity 0040)1w ago

people-search.tsMerge workspace invitations: nobody joins a workspace without saying yes, people are found by username, your own invites can bring someone in, and nobody is left without a workspace (identity 0040)1w ago

people.test.tsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

people.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

perf.server.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

perf.test.tsMerge project overview: one branch_drift call, spliced histories, cached tags, 6 repos calls instead of 251w ago

perf.tsMerge branch 'worktree-agent-ac7d6e0b40efe9fe8' into ship-batch2d ago

phone.test.tsMerge the workspace shell: navigation and phone shell, g1t as orchestrator, agents in roles with audience-checked reads, reactions and custom emoji, live notifications and browser push, the homepage tour (agents 0002, chat 0002)1w ago

phone.tsMerge the workspace shell: navigation and phone shell, g1t as orchestrator, agents in roles with audience-checked reads, reactions and custom emoji, live notifications and browser push, the homepage tour (agents 0002, chat 0002)1w ago

pins.test.tsNavigation: the workspace's own page with tabs, pins and recent in the sidebar1w ago

pins.tsProjects show what they are by their icon: the projects list and grid, and the sidebar's Pinned, Recent and Projects rows, drew the same box for every project. ProjectIcon (components/project-icon.tsx) now picks one from Project.kind: a window for an app or site, books for a library, a terminal for a tool or CLI, an open book for documentation, and the box for anything else. The sidebar still shows a lock for a private project, where nothing else on its row says so; the list keeps its private badge beside the name. Sidebar shortcuts (shortcutOf, ShortcutProject) carry the kind. Checked: apps/web typecheck, lib/pins tests.2d ago

presence.test.tsChat has sounds. A short warm cue plays for a message in a conversation you have open but aren't looking at, a direct message, a mention of you and an agent finishing something for you, with a soft tick for sending that is off until you turn it on; never for your own messages, a muted conversation, a conversation you are looking at in a window that has the front, or while you have paused notifications, and never twice for one message or more than once in a second and a half. The cues are made by the browser itself in two sets, Soft and Bright, so nothing is downloaded; the Sounds and notifications section of your notification settings turns them on and off, picks the set and the volume, plays each one, and offers desktop notifications for when the window is behind; Do not disturb is the one pause everyone sees, set from the chat sidebar's bell, the settings, or your avatar menu, now with a two-hour choice, and the top bar's bell shows a dot while it's on. Your choices are kept with your account. The sound layer sits behind one small interface so the desktop app can play them natively later. The chat and notifications guides say the rules.6d ago

presence.tsChat has sounds. A short warm cue plays for a message in a conversation you have open but aren't looking at, a direct message, a mention of you and an agent finishing something for you, with a soft tick for sending that is off until you turn it on; never for your own messages, a muted conversation, a conversation you are looking at in a window that has the front, or while you have paused notifications, and never twice for one message or more than once in a second and a half. The cues are made by the browser itself in two sets, Soft and Bright, so nothing is downloaded; the Sounds and notifications section of your notification settings turns them on and off, picks the set and the volume, plays each one, and offers desktop notifications for when the window is behind; Do not disturb is the one pause everyone sees, set from the chat sidebar's bell, the settings, or your avatar menu, now with a two-hour choice, and the top bar's bell shows a dot while it's on. Your choices are kept with your account. The sound layer sits behind one small interface so the desktop app can play them natively later. The chat and notifications guides say the rules.6d ago

production-image.test.tsMerge branch 'worktree-agent-a5b0d9e2733e9e382'5d ago

production-image.tsMerge branch 'worktree-agent-a5b0d9e2733e9e382'5d ago

project-kind.test.tsAn app that deploys itself says so, instead of "Deployed elsewhere". A project whose own workflows (or another pipeline) deploy it, as g1t's do, read as if g1t had nothing to do with it: grey "Deployed elsewhere" in the workspace's Projects, "App, deployed elsewhere" on its badge, and the API said `runs` is `g1t` "when g1t deploys it". What tells the two apart is who hosts it, so they are now named that way: App or site, hosted on g1t.page (badge App on g1t.page) and App or site that deploys itself (badge App, deploys itself), in settings, the badge's menu, the overview's "Where does it run?" (It deploys itself), the Deployments pages, the guides and the API's descriptions; `runs` keeps its values. In Projects an app that deploys itself shows the rocket, Deploys itself, and the Deploys filter counts it. Its production card says which workflow run deployed it when g1t Actions did ("deployed 16m ago by its workflow"), else "by its own pipeline". An app nobody has placed that deploys production itself (a successful production deployment from a workflow or the API, not g1t.page's) now deploys itself everywhere, where before only its overview inferred it: the projects service keeps that it has (migration 0013, `reports_production`), and Deployments being on or a setting still win. Checked: the site's typecheck (tsc -b --force), its 911 tests and a production build; the projects service's typecheck and 49 tests (new: an app that deploys production itself); the API's openapi tests with the docs copy regenerated; the deploy tool's 92 tests.3d ago

project-kind.tsAn app that deploys itself says so, instead of "Deployed elsewhere". A project whose own workflows (or another pipeline) deploy it, as g1t's do, read as if g1t had nothing to do with it: grey "Deployed elsewhere" in the workspace's Projects, "App, deployed elsewhere" on its badge, and the API said `runs` is `g1t` "when g1t deploys it". What tells the two apart is who hosts it, so they are now named that way: App or site, hosted on g1t.page (badge App on g1t.page) and App or site that deploys itself (badge App, deploys itself), in settings, the badge's menu, the overview's "Where does it run?" (It deploys itself), the Deployments pages, the guides and the API's descriptions; `runs` keeps its values. In Projects an app that deploys itself shows the rocket, Deploys itself, and the Deploys filter counts it. Its production card says which workflow run deployed it when g1t Actions did ("deployed 16m ago by its workflow"), else "by its own pipeline". An app nobody has placed that deploys production itself (a successful production deployment from a workflow or the API, not g1t.page's) now deploys itself everywhere, where before only its overview inferred it: the projects service keeps that it has (migration 0013, `reports_production`), and Deployments being on or a setting still win. Checked: the site's typecheck (tsc -b --force), its 911 tests and a production build; the projects service's typecheck and 49 tests (new: an app that deploys production itself); the API's openapi tests with the docs copy regenerated; the deploy tool's 92 tests.3d ago

project-list.test.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

project-list.tsAn app that deploys itself says so, instead of "Deployed elsewhere". A project whose own workflows (or another pipeline) deploy it, as g1t's do, read as if g1t had nothing to do with it: grey "Deployed elsewhere" in the workspace's Projects, "App, deployed elsewhere" on its badge, and the API said `runs` is `g1t` "when g1t deploys it". What tells the two apart is who hosts it, so they are now named that way: App or site, hosted on g1t.page (badge App on g1t.page) and App or site that deploys itself (badge App, deploys itself), in settings, the badge's menu, the overview's "Where does it run?" (It deploys itself), the Deployments pages, the guides and the API's descriptions; `runs` keeps its values. In Projects an app that deploys itself shows the rocket, Deploys itself, and the Deploys filter counts it. Its production card says which workflow run deployed it when g1t Actions did ("deployed 16m ago by its workflow"), else "by its own pipeline". An app nobody has placed that deploys production itself (a successful production deployment from a workflow or the API, not g1t.page's) now deploys itself everywhere, where before only its overview inferred it: the projects service keeps that it has (migration 0013, `reports_production`), and Deployments being on or a setting still win. Checked: the site's typecheck (tsc -b --force), its 911 tests and a production build; the projects service's typecheck and 49 tests (new: an app that deploys production itself); the API's openapi tests with the docs copy regenerated; the deploy tool's 92 tests.3d ago

project-nav.test.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

project-nav.tsEvery page's script loads again (#51 ): since Deploy #356 , pages threw "Cannot read properties of undefined (reading 'filter')" as they loaded, so nothing on them answered a click and every Select showed empty. A project's tabs (lib/project-nav.ts) and the phone's soon pages (components/mobile.tsx) read the roadmap as their module loaded, and the bundle had put the roadmap in a chunk that imports theirs back, so it was read before it was there. Both are now built on first use, which no order of chunks can break. Explain moves into components/home.tsx, so Home's parts no longer import each other. Checked: a production build loaded route by route in Chrome (root, then each route's module): 0 of 257 fail, where the build before this fails all 257 with the same error and stack as production; the site's typecheck and tests (955).2d ago

provenance.server.tsAgents as a team: lifecycle, merge queue, billing and a new shell2w ago

public-cache.test.tsSigned-out page cache: a repository's kept page is served only while the repository is still public, so one made private or deleted never shows from any data centre's copy1w ago

public-cache.tsSigned-out page cache: a repository's kept page is served only while the repository is still public, so one made private or deleted never shows from any data centre's copy1w ago

pull-detail.test.tsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

reconcile.test.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

reconnect.tsRetrying a read that gets no answer never becomes a loop (after 71274eeb): a refresh of the page already showing kept trying every 5 s for as long as it took, which is right while the browser is offline but not on a network that is up and never answers (an extension or proxy blocking g1t's .data requests, a security policy, broken DNS): the page sat on old data under "Can't reach g1t" forever and never showed what was wrong, and a tab out of view kept asking too. Now a refresh counts only the time it waited while the browser said it was online, and gives up after 2 minutes of that (REFRESH_PATIENCE_MS); offline still waits as long as it lasts. Nothing is retried from a tab out of view: a wait that is over holds until the tab is looked at. While the error page tries on its own (lib/reconnect.ts, 3 s, 6 s, 12 s, then every 30 s), reads are tried once each instead of retrying underneath it, so a broken network costs one request every 30 s on a visible error page and none on a hidden one. A real error (an answer that is an error) was already never retried and stops the error page's trying. keepTrying in lib/resilient-fetch.ts makes the call. Checked: apps/web tsc -b; npm test (985; new: offline waited out however long, up-and-silent gives up for a refresh and a click, one try while the error page tries).2d ago

refresh.tsMerge branch 'worktree-agent-a9934473c8cf2bca6'4d ago

registration.server.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

registration.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

registry-paths.test.tsMerge branch 'worktree-agent-a8385d293d42c913a'1w ago

registry-paths.tsMerge branch 'worktree-agent-a8385d293d42c913a'1w ago

renamed.server.tsMerge branch 'worktree-agent-a8385d293d42c913a'1w ago

repo-lifecycle.test.tsInvite-only launch: sign in with GitHub, repository access and lifecycle, many emails, a new look1w ago

repo-lifecycle.tsOnly projects have descriptions now; a repository no longer keeps one of its own (#45 ). Every repository already had a project of its own name, which showed the repository's description until it was given another, so the same line lived in two places with two settings pages. Now:2d ago

repo-live.test.tsMerge branch 'worktree-agent-a9934473c8cf2bca6'4d ago

repo-live.tsMerge branch 'main' into worktree-agent-a654d8ba46392bedd2d ago

resilient-fetch.test.tsRetrying a read that gets no answer never becomes a loop (after 71274eeb): a refresh of the page already showing kept trying every 5 s for as long as it took, which is right while the browser is offline but not on a network that is up and never answers (an extension or proxy blocking g1t's .data requests, a security policy, broken DNS): the page sat on old data under "Can't reach g1t" forever and never showed what was wrong, and a tab out of view kept asking too. Now a refresh counts only the time it waited while the browser said it was online, and gives up after 2 minutes of that (REFRESH_PATIENCE_MS); offline still waits as long as it lasts. Nothing is retried from a tab out of view: a wait that is over holds until the tab is looked at. While the error page tries on its own (lib/reconnect.ts, 3 s, 6 s, 12 s, then every 30 s), reads are tried once each instead of retrying underneath it, so a broken network costs one request every 30 s on a visible error page and none on a hidden one. A real error (an answer that is an error) was already never retried and stops the error page's trying. keepTrying in lib/resilient-fetch.ts makes the call. Checked: apps/web tsc -b; npm test (985; new: offline waited out however long, up-and-silent gives up for a refresh and a click, one try while the error page tries).2d ago

resilient-fetch.tsRetrying a read that gets no answer never becomes a loop (after 71274eeb): a refresh of the page already showing kept trying every 5 s for as long as it took, which is right while the browser is offline but not on a network that is up and never answers (an extension or proxy blocking g1t's .data requests, a security policy, broken DNS): the page sat on old data under "Can't reach g1t" forever and never showed what was wrong, and a tab out of view kept asking too. Now a refresh counts only the time it waited while the browser said it was online, and gives up after 2 minutes of that (REFRESH_PATIENCE_MS); offline still waits as long as it lasts. Nothing is retried from a tab out of view: a wait that is over holds until the tab is looked at. While the error page tries on its own (lib/reconnect.ts, 3 s, 6 s, 12 s, then every 30 s), reads are tried once each instead of retrying underneath it, so a broken network costs one request every 30 s on a visible error page and none on a hidden one. A real error (an answer that is an error) was already never retried and stops the error page's trying. keepTrying in lib/resilient-fetch.ts makes the call. Checked: apps/web tsc -b; npm test (985; new: offline waited out however long, up-and-silent gives up for a refresh and a click, one try while the error page tries).2d ago

roadmap.tsAgents in any harness coordinate through one inbox, the merge queue lands only what an attempt itself tested, and workspaces have Insights. Each session keeps the coordination updates that concern it (session_updates, migration 0038): claims, releases, hand-offs and expiries, overlaps, dependency changes, decisions, integration results and sessions starting and ending, about the issues and pull requests it is on or tied to, or every one while it is on nothing yet; never its own start, end or claims. An update is written in the same batch as the change it announces, guarded so it exists only once the change is made, with an id derived from the change, so a retry or two requests at once deliver it once; a blocker clearing and queue results follow their change under a key that names it. get_session's `updates` lists up to 100, oldest first, and reading never clears them; record_session_entries' `acknowledge` clears exact IDs for that session only and says how many were its (`acknowledged`). The CLI has `g1t session updates` and `g1t session acknowledge ID --note` (the decision note and the acknowledgement in one idempotent call, `--session` over G1T_SESSION), and those and `heartbeat` say which files the working copy changed (git status; G1T_TOUCH=0 turns it off), so overlapping edits are noticed whichever harness makes them; a pair of sessions is told again when another file in common joins. The Claude Code hooks fetch the inbox in the background, show it at every prompt and before tools when it changes or every two minutes, refresh after an acknowledgement, and never acknowledge. The AGENTS.md snippet, the skills, MCP's instructions and llms.txt say the same steps for every harness: scope claims (Scope, Interfaces, Depends on, Checks), proceed on complementary work, agree ownership before overlapping edits, record public summaries and evidence, never private reasoning. Claims resolve ownership in one statement, so simultaneous claims by one session converge. The merge queue: every attempt has its own id and branch (0037); a claimed build, an accepted build, a report and a callback each check their attempt; a dispatch registers its runs first and Actions fails the dispatch when it cannot; an attempt is decided only by its own runs (and other tools' statuses set after it was claimed), never an earlier attempt's or the pull request's own on the same commit (0040 keeps what it built and how many runs it registered); landing re-reads the rules, honours a bypass asked for when queued (0039), checks no predecessor remains, carries the exact tested commit to Git's compare-and-swap, records publication intent that fences removal, closing, drafting and changing base, resumes after a crash, and a publisher that lost a race no longer undoes the winner's landing; a retried attempt's branch is deleted; an entry whose branch or pull request is gone leaves the queue with a note instead of blocking those behind it; diagnosis rebuilds exactly what was tested and names a failing combination, not a proven cause. A session's page shows what awaits its agent as the coordination feed does. Workspace Insights (/:owner/-/insights): delivery metrics across a workspace's readable repositories (work service delivery_insights). Deploy Work migrations 0037 to 0040 first (the pipeline does), then Repos and Actions before Work, and the site with Work. Checked: cargo test for g1t-work (143), g1t-repos (311), g1t-actions-service (108), g1t-api (201), g1t (149), g1t-rules (72), g1t-contracts (306) and wasm32 builds; cargo check --workspace; the conformance suite against freshly built API and Work Workers (5, none skipped); the Claude Code hooks (49); the site's typecheck, tests (919) and a production build; the deploy tool's 92 tests; the docs build; and two unscripted local rehearsals in which two concurrent agents, given only a task and AGENTS.md, found each other, agreed a split and made complementary edits.3d ago

robots.test.tsMerge main into project-descriptions2d ago

robots.tsMerge main into project-descriptions2d ago

rules.test.tsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

rules.tsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

run-artifacts.server.tsThe artifacts service is services/artifacts, the Worker g1t-artifacts, bound as ARTIFACTS by the API, the site and the agents; its live rooms move to it with a Durable Object transfer from g1t-docs-service, and its database, bucket, indexes and queue keep their names. The git store's binding and settings are GITSTORE, its ops scripts gitstore-*, and workflow run artifacts keep their compatible API under run_artifacts modules. The deploy tool puts a Worker that has never deployed before the Workers in its stage that bind to it, and the deploy guide gives the cutover runbook.1w ago

run-artifacts.test.tsThe artifacts service is services/artifacts, the Worker g1t-artifacts, bound as ARTIFACTS by the API, the site and the agents; its live rooms move to it with a Durable Object transfer from g1t-docs-service, and its database, bucket, indexes and queue keep their names. The git store's binding and settings are GITSTORE, its ops scripts gitstore-*, and workflow run artifacts keep their compatible API under run_artifacts modules. The deploy tool puts a Worker that has never deployed before the Workers in its stage that bind to it, and the deploy guide gives the cutover runbook.1w ago

run-artifacts.tsThe artifacts service is services/artifacts, the Worker g1t-artifacts, bound as ARTIFACTS by the API, the site and the agents; its live rooms move to it with a Durable Object transfer from g1t-docs-service, and its database, bucket, indexes and queue keep their names. The git store's binding and settings are GITSTORE, its ops scripts gitstore-*, and workflow run artifacts keep their compatible API under run_artifacts modules. The deploy tool puts a Worker that has never deployed before the Workers in its stage that bind to it, and the deploy guide gives the cutover runbook.1w ago

run-controls.test.tsMerge the run page fixes: Cancel run shows on every view of a run that is still going, including an earlier attempt's and job links, and pages refresh when a run is re-run from anywhere; a job's log follows to its true end, above anything covering the bottom of the window, and Jump to latest lands there.5d ago

run-controls.tsMerge the run page fixes: Cancel run shows on every view of a run that is still going, including an earlier attempt's and job links, and pages refresh when a run is re-run from anywhere; a job's log follows to its true end, above anything covering the bottom of the window, and Jump to latest lands there.5d ago

run-graph.test.tsMerge branch 'worktree-agent-af75396ad4c6ee7a8' into claude/cool-newton-7nm8r95d ago

run-graph.tsMerge branch 'worktree-agent-af75396ad4c6ee7a8' into claude/cool-newton-7nm8r95d ago

runners.server.tsMerge branch 'worktree-agent-a56293a6f836cd343'4d ago

runners.test.tsMerge branch 'worktree-agent-a56293a6f836cd343'4d ago

runners.tsMerge branch 'worktree-agent-a56293a6f836cd343'4d ago

same-origin.tsMerge main into Artifacts Phase 21w ago

search.test.tsSearch across all of g1t, Explore, and a command palette1w ago

search.tsSearch across all of g1t, Explore, and a command palette1w ago

secrets.server.tsMerge brokered secrets: a secret a job uses only to call an API can be tied to the hosts it is for, and the job gets a stand-in in its place, which g1t fills in on the way out to those hosts and nowhere else, so the value never enters the sandbox; and a request to an allowed host that fails upstream is retried when it can be, or answered with a 502 that says why, rather than reset.5d ago

security-alerts.test.tsMerge main (membership, two-factor, GitHub repo roles) into tokens1w ago

security-alerts.tsMerge main (membership, two-factor, GitHub repo roles) into tokens1w ago

security-overview.test.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

security-overview.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

security-suite.server.tsPricing says it plainly: models at the provider's price, the agent rate for what g1t runs around every model call (the gateway, secrets, routing, context and pass-through to your own provider, so your own keys too), everything else at cost plus 20%, your own runners free, no seats; Security and quality comes with the plan with no separate fee, and live activations end. Each agent has an effort setting, Auto to Max, with what a typical task has cost at each level, and Spend's Spend less, keep quality suggests a lower level only when the agent's own past work shows quality held, to apply or dismiss. The pricing, spend and agents guides say how.1w ago

security-suite.test.tsMerge branch 'worktree-agent-a3abfcce648e87dca'1w ago

security-suite.tsMerge branch 'worktree-agent-a3abfcce648e87dca'1w ago

services.server.tsA workspace's AI is its own setting, not an integration. AI providers and what Ask answers with were connections like Sentry's or Linear's: rows of `connections`, cards in the Integrations directory, `provider` values of the integration operations. Integrations are becoming a directory others publish to; a workspace's AI is a base setting g1t looks after itself. Contracts: `g1t_contracts::ai` (and packages/contracts ai.ts) has AiProvider (the 14), AiGroup (labs, platforms, self_hosted), AiConfig, AiConnection and their methods' arguments, and the Ask types moved there; AskSetup names the provider by `provider_id` and `provider_name` (were connection_id, connection_name). `integrations::Provider` is sentry, datadog, webhook, jira and linear; ProviderKind loses Models, ConnectionConfig its base_url, auth_header and model, Connection its models. Storage: integrations migration 0011 makes `ai_providers` and `ai_routes`, moves the AI rows out of `connections` with their ids (sealed keys are bound to the id, and managed_agents and agent_runs keep finding them), keeps Ask's route and drops `model_routes`; renaming and deleting a workspace cover both tables. The service keeps them in src/ai.rs with methods of their own (ai_list, ai_add, ai_update, ai_remove, ai_test, ask_settings, set_ask_settings, ask); a provider's key is checked when added and on each test; removing one turns Ask off when it answered with it and forgets its managed agent. Sessions on issues find the workspace's Anthropic provider there. API: list_ai_providers, add_ai_provider, update_ai_provider, remove_ai_provider, test_ai_provider (/workspaces/{workspace}/ai/providers[/{id}[/test]]), get_ask_settings and set_ask_settings (/workspaces/{workspace}/ai/ask), under the OpenAPI tag AI; reading takes workspace:read, changing workspace:admin and an owner's own token; the integration operations take only integration providers. MCP: a tool `ai` (list_providers, add_provider, update_provider, remove_provider, test_provider, get_ask, set_ask). The CLI's operation table and command reference are regenerated. Site: AI at /<workspace>/-/ai (the sidebar's AI → Models, the command palette's AI models) holds Ask g1t, the providers and Add a provider, grouped Frontier labs, Model platforms, Self-hosted or your own gateway; AI providers are no longer in the Integrations directory or Apps, and the integrations pages are alerts and trackers only; -/integrations/models and ?add=<an AI provider> lead to -/ai. Ask's settings link, and @g1t's "not set up yet" in Chat, point there. Docs: the AI providers guide (with its API), and the integrations, Ask, sessions, CLI, plugin and MCP pages say so. Deploy the integrations migration first, then integrations, then api, chat and web. Checked: cargo test for g1t-contracts (309), g1t-integrations (70), g1t-api (204, the OpenAPI document regenerated) and the CLI (105 and 44, the command reference current); cargo check --workspace; wasm32 checks of api and integrations; every migration applied in order to SQLite with an AI provider, a Sentry connection and Ask's route; the site's tsc -b --force, 927 tests and a production build; contracts' 48; chat's 76 and typecheck; the deploy tool's 92; the docs build (521 pages).3d ago

session-card.test.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

session-card.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

session-entries.test.tsA session's transcript reads as what it said, with its tool calls between: "What it recorded" on a session's page (and What happened on a pull request's Session tab) drew every entry as a full-height row. A tool call said its tool twice ("bash bash: curl …", "g1t:issue g1t:issue: work"), a message's Markdown showed its ** and backticks as written, the prompt's <issue> and <comment> blocks were raw text in a <pre>, and a run of eight tool calls was eight bordered rows. Tool calls between two messages now run together (lib/session-entries.ts transcript): up to three in a tight list, more folded into one line ("7 tool calls · bash, write, read") that opens to it, each result folded under the call before it. Each call says its tool once (readToolCall takes the "<tool>: " off the front of its text, and finds the tool in the text when the entry has none, as from the Claude Code plugin): a command as code in one line, a long one unfolding in full; a file by its path with a file icon; an MCP tool by resource and action ("issue · claim", "pull request · create") with the g1t mark for g1t's, and its server's name for others (mcp__claude_ai_Google_Drive__… is Google Drive). Messages render with the site's Markdown, references linked within the repository. The prompt folds under its first line and opens to it rendered, its <issue> and <comment> blocks as quotes under a label (quotedBlocks). Relative times stay on prompts, messages, skills and notes. The sessions guide says how the transcript reads.2d ago

session-entries.tsA session's transcript reads as what it said, with its tool calls between: "What it recorded" on a session's page (and What happened on a pull request's Session tab) drew every entry as a full-height row. A tool call said its tool twice ("bash bash: curl …", "g1t:issue g1t:issue: work"), a message's Markdown showed its ** and backticks as written, the prompt's <issue> and <comment> blocks were raw text in a <pre>, and a run of eight tool calls was eight bordered rows. Tool calls between two messages now run together (lib/session-entries.ts transcript): up to three in a tight list, more folded into one line ("7 tool calls · bash, write, read") that opens to it, each result folded under the call before it. Each call says its tool once (readToolCall takes the "<tool>: " off the front of its text, and finds the tool in the text when the entry has none, as from the Claude Code plugin): a command as code in one line, a long one unfolding in full; a file by its path with a file icon; an MCP tool by resource and action ("issue · claim", "pull request · create") with the g1t mark for g1t's, and its server's name for others (mcp__claude_ai_Google_Drive__… is Google Drive). Messages render with the site's Markdown, references linked within the repository. The prompt folds under its first line and opens to it rendered, its <issue> and <comment> blocks as quotes under a label (quotedBlocks). Relative times stay on prompts, messages, skills and notes. The sessions guide says how the transcript reads.2d ago

session.server.tsThe docs folder is gone, and what it held lives where people read it: how a self-hosted g1t runs and how to deploy g1t to Cloudflare are pages on docs.g1t.sh under Run g1t yourself, and speed, rate limits and operating g1t.sh are sections of CONTRIBUTING.md; code that cited a file in docs/ now points to the page or section that covers it, or says what it means itself, and applied migrations and the runner images are left as they were.1w ago

shiki.tsYou choose how g1t looks: Auto follows your system, Light and Dark stay put, set from the account menu, the phone's You and help sheet, Settings → Account or the command palette; the page is drawn in your choice from the first byte, every colour has a light value that reads at 4.5:1 or better, code and diffs are highlighted for both, diagrams and the document editor follow along, the product tour and the g1t mark's avatar stay dark, and Dark is what you see until you choose; the workspaces guide says how.1w ago

shortcut.test.tsUsage, Billing settings and prepaid AI credit; fixes from the UX audit1w ago

shortcut.tsUsage, Billing settings and prepaid AI credit; fixes from the UX audit1w ago

socket-ticket.server.tsThe phone app: Chat in full with phone notifications, plus Code and Customize (#4 )6d ago

socket-ticket.test.tsThe phone app: Chat in full with phone notifications, plus Code and Customize (#4 )6d ago

socket-ticket.tsThe phone app: Chat in full with phone notifications, plus Code and Customize (#4 )6d ago

sound-events.tsOne tab pings, once, and only for what is for you. Every open tab of g1t heard the same arrival, so each played its sound and showed its desktop toast; now a tab claims the arrival across the browser before pinging (lib/one-tab.ts: a Web Lock held while a ten-minute mark is written to the localStorage every tab shares, so a tab that comes second finds the mark), a tab in front claims at once and the others wait 400 ms, so the one you are looking at is the one that pings, and the sound and the toast for an arrival go together. A message heard over both the conversation's socket and the feed is claimed once, by its id. g1t's answer no longer pings as it starts: it arrives empty and thinking the moment a question is asked, which pinged right after your own message; it now pings once, when it is written. Where Web Locks or storage are missing, every tab pings as before. Checked: the site's typecheck and its 893 tests (new: the claim's rule and the answer's silence until it is written), and a production build whose chunks load first without error.4d ago

sound-synth.test.tsChat has sounds. A short warm cue plays for a message in a conversation you have open but aren't looking at, a direct message, a mention of you and an agent finishing something for you, with a soft tick for sending that is off until you turn it on; never for your own messages, a muted conversation, a conversation you are looking at in a window that has the front, or while you have paused notifications, and never twice for one message or more than once in a second and a half. The cues are made by the browser itself in two sets, Soft and Bright, so nothing is downloaded; the Sounds and notifications section of your notification settings turns them on and off, picks the set and the volume, plays each one, and offers desktop notifications for when the window is behind; Do not disturb is the one pause everyone sees, set from the chat sidebar's bell, the settings, or your avatar menu, now with a two-hour choice, and the top bar's bell shows a dot while it's on. Your choices are kept with your account. The sound layer sits behind one small interface so the desktop app can play them natively later. The chat and notifications guides say the rules.6d ago

sound-synth.tsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

sounds.tsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

stale-build.test.tsMerge branch 'worktree-agent-a64e51c8f386f46ac'1w ago

stale-build.tsMerge branch 'worktree-agent-a64e51c8f386f46ac'1w ago

starter-workflow.test.tsFast pages, required checks on the branch, self-hosted runners, honest incidents1w ago

starter-workflow.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

status.test.tsMerge branch 'worktree-agent-ad7c6d88d93adc817'1w ago

status.tsMerge branch 'worktree-agent-ad7c6d88d93adc817'1w ago

structured-data.tsMerge site-audit: SEO routes, analytics with consent first in Europe, security page, founder and X account, docs header; the tour opens on the pull request and leads with Code, on the real shell, with Agents and Docs working today1w ago

stuck.test.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

stuck.tsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

tab-paths.test.tsMerge branch 'worktree-agent-ac2922d9eb84a7909'2d ago

tab-paths.tsMerge branch 'worktree-agent-ac2922d9eb84a7909'2d ago

tar.test.tsA Windows runner finds the folders a build's own Wrangler config names. Resolving the config's path on Windows gave a verbatim path (\?\C:\…), in which / and .. are not read, so "directory": "../client" joined to it named nothing; drive paths now lose the prefix. The tar read-back test is skipped on Windows, whose tar takes C: for a remote host and whose files keep no Unix modes or symlinks; CI reads the archive back on Linux.5d ago

tar.tsMerge GitHub pull request #8 (claude/cool-newton-7nm8r9): Docker deployments on g1t.page behind a switch that stays off until Cloudflare's container APIs are checked live, workflows a workspace manages in its repositories, tarball and zipball downloads, refs/pull/* kept to g1t alone, framework detection for g1t.page builds, and a workflow's event payload matching GitHub's.5d ago

teams.test.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

teams.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

theme.test.tsYou choose how g1t looks: Auto follows your system, Light and Dark stay put, set from the account menu, the phone's You and help sheet, Settings → Account or the command palette; the page is drawn in your choice from the first byte, every colour has a light value that reads at 4.5:1 or better, code and diffs are highlighted for both, diagrams and the document editor follow along, the product tour and the g1t mark's avatar stay dark, and Dark is what you see until you choose; the workspaces guide says how.1w ago

theme.tsg1t follows your system's light or dark setting until you choose otherwise, on this browser; the workspaces guide says so, and the privacy policy lists the theme, dock and sidebar cookies.1w ago

time-ago.test.tsTimes read as people say them, and old ones as their date: every TimeAgo across the site (283 of them) said "1176d ago" for a commit three years old, since it only knew days, hours and minutes. It now says "just now", "5m ago", "3h ago", "yesterday", "4d ago" and "3w ago", then the date, "Sep 30" this year and "Jul 24, 2023" before, from lib/time-ago.ts, and the moment in full ("Oct 13, 2026, 3:04 PM UTC") shows in the site's tooltip on hover. A deployment's status no longer wraps its time in a second hint of its own. A time that can't be read shows nothing rather than "NaN". Checked: the site's typecheck, its 893 tests (new: time-ago's wording for each span, a clock slightly ahead, an unreadable time and the full moment) and a production build.4d ago

time-ago.tsTimes read as people say them, and old ones as their date: every TimeAgo across the site (283 of them) said "1176d ago" for a commit three years old, since it only knew days, hours and minutes. It now says "just now", "5m ago", "3h ago", "yesterday", "4d ago" and "3w ago", then the date, "Sep 30" this year and "Jul 24, 2023" before, from lib/time-ago.ts, and the moment in full ("Oct 13, 2026, 3:04 PM UTC") shows in the site's tooltip on hover. A deployment's status no longer wraps its time in a second hint of its own. A time that can't be read shows nothing rather than "NaN". Checked: the site's typecheck, its 893 tests (new: time-ago's wording for each span, a clock slightly ahead, an unreadable time and the full moment) and a production build.4d ago

time-zone.test.tsMerge leftovers: plan activity filtered in the query, pushes counted by account, ghost during the deletion window, profile time zones (identity 0039)1w ago

time-zone.tsMerge leftovers: plan activity filtered in the query, pushes counted by account, ghost during the deletion window, profile time zones (identity 0039)1w ago

token-scopes.test.tsMerge branch 'worktree-agent-a4babc5b5815f6f3a'4d ago

token-scopes.tsThe Agent preset creates repositories, with a new repo:create scope (#52 )2d ago

tokens.test.tsYou choose how g1t looks: Auto follows your system, Light and Dark stay put, set from the account menu, the phone's You and help sheet, Settings → Account or the command palette; the page is drawn in your choice from the first byte, every colour has a light value that reads at 4.5:1 or better, code and diffs are highlighted for both, diagrams and the document editor follow along, the product tour and the g1t mark's avatar stay dark, and Dark is what you see until you choose; the workspaces guide says how.1w ago

tour.test.tsThe landing page starts with Home, where the day starts: the hero tour opens on Priya's Home (notifications that settled overnight leave Where you're needed, Start here picks her pull request in review, What changed says what landed and is live), then she opens #431 from the summary and watches it pass its checks, get approved and merge, then its session, then #web hearing it shipped. Home's two pillars (Where you're needed, What changed) move out of Agents into a Home section right after the facts, and One workspace names Home first. The tour no longer stamps "Works today" under each step or says "Everything here works today"; the product is live. Checked: node --test app/lib/tour.test.ts (10); apps/web tsc -b; the landing page rendered in the dev server, tour stepped through Home, #431 , its session and #web.2d ago

tour.tsThe landing page starts with Home, where the day starts: the hero tour opens on Priya's Home (notifications that settled overnight leave Where you're needed, Start here picks her pull request in review, What changed says what landed and is live), then she opens #431 from the summary and watches it pass its checks, get approved and merge, then its session, then #web hearing it shipped. Home's two pillars (Where you're needed, What changed) move out of Agents into a Home section right after the facts, and One workspace names Home first. The tour no longer stamps "Works today" under each step or says "Everything here works today"; the product is live. Checked: node --test app/lib/tour.test.ts (10); apps/web tsc -b; the landing page rendered in the dev server, tour stepped through Home, #431 , its session and #web.2d ago

turnstile.server.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

turnstile.test.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

turnstile.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

two-factor.server.tsMerge main (membership, two-factor, GitHub repo roles) into tokens1w ago

usage.test.tsMerge branch 'worktree-agent-a56293a6f836cd343'4d ago

usage.tsMerge branch 'worktree-agent-a56293a6f836cd343'4d ago

use-arrived.tsMerge remote-tracking branch 'g1t/main' into HEAD2d ago

use-home-live.tsMerge branch 'main' into worktree-agent-a654d8ba46392bedd2d ago

use-repo-live.tsMerge branch 'main' into worktree-agent-a654d8ba46392bedd2d ago

usercontent.test.tsMerge g1tusercontent.com: registry answers run nothing in a browser, the site's pages run only their own scripts, repository files and avatars on their own origin, raw files rate limited per address1w ago

usercontent.tsAn agent can put an image in a pull request, an issue or a comment without committing it: on flagon-io/session-lab #2 a cloud session asked to show a screenshot of its preview in the pull request's description found no way to upload one (not in g1t --help, g1t api --list or the docs), so it committed the PNG, linked it at that commit and removed it in a second commit. The site has no comment-image upload either; images only ever came from repository files. Now upload_attachment (POST /repos/{owner}/{repo}/attachments, issues:write, the issue tool's attach action over MCP) takes name, data (the file base64-encoded, as upload_sarif takes its file) and an optional content_type, and answers in snake_case with id, name, url, markdown (![name](url), ready to paste), content_type and size. The work service decides, as add_comment does: a confirmed account that can read the repository, which is not archived. The type is read from the bytes (PNG, JPEG, GIF or WebP; a content_type that says otherwise is refused), at most 10 MB; too large is the new FailureCode too_large (413), another type the new unsupported_type (415), in Rust and TS contracts and the OpenAPI error codes. The bytes go in the usercontent namespace (AVATARS, now bound to work) under attachments/<id>, where the id is 256 random bits so the address is its permission, as Docs files' is; a row in the new attachments table (migration 0048) names the repository, and repo.purged deletes the bytes, then the rows. The usercontent origin serves /attachments/<id> as it serves avatars (stored type, nosniff, sandboxed). The API forms url from USERCONTENT_URL as the site does (g1tusercontent.com for g1t.sh, SITE_URL/-/usercontent elsewhere; the self-host config now passes it to the API). The CLI has g1t attach FILE... (prints each image's Markdown; --json the fields, a list for several) and --attach FILE (repeatable) on g1t issue create, issue comment, pr create, pr comment and pr ready, which uploads and appends the Markdown after the text. A session's run credential gets upload_attachment through issues:write, bound to its own repository (tested), and the cloud session's system prompt says to upload screenshots with g1t attach and never commit images just to link them. Docs: the CLI guide (Images in descriptions and comments), pull requests (Images), sessions, the MCP reference, self-hosting's USERCONTENT_URL row, reference.json and the regenerated openapi.json, CLI ops table and CLI reference. Deploy order: apply work migration 0048 and deploy work (new AVATARS binding), then the site (usercontent serves /attachments), then api, then integrations (prompt) and the docs; g1t attach and --attach ship with the next CLI release, which cloud sessions need before the prompt's advice works. Checked: cargo test -p g1t-contracts (319), -p g1t-work (209), -p g1t-api (211; openapi regenerated with G1T_WRITE_OPENAPI), -p g1t (112 + 51; ops_table and cli_reference regenerated with G1T_WRITE_OPS and G1T_WRITE_DOCS), -p g1t-integrations (76); cargo check -p g1t-api, -p g1t-work and -p g1t-integrations --target wasm32-unknown-unknown; node --test apps/web/app/lib/token-scopes.test.ts.2d ago

webhooks.server.tsWebhooks: every event, to your own addresses, signed and retried2w ago

website-token.test.tsThe phone app: Chat in full with phone notifications, plus Code and Customize (#4 )6d ago

website-token.tsThe phone app: Chat in full with phone notifications, plus Code and Customize (#4 )6d ago

workspace-choice.test.tsUsage, Billing settings and prepaid AI credit; fixes from the UX audit1w ago

workspace-choice.tsUsage, Billing settings and prepaid AI credit; fixes from the UX audit1w ago

workspace-deletion.test.tsPackages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member1w ago

workspace-deletion.tsPackages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member1w ago

workspace-gate.test.tsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.5d ago

workspace-gate.tsg1t is described as the product it is in alpha, and its shell takes the new shape: the README, the docs home and docs/PLAN.md label every feature Live, Preview or Coming; a floating dock with the g1t mark, Today, Chat, Notifications, Agents, Code, Artifacts, your pinned apps and the Apps launcher, with People, Workspace and the account menu (now holding help) at its foot; the workspace's logo and switcher heading a flat sidebar that folds away with Ctrl B, or leading the header's breadcrumbs where there is none; the page in a rounded panel with a full-width header; Today as the front page, with first-time acceptance of agents' pull requests, what needs you and what's at stake, spend today and one item to start with, every number from a service; Notifications in place of the Inbox at /notifications; a bottom bar and More sheet on phones; and sign-in, sign-up, two-factor, reset, invites and choosing a workspace on standalone pages with no app around them.1w ago

workspace-insights.test.tsAgents in any harness coordinate through one inbox, the merge queue lands only what an attempt itself tested, and workspaces have Insights. Each session keeps the coordination updates that concern it (session_updates, migration 0038): claims, releases, hand-offs and expiries, overlaps, dependency changes, decisions, integration results and sessions starting and ending, about the issues and pull requests it is on or tied to, or every one while it is on nothing yet; never its own start, end or claims. An update is written in the same batch as the change it announces, guarded so it exists only once the change is made, with an id derived from the change, so a retry or two requests at once deliver it once; a blocker clearing and queue results follow their change under a key that names it. get_session's `updates` lists up to 100, oldest first, and reading never clears them; record_session_entries' `acknowledge` clears exact IDs for that session only and says how many were its (`acknowledged`). The CLI has `g1t session updates` and `g1t session acknowledge ID --note` (the decision note and the acknowledgement in one idempotent call, `--session` over G1T_SESSION), and those and `heartbeat` say which files the working copy changed (git status; G1T_TOUCH=0 turns it off), so overlapping edits are noticed whichever harness makes them; a pair of sessions is told again when another file in common joins. The Claude Code hooks fetch the inbox in the background, show it at every prompt and before tools when it changes or every two minutes, refresh after an acknowledgement, and never acknowledge. The AGENTS.md snippet, the skills, MCP's instructions and llms.txt say the same steps for every harness: scope claims (Scope, Interfaces, Depends on, Checks), proceed on complementary work, agree ownership before overlapping edits, record public summaries and evidence, never private reasoning. Claims resolve ownership in one statement, so simultaneous claims by one session converge. The merge queue: every attempt has its own id and branch (0037); a claimed build, an accepted build, a report and a callback each check their attempt; a dispatch registers its runs first and Actions fails the dispatch when it cannot; an attempt is decided only by its own runs (and other tools' statuses set after it was claimed), never an earlier attempt's or the pull request's own on the same commit (0040 keeps what it built and how many runs it registered); landing re-reads the rules, honours a bypass asked for when queued (0039), checks no predecessor remains, carries the exact tested commit to Git's compare-and-swap, records publication intent that fences removal, closing, drafting and changing base, resumes after a crash, and a publisher that lost a race no longer undoes the winner's landing; a retried attempt's branch is deleted; an entry whose branch or pull request is gone leaves the queue with a note instead of blocking those behind it; diagnosis rebuilds exactly what was tested and names a failing combination, not a proven cause. A session's page shows what awaits its agent as the coordination feed does. Workspace Insights (/:owner/-/insights): delivery metrics across a workspace's readable repositories (work service delivery_insights). Deploy Work migrations 0037 to 0040 first (the pipeline does), then Repos and Actions before Work, and the site with Work. Checked: cargo test for g1t-work (143), g1t-repos (311), g1t-actions-service (108), g1t-api (201), g1t (149), g1t-rules (72), g1t-contracts (306) and wasm32 builds; cargo check --workspace; the conformance suite against freshly built API and Work Workers (5, none skipped); the Claude Code hooks (49); the site's typecheck, tests (919) and a production build; the deploy tool's 92 tests; the docs build; and two unscripted local rehearsals in which two concurrent agents, given only a task and AGENTS.md, found each other, agreed a split and made complementary edits.3d ago

workspace-insights.tsAgents in any harness coordinate through one inbox, the merge queue lands only what an attempt itself tested, and workspaces have Insights. Each session keeps the coordination updates that concern it (session_updates, migration 0038): claims, releases, hand-offs and expiries, overlaps, dependency changes, decisions, integration results and sessions starting and ending, about the issues and pull requests it is on or tied to, or every one while it is on nothing yet; never its own start, end or claims. An update is written in the same batch as the change it announces, guarded so it exists only once the change is made, with an id derived from the change, so a retry or two requests at once deliver it once; a blocker clearing and queue results follow their change under a key that names it. get_session's `updates` lists up to 100, oldest first, and reading never clears them; record_session_entries' `acknowledge` clears exact IDs for that session only and says how many were its (`acknowledged`). The CLI has `g1t session updates` and `g1t session acknowledge ID --note` (the decision note and the acknowledgement in one idempotent call, `--session` over G1T_SESSION), and those and `heartbeat` say which files the working copy changed (git status; G1T_TOUCH=0 turns it off), so overlapping edits are noticed whichever harness makes them; a pair of sessions is told again when another file in common joins. The Claude Code hooks fetch the inbox in the background, show it at every prompt and before tools when it changes or every two minutes, refresh after an acknowledgement, and never acknowledge. The AGENTS.md snippet, the skills, MCP's instructions and llms.txt say the same steps for every harness: scope claims (Scope, Interfaces, Depends on, Checks), proceed on complementary work, agree ownership before overlapping edits, record public summaries and evidence, never private reasoning. Claims resolve ownership in one statement, so simultaneous claims by one session converge. The merge queue: every attempt has its own id and branch (0037); a claimed build, an accepted build, a report and a callback each check their attempt; a dispatch registers its runs first and Actions fails the dispatch when it cannot; an attempt is decided only by its own runs (and other tools' statuses set after it was claimed), never an earlier attempt's or the pull request's own on the same commit (0040 keeps what it built and how many runs it registered); landing re-reads the rules, honours a bypass asked for when queued (0039), checks no predecessor remains, carries the exact tested commit to Git's compare-and-swap, records publication intent that fences removal, closing, drafting and changing base, resumes after a crash, and a publisher that lost a race no longer undoes the winner's landing; a retried attempt's branch is deleted; an entry whose branch or pull request is gone leaves the queue with a note instead of blocking those behind it; diagnosis rebuilds exactly what was tested and names a failing combination, not a proven cause. A session's page shows what awaits its agent as the coordination feed does. Workspace Insights (/:owner/-/insights): delivery metrics across a workspace's readable repositories (work service delivery_insights). Deploy Work migrations 0037 to 0040 first (the pipeline does), then Repos and Actions before Work, and the site with Work. Checked: cargo test for g1t-work (143), g1t-repos (311), g1t-actions-service (108), g1t-api (201), g1t (149), g1t-rules (72), g1t-contracts (306) and wasm32 builds; cargo check --workspace; the conformance suite against freshly built API and Work Workers (5, none skipped); the Claude Code hooks (49); the site's typecheck, tests (919) and a production build; the deploy tool's 92 tests; the docs build; and two unscripted local rehearsals in which two concurrent agents, given only a task and AGENTS.md, found each other, agreed a split and made complementary edits.3d ago

workspace-nav.test.tsMerge main into project-descriptions2d ago

workspace-nav.tsMerge main into project-descriptions2d ago

workspace-projects.server.tsProjects show what they are by their icon: the projects list and grid, and the sidebar's Pinned, Recent and Projects rows, drew the same box for every project. ProjectIcon (components/project-icon.tsx) now picks one from Project.kind: a window for an app or site, books for a library, a terminal for a tool or CLI, an open book for documentation, and the box for anything else. The sidebar still shows a lock for a private project, where nothing else on its row says so; the list keeps its private badge beside the name. Sidebar shortcuts (shortcutOf, ShortcutProject) carry the kind. Checked: apps/web typecheck, lib/pins tests.2d ago

workspace-usage.server.tsMerge the workspace shell: navigation and phone shell, g1t as orchestrator, agents in roles with audience-checked reads, reactions and custom emoji, live notifications and browser push, the homepage tour (agents 0002, chat 0002)1w ago

write-up.test.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

write-up.tsThe site no longer has g1t's own agents. g1t is where the coding agents you already use do their work, so the Agents mode is gone from the rail, with every page under /<workspace>/-/agents: personas, faces, skills, abilities, routines, computers, sessions and agent spend. Customize (skills, connectors, plugins, marketplaces), the context hub, model providers, the AI Gateway page, the Spend page and pill, Ask g1t, and the repository's Agents, Plans and agent settings pages go too, and their addresses now answer 404.5d ago

zip.test.tsMerge GitHub pull request #8 (claude/cool-newton-7nm8r9): Docker deployments on g1t.page behind a switch that stays off until Cloudflare's container APIs are checked live, workflows a workspace manages in its repositories, tarball and zipball downloads, refs/pull/* kept to g1t alone, framework detection for g1t.page builds, and a workflow's event payload matching GitHub's.5d ago

zip.tsMerge GitHub pull request #8 (claude/cool-newton-7nm8r9): Docker deployments on g1t.page behind a switch that stays off until Cloudflare's container APIs are checked live, workflows a workspace manages in its repositories, tarball and zipball downloads, refs/pull/* kept to g1t alone, framework detection for g1t.page builds, and a workflow's event payload matching GitHub's.5d ago