Pick any line to see why it is the way it is: the commit, the pull request and issue it came from, and what the agent was thinking.
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 1 | //! Memory that fills itself (see `g1t_contracts::capture`). |
| 2 | //! | |
| 3 | //! Candidates arrive from four places: | |
| 4 | //! | |
| 5 | //! - **Runs.** At the end of a run the harness asks the agent what it | |
| 6 | //! learned and reports it with the run's token (`report_learned`). | |
| 7 | //! - **Reviews.** A person's request for changes, or a comment that | |
| 8 | //! corrects the agent ("use the shared client instead"), on a pull | |
| 9 | //! request, becomes a convention candidate quoting the comment. | |
| 10 | //! - **Merges.** A merged pull request's title, why and files become a | |
| 11 | //! decision candidate. No model is asked: the pull request's own words | |
| 12 | //! are the decision, and a candidate waits for review anyway. | |
| 13 | //! - **Docs.** The context service reads a project's README, AGENTS.md, | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 14 | //! CONTRIBUTING, docs on how to work in it, and manifests, and sends |
| 15 | //! what they say (`capture_memories`). Once it has read every file, the | |
| 16 | //! waiting candidates its docs no longer suggest are let go | |
| 17 | //! (`prune_doc_candidates`). | |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 18 | //! |
| 19 | //! The same thing said again, by an independent source, is one memory seen | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 20 | //! twice, which keeps it (`promotes`). The same thing in other words (a |
| 21 | //! sentence that grew, one cut short) is the memory already there | |
| 22 | //! (`same_memory`). A dismissed memory stays dismissed, so the same thing | |
| 23 | //! is never suggested again. Nothing that looks like a secret is stored, in | |
| 24 | //! the text or in its evidence. | |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 25 | |
| 26 | use std::collections::HashMap; | |
| 27 | ||
| 28 | use g1t_contracts::agents::*; | |
| 29 | use g1t_contracts::capture::*; | |
| 30 | use g1t_contracts::events::{Event, NewEvent, Publish}; | |
| 31 | use g1t_contracts::repos::{PathByIdArgs, RepoPath}; | |
| 32 | use g1t_contracts::time::rfc3339; | |
| 33 | use g1t_contracts::work::{Comment, CommentKind, Pull, PullStatus, Verdict}; | |
| 34 | use g1t_contracts::{FailureCode, Outcome, PrincipalKind, new_id}; | |
| 35 | use g1t_kit::now_ms; | |
| 36 | use serde::Deserialize; | |
| 37 | use worker::Result; | |
| 38 | use worker::wasm_bindgen::JsValue; | |
| 39 | ||
| 40 | use crate::Work; | |
| Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step | 41 | use crate::rows::PULL_COLUMNS; |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 42 | use crate::checks::hash; |
| 43 | use crate::rows::{CommentRow, PullRow}; | |
| 44 | ||
| 45 | /// The most memories one project, or one workspace, keeps (as memory.rs). | |
| 46 | const MAX_PER_SCOPE: u32 = 500; | |
| 47 | /// How sure g1t is of what an agent says it learned. | |
| 48 | const RUN_CONFIDENCE: f64 = 0.6; | |
| 49 | /// Of a person's correction in a review. | |
| 50 | const REVIEW_CONFIDENCE: f64 = 0.7; | |
| 51 | /// Of a merged pull request's decision, summed up from its own words. | |
| 52 | const MERGE_CONFIDENCE: f64 = 0.5; | |
| 53 | /// The longest correction or decision kept, in characters. | |
| 54 | const MAX_SUMMARY_CHARS: usize = 400; | |
| 55 | /// The most files a decision names. | |
| 56 | const MAX_FILES_NAMED: usize = 6; | |
| 57 | /// Who people's and agents' names are not. | |
| g1t is one name: its agent's work, commits and comments show as @g1t, and nobody can claim g1t or g1t-agent | 58 | const NOT_PEOPLE: [&str; 2] = ["g1t", "agent"]; |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 59 | |
| 60 | /// Words that mark a sentence as telling the agent how things are done. | |
| 61 | const CORRECTIVE: [&str; 20] = [ | |
| 62 | "instead", | |
| 63 | "don't", | |
| 64 | "do not", | |
| 65 | "never", | |
| 66 | "always", | |
| 67 | "please use", | |
| 68 | "we use", | |
| 69 | "use the", | |
| 70 | "should", | |
| 71 | "shouldn't", | |
| 72 | "prefer", | |
| 73 | "avoid", | |
| 74 | "must", | |
| 75 | "rather than", | |
| 76 | "convention", | |
| 77 | "not how", | |
| 78 | "we don't", | |
| 79 | "make sure", | |
| 80 | "remember to", | |
| 81 | "the right way", | |
| 82 | ]; | |
| 83 | ||
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 84 | #[derive(Clone, Deserialize)] |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 85 | struct Existing { |
| 86 | id: String, | |
| 87 | status: String, | |
| 88 | sources: String, | |
| 89 | confidence: Option<f64>, | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 90 | text: String, |
| 91 | } | |
| 92 | ||
| 93 | /// The most memories of one scope compared with a new one for being the | |
| 94 | /// same thing in other words: every kept and waiting one (at most 500) and | |
| 95 | /// the most recently dismissed. | |
| 96 | const MAX_COMPARED: u32 = 2000; | |
| 97 | ||
| 98 | /// The memory in `among` that says the same thing as `text`, if any. | |
| 99 | fn same_as<'a>(among: &'a [Existing], text: &str) -> Option<&'a Existing> { | |
| 100 | among.iter().find(|existing| same_memory(&existing.text, text)) | |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 101 | } |
| 102 | ||
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 103 | /// Whether `existing` is a waiting candidate that the source `reference` |
| 104 | /// said before and now says in other words (`print`). | |
| 105 | fn reworded_by(existing: &Existing, reference: &str, print: &str) -> bool { | |
| 106 | let sources: Vec<String> = serde_json::from_str(&existing.sources).unwrap_or_default(); | |
| 107 | existing.status == MemoryStatus::Candidate.as_str() | |
| 108 | && fingerprint(&existing.text) != print | |
| 109 | && sources.iter().any(|source| source == reference) | |
| 110 | } | |
| 111 | ||
| 112 | /// Whether a candidate came from docs and manifests alone. | |
| 113 | fn only_from_docs(sources: &str) -> bool { | |
| 114 | let list: Vec<String> = serde_json::from_str(sources).unwrap_or_default(); | |
| 115 | !list.is_empty() && list.iter().all(|source| source.starts_with("doc:")) | |
| 116 | } | |
| 117 | ||
| 118 | /// Of a project's waiting doc candidates, the ids its docs no longer | |
| 119 | /// suggest: none of `texts` is the same memory. | |
| 120 | fn unsuggested(candidates: &[Existing], texts: &[String]) -> Vec<String> { | |
| 121 | candidates | |
| 122 | .iter() | |
| 123 | .filter(|candidate| only_from_docs(&candidate.sources)) | |
| 124 | .filter(|candidate| !texts.iter().any(|text| same_memory(&candidate.text, text))) | |
| 125 | .map(|candidate| candidate.id.clone()) | |
| 126 | .collect() | |
| 127 | } | |
| 128 | ||
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 129 | #[derive(Deserialize)] |
| 130 | struct RunTicket { | |
| 131 | id: String, | |
| 132 | workspace: String, | |
| 133 | repo_id: String, | |
| 134 | number: Option<u32>, | |
| 135 | token_hash: String, | |
| 136 | } | |
| 137 | ||
| 138 | /// `text` cut to `max` characters at a word, with an ellipsis if cut. | |
| 139 | fn clip(text: &str, max: usize) -> String { | |
| 140 | let text = text.split_whitespace().collect::<Vec<_>>().join(" "); | |
| 141 | if text.chars().count() <= max { | |
| 142 | return text; | |
| 143 | } | |
| 144 | let cut: String = text.chars().take(max - 1).collect(); | |
| 145 | let cut = cut.rsplit_once(' ').map_or(cut.as_str(), |(head, _)| head).to_owned(); | |
| 146 | format!("{cut}…") | |
| 147 | } | |
| 148 | ||
| 149 | /// Markdown reduced to plain prose: no code blocks, quotes, headings or | |
| 150 | /// list markers. | |
| 151 | fn prose(markdown: &str) -> Vec<String> { | |
| 152 | let mut lines = Vec::new(); | |
| 153 | let mut fenced = false; | |
| 154 | for line in markdown.lines() { | |
| 155 | let trimmed = line.trim(); | |
| 156 | if trimmed.starts_with("```") || trimmed.starts_with("~~~") { | |
| 157 | fenced = !fenced; | |
| 158 | continue; | |
| 159 | } | |
| 160 | if fenced || trimmed.starts_with('>') || trimmed.starts_with('#') || trimmed.starts_with("<!--") { | |
| 161 | lines.push(String::new()); | |
| 162 | continue; | |
| 163 | } | |
| 164 | let item = trimmed.trim_start_matches(['-', '*', '+']).trim_start(); | |
| 165 | lines.push(item.to_owned()); | |
| 166 | } | |
| 167 | lines | |
| 168 | } | |
| 169 | ||
| 170 | /// The sentences of a comment, as plain prose. | |
| 171 | fn sentences(body: &str) -> Vec<String> { | |
| 172 | let mut out = Vec::new(); | |
| 173 | for line in prose(body) { | |
| 174 | let mut current = String::new(); | |
| 175 | for c in line.chars() { | |
| 176 | current.push(c); | |
| 177 | if matches!(c, '.' | '!' | '?') { | |
| 178 | let sentence = current.trim().to_owned(); | |
| 179 | if !sentence.is_empty() { | |
| 180 | out.push(sentence); | |
| 181 | } | |
| 182 | current.clear(); | |
| 183 | } | |
| 184 | } | |
| 185 | let rest = current.trim(); | |
| 186 | if !rest.is_empty() { | |
| 187 | out.push(rest.to_owned()); | |
| 188 | } | |
| 189 | } | |
| 190 | out | |
| 191 | } | |
| 192 | ||
| 193 | /// Whether a sentence tells the agent how something is done here. | |
| 194 | fn corrective(sentence: &str) -> bool { | |
| 195 | let lower = format!(" {} ", sentence.to_lowercase()); | |
| 196 | !sentence.trim_end().ends_with('?') && CORRECTIVE.iter().any(|word| lower.contains(word)) | |
| 197 | } | |
| 198 | ||
| 199 | /// What a person's review comment teaches, as one convention: its first | |
| 200 | /// sentence that says how things are done, or, for a request for changes, | |
| 201 | /// its first real sentence. None when it teaches nothing reusable: a | |
| 202 | /// question, a thank-you, a nit too short to mean anything. | |
| 203 | pub(crate) fn correction(body: &str, requested_changes: bool) -> Option<String> { | |
| 204 | let all = sentences(body); | |
| 205 | let found = all | |
| 206 | .iter() | |
| 207 | .find(|sentence| sentence.chars().count() >= 15 && corrective(sentence)) | |
| 208 | .or_else(|| { | |
| 209 | requested_changes.then(|| { | |
| 210 | all.iter() | |
| 211 | .find(|sentence| sentence.chars().count() >= 20 && !sentence.ends_with('?')) | |
| 212 | })? | |
| 213 | })?; | |
| 214 | Some(clip(found, MAX_SUMMARY_CHARS)) | |
| 215 | } | |
| 216 | ||
| 217 | /// A merged pull request's decision, from its own words: its title, the | |
| 218 | /// first paragraph of its description (why), and the files it changed. | |
| 219 | /// None when it says no more than its title. | |
| 220 | pub(crate) fn decision(number: u32, title: &str, body: Option<&str>, files: &[String]) -> Option<String> { | |
| 221 | let body = body.unwrap_or_default(); | |
| 222 | let mut paragraph = String::new(); | |
| 223 | for line in prose(body) { | |
| 224 | if line.trim().is_empty() { | |
| 225 | if !paragraph.is_empty() { | |
| 226 | break; | |
| 227 | } | |
| 228 | continue; | |
| 229 | } | |
| 230 | // "Summary:" and the like say nothing by themselves. | |
| 231 | if paragraph.is_empty() && line.trim_end_matches(':').split_whitespace().count() <= 1 { | |
| 232 | continue; | |
| 233 | } | |
| 234 | if !paragraph.is_empty() { | |
| 235 | paragraph.push(' '); | |
| 236 | } | |
| 237 | paragraph.push_str(line.trim()); | |
| 238 | } | |
| 239 | if paragraph.chars().count() < 30 { | |
| 240 | return None; | |
| 241 | } | |
| 242 | let why = clip(¶graph, MAX_SUMMARY_CHARS); | |
| 243 | let mut text = format!("Decided in #{number}, \"{}\": {why}", clip(title, 120)); | |
| 244 | if !files.is_empty() { | |
| 245 | let named: Vec<&str> = files.iter().take(MAX_FILES_NAMED).map(String::as_str).collect(); | |
| 246 | let more = files.len().saturating_sub(MAX_FILES_NAMED); | |
| 247 | text.push_str(&format!( | |
| 248 | " Changed {}{}.", | |
| 249 | named.join(", "), | |
| 250 | if more > 0 { format!(" and {more} more") } else { String::new() } | |
| 251 | )); | |
| 252 | } | |
| 253 | Some(clip(&text, MAX_MEMORY_CHARS)) | |
| 254 | } | |
| 255 | ||
| 256 | /// The JSON list of sources, with `reference` added once. | |
| 257 | fn with_source(sources: &str, reference: &str) -> Vec<String> { | |
| 258 | let mut list: Vec<String> = serde_json::from_str(sources).unwrap_or_default(); | |
| 259 | if !list.iter().any(|seen| seen == reference) { | |
| 260 | list.push(reference.to_owned()); | |
| 261 | } | |
| 262 | list | |
| 263 | } | |
| 264 | ||
| 265 | /// An item tidied, or why it cannot be kept. | |
| 266 | fn checked(item: &CaptureItem) -> std::result::Result<(String, Option<String>), &'static str> { | |
| 267 | let text = item.text.split_whitespace().collect::<Vec<_>>().join(" "); | |
| 268 | if text.chars().count() < 8 { | |
| 269 | return Err("too short"); | |
| 270 | } | |
| 271 | if text.chars().count() > MAX_MEMORY_CHARS { | |
| 272 | return Err("too long"); | |
| 273 | } | |
| 274 | if secret_in(&text).is_some() { | |
| 275 | return Err("looks like a secret"); | |
| 276 | } | |
| 277 | let evidence = item | |
| 278 | .evidence | |
| 279 | .as_deref() | |
| 280 | .map(|evidence| clip(evidence, MAX_EVIDENCE_CHARS)) | |
| 281 | .filter(|evidence| !evidence.is_empty()); | |
| 282 | // Evidence quoting a key would put the key in memory all the same. | |
| 283 | if evidence.as_deref().is_some_and(|evidence| secret_in(evidence).is_some()) { | |
| 284 | return Err("looks like a secret"); | |
| 285 | } | |
| 286 | Ok((text, evidence)) | |
| 287 | } | |
| 288 | ||
| 289 | impl Work { | |
| 290 | /// The `owner/name` of a repository, by id, remembered for one capture. | |
| 291 | async fn path_of(&self, cache: &mut HashMap<String, Option<RepoPath>>, repo_id: &str) -> Result<Option<RepoPath>> { | |
| 292 | if let Some(found) = cache.get(repo_id) { | |
| 293 | return Ok(found.clone()); | |
| 294 | } | |
| 295 | let found: Option<RepoPath> = | |
| 296 | g1t_kit::call(&self.repos, "path_by_id", &PathByIdArgs { id: repo_id.to_owned() }).await?; | |
| 297 | cache.insert(repo_id.to_owned(), found.clone()); | |
| 298 | Ok(found) | |
| 299 | } | |
| 300 | ||
| 301 | /// Tells subscribers (the context service's search) that a memory | |
| 302 | /// changed. Never fails what changed it. | |
| 303 | pub(crate) async fn memory_changed(&self, id: &str, workspace: &str, status: &str, repo_id: Option<&str>) { | |
| 304 | let event = NewEvent { | |
| 305 | kind: "memory.changed", | |
| 306 | source: "work", | |
| 307 | repo_id: repo_id.map(str::to_owned), | |
| 308 | actor: None, | |
| 309 | data: MemoryChanged { | |
| 310 | memory_id: id.to_owned(), | |
| 311 | workspace: workspace.to_owned(), | |
| 312 | status: status.to_owned(), | |
| 313 | }, | |
| 314 | }; | |
| 315 | let sent: Result<()> = g1t_kit::call(&self.events, "publish", &Publish { events: vec![event] }).await; | |
| 316 | if let Err(error) = sent { | |
| 317 | worker::console_error!("memory.changed for {id} was not published: {error}"); | |
| 318 | } | |
| 319 | } | |
| 320 | ||
| 321 | /// Adds candidates, or counts them as another sighting of what is there. | |
| 322 | pub(crate) async fn capture(&self, workspace: &str, items: &[CaptureItem], by: &str) -> Result<Captured> { | |
| 323 | let workspace = workspace.to_lowercase(); | |
| 324 | let mut done = Captured::default(); | |
| 325 | let mut paths = HashMap::new(); | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 326 | // Each scope's memories, read once a capture, for near-duplicates. |
| 327 | let mut scopes: HashMap<(&'static str, String), Vec<Existing>> = HashMap::new(); | |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 328 | for item in items.iter().take(MAX_CAPTURE) { |
| 329 | let Ok((text, evidence)) = checked(item) else { | |
| 330 | done.refused += 1; | |
| 331 | continue; | |
| 332 | }; | |
| 333 | let (key, repo) = match (item.scope, &item.repo_id) { | |
| 334 | (MemoryScope::Workspace, _) => (workspace.clone(), item.repo_id.clone()), | |
| 335 | (MemoryScope::Project, Some(repo_id)) => (repo_id.clone(), Some(repo_id.clone())), | |
| 336 | (MemoryScope::Project, None) => { | |
| 337 | done.refused += 1; | |
| 338 | continue; | |
| 339 | } | |
| 340 | }; | |
| 341 | let path = match &repo { | |
| 342 | Some(repo_id) => self.path_of(&mut paths, repo_id).await?, | |
| 343 | None => None, | |
| 344 | }; | |
| 345 | // A project's memory belongs to a repository of this workspace. | |
| 346 | if item.scope == MemoryScope::Project | |
| 347 | && !path.as_ref().is_some_and(|path| path.namespace.to_lowercase() == workspace) | |
| 348 | { | |
| 349 | done.refused += 1; | |
| 350 | continue; | |
| 351 | } | |
| 352 | let print = fingerprint(&text); | |
| 353 | let now = rfc3339(now_ms()); | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 354 | let mut existing = self |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 355 | .db |
| 356 | .prepare( | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 357 | "SELECT id, status, sources, confidence, text FROM memories |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 358 | WHERE scope = ? AND scope_key = ? AND (fingerprint = ? OR text = ?) LIMIT 1", |
| 359 | ) | |
| 360 | .bind(&[item.scope.as_str().into(), key.as_str().into(), print.as_str().into(), text.as_str().into()])? | |
| 361 | .first::<Existing>(None) | |
| 362 | .await?; | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 363 | // The same thing in other words: a sentence that grew, or one |
| 364 | // cut short, is the memory already there, kept, waiting or | |
| 365 | // dismissed. | |
| 366 | let scope_key = (item.scope.as_str(), key.clone()); | |
| 367 | if existing.is_none() { | |
| 368 | if !scopes.contains_key(&scope_key) { | |
| 369 | let rows = self | |
| 370 | .db | |
| 371 | .prepare( | |
| 372 | "SELECT id, status, sources, confidence, text FROM memories | |
| 373 | WHERE scope = ? AND scope_key = ? | |
| 374 | ORDER BY status = 'dismissed', updated_at DESC LIMIT ?", | |
| 375 | ) | |
| 376 | .bind(&[item.scope.as_str().into(), key.as_str().into(), MAX_COMPARED.into()])? | |
| 377 | .all() | |
| 378 | .await? | |
| 379 | .results::<Existing>()?; | |
| 380 | scopes.insert(scope_key.clone(), rows); | |
| 381 | } | |
| 382 | existing = scopes.get(&scope_key).and_then(|among| same_as(among, &text)).cloned(); | |
| 383 | } | |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 384 | if let Some(existing) = existing { |
| 385 | done.merged += 1; | |
| 386 | if existing.status == MemoryStatus::Dismissed.as_str() { | |
| 387 | continue; | |
| 388 | } | |
| 389 | let sources = with_source(&existing.sources, &item.reference); | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 390 | // A waiting candidate its own source now says in other words |
| 391 | // (a doc's line read whole, or edited) takes the new words, | |
| 392 | // kind and confidence. Kept memory keeps its words. | |
| 393 | let reworded = reworded_by(&existing, &item.reference, &print); | |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 394 | let confidence = match (existing.confidence, item.confidence) { |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 395 | _ if reworded => item.confidence, |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 396 | (Some(a), Some(b)) => Some(a.max(b)), |
| 397 | (a, b) => a.or(b), | |
| 398 | }; | |
| 399 | let keep = existing.status == MemoryStatus::Candidate.as_str() | |
| 400 | && promotes(sources.len(), item.source, confidence); | |
| 401 | if keep { | |
| 402 | done.kept += 1; | |
| 403 | } | |
| 404 | let status = if keep { MemoryStatus::Kept.as_str() } else { existing.status.as_str() }; | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 405 | let (new_text, new_print) = |
| 406 | if reworded { (text.clone(), print.clone()) } else { (existing.text.clone(), fingerprint(&existing.text)) }; | |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 407 | self.db |
| 408 | .prepare( | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 409 | "UPDATE memories SET sources = ?, confidence = ?, status = ?, text = ?, fingerprint = ?, |
| 410 | kind = COALESCE(?, kind), evidence = COALESCE(?, evidence), updated_at = ? | |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 411 | WHERE id = ?", |
| 412 | ) | |
| 413 | .bind(&[ | |
| 414 | serde_json::to_string(&sources)?.into(), | |
| 415 | confidence.map_or(JsValue::NULL, JsValue::from), | |
| 416 | status.into(), | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 417 | new_text.as_str().into(), |
| 418 | new_print.as_str().into(), | |
| 419 | if reworded { item.kind.as_str().into() } else { JsValue::NULL }, | |
| 420 | if reworded { evidence.clone().map_or(JsValue::NULL, JsValue::from) } else { JsValue::NULL }, | |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 421 | now.as_str().into(), |
| 422 | existing.id.as_str().into(), | |
| 423 | ])? | |
| 424 | .run() | |
| 425 | .await?; | |
| 426 | if keep { | |
| 427 | self.memory_changed(&existing.id, &workspace, status, repo.as_deref()).await; | |
| 428 | } | |
| 429 | continue; | |
| 430 | } | |
| 431 | let count = self | |
| 432 | .db | |
| 433 | .prepare("SELECT count(*) AS value FROM memories WHERE scope = ? AND scope_key = ? AND status != 'dismissed'") | |
| 434 | .bind(&[item.scope.as_str().into(), key.as_str().into()])? | |
| 435 | .first::<u32>(Some("value")) | |
| 436 | .await? | |
| 437 | .unwrap_or_default(); | |
| 438 | if count >= MAX_PER_SCOPE { | |
| 439 | done.refused += 1; | |
| 440 | continue; | |
| 441 | } | |
| 442 | let keep = promotes(1, item.source, item.confidence); | |
| 443 | let status = if keep { MemoryStatus::Kept } else { MemoryStatus::Candidate }; | |
| 444 | let id = new_id("mem", now_ms()); | |
| 445 | let repo_text = path.as_ref().map(|path| format!("{}/{}", path.namespace, path.name)); | |
| 446 | self.db | |
| 447 | .prepare( | |
| 448 | "INSERT INTO memories | |
| 449 | (id, scope, scope_key, workspace, repo, text, kind, source_kind, source_run, | |
| 450 | source_repo, source_number, created_by, pinned, created_at, updated_at, | |
| 451 | status, confidence, source_ref, evidence, fingerprint, sources) | |
| 452 | VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, 0, ?, ?, ?, ?, ?, ?, ?, ?)", | |
| 453 | ) | |
| 454 | .bind(&[ | |
| 455 | id.as_str().into(), | |
| 456 | item.scope.as_str().into(), | |
| 457 | key.as_str().into(), | |
| 458 | workspace.as_str().into(), | |
| 459 | if item.scope == MemoryScope::Project { repo_text.clone().map_or(JsValue::NULL, JsValue::from) } else { JsValue::NULL }, | |
| 460 | text.as_str().into(), | |
| 461 | item.kind.as_str().into(), | |
| 462 | item.source.as_str().into(), | |
| 463 | item.run_id.as_deref().map_or(JsValue::NULL, JsValue::from), | |
| 464 | repo_text.map_or(JsValue::NULL, JsValue::from), | |
| 465 | item.number.map_or(JsValue::NULL, JsValue::from), | |
| 466 | by.into(), | |
| 467 | now.as_str().into(), | |
| 468 | now.as_str().into(), | |
| 469 | status.as_str().into(), | |
| 470 | item.confidence.map_or(JsValue::NULL, JsValue::from), | |
| 471 | item.reference.as_str().into(), | |
| 472 | evidence.map_or(JsValue::NULL, JsValue::from), | |
| 473 | print.as_str().into(), | |
| 474 | serde_json::to_string(&[&item.reference])?.into(), | |
| 475 | ])? | |
| 476 | .run() | |
| 477 | .await?; | |
| 478 | done.added += 1; | |
| 479 | if keep { | |
| 480 | done.kept += 1; | |
| 481 | } | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 482 | // The same thing twice in one capture is one memory. |
| 483 | if let Some(among) = scopes.get_mut(&scope_key) { | |
| 484 | among.push(Existing { | |
| 485 | id: id.clone(), | |
| 486 | status: status.as_str().to_owned(), | |
| 487 | sources: serde_json::to_string(&[&item.reference])?, | |
| 488 | confidence: item.confidence, | |
| 489 | text: text.clone(), | |
| 490 | }); | |
| 491 | } | |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 492 | self.memory_changed(&id, &workspace, status.as_str(), repo.as_deref()).await; |
| 493 | } | |
| 494 | Ok(done) | |
| 495 | } | |
| 496 | ||
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 497 | /// Removes a project's waiting candidates that came from its docs alone |
| 498 | /// and that its docs, as read now, no longer suggest. Kept and dismissed | |
| 499 | /// memory is never touched, nor a candidate another source saw too. | |
| 500 | pub(crate) async fn prune_doc_candidates(&self, a: PruneDocCandidatesArgs) -> Result<Pruned> { | |
| 501 | let workspace = a.workspace.to_lowercase(); | |
| 502 | let candidates = self | |
| 503 | .db | |
| 504 | .prepare( | |
| 505 | "SELECT id, status, sources, confidence, text FROM memories | |
| 506 | WHERE workspace = ? AND scope = 'project' AND scope_key = ? AND status = 'candidate' | |
| 507 | AND source_kind = 'doc' | |
| 508 | LIMIT ?", | |
| 509 | ) | |
| 510 | .bind(&[workspace.as_str().into(), a.repo_id.as_str().into(), MAX_COMPARED.into()])? | |
| 511 | .all() | |
| 512 | .await? | |
| 513 | .results::<Existing>()?; | |
| 514 | let gone = unsuggested(&candidates, &a.texts); | |
| 515 | if gone.is_empty() { | |
| 516 | return Ok(Pruned::default()); | |
| 517 | } | |
| 518 | let removed = self | |
| 519 | .db | |
| 520 | .prepare( | |
| 521 | "DELETE FROM memories WHERE workspace = ? AND scope_key = ? AND status = 'candidate' | |
| 522 | AND id IN (SELECT value FROM json_each(?)) RETURNING id", | |
| 523 | ) | |
| 524 | .bind(&[workspace.as_str().into(), a.repo_id.as_str().into(), serde_json::to_string(&gone)?.into()])? | |
| 525 | .all() | |
| 526 | .await? | |
| 527 | .results::<serde_json::Value>()? | |
| 528 | .len(); | |
| 529 | worker::console_log!("pruned {removed} doc candidates of {} in {workspace}", a.repo_id); | |
| 530 | Ok(Pruned { removed: removed as u32 }) | |
| 531 | } | |
| 532 | ||
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 533 | pub(crate) async fn capture_memories(&self, a: CaptureMemoriesArgs) -> Result<Captured> { |
| 534 | self.capture(&a.workspace, &a.items, a.by.as_deref().unwrap_or("g1t")).await | |
| 535 | } | |
| 536 | ||
| 537 | pub(crate) async fn report_learned(&self, a: ReportLearnedArgs) -> Result<Outcome<Captured>> { | |
| 538 | let run = self | |
| 539 | .db | |
| 540 | .prepare("SELECT id, workspace, repo_id, number, token_hash FROM agent_runs WHERE id = ?") | |
| 541 | .bind(&[a.run_id.as_str().into()])? | |
| 542 | .first::<RunTicket>(None) | |
| 543 | .await? | |
| 544 | .filter(|run| !a.token.is_empty() && run.token_hash == hash(&a.token)); | |
| 545 | let Some(run) = run else { | |
| 546 | return Ok(Outcome::fail(FailureCode::NotFound, "Run not found.")); | |
| 547 | }; | |
| 548 | let items: Vec<CaptureItem> = a | |
| 549 | .items | |
| 550 | .iter() | |
| 551 | .take(MAX_LEARNED) | |
| 552 | .map(|learned| CaptureItem { | |
| 553 | scope: if learned.scope.as_deref() == Some("workspace") { | |
| 554 | MemoryScope::Workspace | |
| 555 | } else { | |
| 556 | MemoryScope::Project | |
| 557 | }, | |
| 558 | repo_id: Some(run.repo_id.clone()), | |
| 559 | kind: learned.kind.as_deref().and_then(MemoryKind::parse).unwrap_or_default(), | |
| 560 | text: learned.text.clone(), | |
| 561 | confidence: Some(RUN_CONFIDENCE), | |
| 562 | source: CaptureSource::Run, | |
| 563 | reference: format!("run:{}", run.id), | |
| 564 | evidence: learned.evidence.clone(), | |
| 565 | number: run.number, | |
| 566 | run_id: Some(run.id.clone()), | |
| 567 | }) | |
| 568 | .collect(); | |
| g1t is one name: its agent's work, commits and comments show as @g1t, and nobody can claim g1t or g1t-agent | 569 | Ok(Outcome::Ok(self.capture(&run.workspace, &items, "g1t").await?)) |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 570 | } |
| 571 | ||
| 572 | pub(crate) async fn list_candidates(&self, a: ListCandidatesArgs) -> Result<Outcome<Vec<Memory>>> { | |
| 573 | let workspace = a.workspace.to_lowercase(); | |
| 574 | if !a.viewer.as_ref().is_some_and(|viewer| viewer.is_member(&workspace)) { | |
| 575 | return Ok(Outcome::fail(FailureCode::Forbidden, "Memory is for members of the workspace and its agents.")); | |
| 576 | } | |
| 577 | let statement = match &a.repo { | |
| 578 | Some(path) => { | |
| 579 | let repo = match self.repo(path, &a.viewer).await? { | |
| 580 | Outcome::Ok(repo) => repo, | |
| 581 | Outcome::Fail(failure) => return Ok(Outcome::Fail(failure)), | |
| 582 | }; | |
| 583 | // The project's own, and the workspace's learned there. | |
| 584 | self.db | |
| 585 | .prepare( | |
| 586 | "SELECT * FROM memories WHERE workspace = ? AND status = 'candidate' | |
| 587 | AND (scope_key = ? OR (scope = 'workspace' AND source_repo = ?)) | |
| 588 | ORDER BY updated_at DESC LIMIT 200", | |
| 589 | ) | |
| 590 | .bind(&[ | |
| 591 | workspace.as_str().into(), | |
| 592 | repo.id.as_str().into(), | |
| 593 | format!("{}/{}", repo.namespace, repo.name).into(), | |
| 594 | ])? | |
| 595 | } | |
| 596 | None => self | |
| 597 | .db | |
| 598 | .prepare("SELECT * FROM memories WHERE workspace = ? AND status = 'candidate' ORDER BY updated_at DESC LIMIT 200") | |
| 599 | .bind(&[workspace.as_str().into()])?, | |
| 600 | }; | |
| 601 | let rows = statement.all().await?.results::<crate::memory::MemoryRow>()?; | |
| 602 | Ok(Outcome::Ok(rows.into_iter().map(Memory::from).collect())) | |
| 603 | } | |
| 604 | ||
| 605 | pub(crate) async fn review_memory(&self, a: ReviewMemoryArgs) -> Result<Outcome<Memory>> { | |
| 606 | let workspace = a.workspace.to_lowercase(); | |
| 607 | if !a.actor.is_member(&workspace) || a.actor.kind != PrincipalKind::User || !a.actor.verified { | |
| 608 | return Ok(Outcome::fail(FailureCode::Forbidden, "Only members of the workspace can review its memory.")); | |
| 609 | } | |
| 610 | let Some(row) = self | |
| 611 | .db | |
| 612 | .prepare("SELECT * FROM memories WHERE id = ? AND workspace = ?") | |
| 613 | .bind(&[a.id.as_str().into(), workspace.as_str().into()])? | |
| 614 | .first::<crate::memory::MemoryRow>(None) | |
| 615 | .await? | |
| 616 | else { | |
| 617 | return Ok(Outcome::fail(FailureCode::NotFound, "Memory not found.")); | |
| 618 | }; | |
| 619 | let text = match a.text.as_deref().map(str::trim).filter(|text| !text.is_empty()) { | |
| 620 | Some(text) => { | |
| 621 | if text.chars().count() > MAX_MEMORY_CHARS { | |
| 622 | return Ok(Outcome::fail(FailureCode::Invalid, format!("Keep a memory to {MAX_MEMORY_CHARS} characters."))); | |
| 623 | } | |
| 624 | if let Some(what) = secret_in(text) { | |
| 625 | return Ok(Outcome::fail( | |
| 626 | FailureCode::Invalid, | |
| 627 | format!("That looks like {what}. Memory never holds secrets; say where the secret lives instead."), | |
| 628 | )); | |
| 629 | } | |
| 630 | Some(text.to_owned()) | |
| 631 | } | |
| 632 | None => None, | |
| 633 | }; | |
| 634 | let status = match a.decision { | |
| 635 | ReviewDecision::Keep => MemoryStatus::Kept, | |
| 636 | ReviewDecision::Dismiss => MemoryStatus::Dismissed, | |
| 637 | }; | |
| 638 | let now = rfc3339(now_ms()); | |
| 639 | let print = text.as_deref().map(fingerprint); | |
| 640 | self.db | |
| 641 | .prepare( | |
| 642 | "UPDATE memories SET status = ?, text = COALESCE(?, text), fingerprint = COALESCE(?, fingerprint), | |
| 643 | kind = COALESCE(?, kind), reviewed_by = ?, reviewed_at = ?, updated_at = ? | |
| 644 | WHERE id = ? AND workspace = ?", | |
| 645 | ) | |
| 646 | .bind(&[ | |
| 647 | status.as_str().into(), | |
| 648 | text.map_or(JsValue::NULL, JsValue::from), | |
| 649 | print.map_or(JsValue::NULL, JsValue::from), | |
| 650 | a.kind.map_or(JsValue::NULL, |kind| kind.as_str().into()), | |
| 651 | a.actor.username.as_str().into(), | |
| 652 | now.as_str().into(), | |
| 653 | now.as_str().into(), | |
| 654 | a.id.as_str().into(), | |
| 655 | workspace.as_str().into(), | |
| 656 | ])? | |
| 657 | .run() | |
| 658 | .await?; | |
| 659 | let repo_id = (row.scope == "project").then(|| row.scope_key.clone()); | |
| 660 | self.memory_changed(&a.id, &workspace, status.as_str(), repo_id.as_deref()).await; | |
| 661 | let reviewed = self | |
| 662 | .db | |
| 663 | .prepare("SELECT * FROM memories WHERE id = ?") | |
| 664 | .bind(&[a.id.as_str().into()])? | |
| 665 | .first::<crate::memory::MemoryRow>(None) | |
| 666 | .await?; | |
| 667 | Ok(match reviewed { | |
| 668 | Some(row) => Outcome::Ok(Memory::from(row)), | |
| 669 | None => Outcome::fail(FailureCode::NotFound, "Memory not found."), | |
| 670 | }) | |
| 671 | } | |
| 672 | ||
| 673 | pub(crate) async fn memories_by_id(&self, a: MemoriesByIdArgs) -> Result<Vec<Memory>> { | |
| 674 | let ids: Vec<&String> = a.ids.iter().take(100).collect(); | |
| 675 | let rows = self | |
| 676 | .db | |
| 677 | .prepare("SELECT * FROM memories WHERE workspace = ? AND id IN (SELECT value FROM json_each(?))") | |
| 678 | .bind(&[a.workspace.to_lowercase().into(), serde_json::to_string(&ids)?.into()])? | |
| 679 | .all() | |
| 680 | .await? | |
| 681 | .results::<crate::memory::MemoryRow>()?; | |
| 682 | Ok(rows.into_iter().map(Memory::from).collect()) | |
| 683 | } | |
| 684 | ||
| 685 | pub(crate) async fn search_memories(&self, a: SearchMemoriesArgs) -> Result<Vec<Memory>> { | |
| 686 | let status = a.status.unwrap_or(MemoryStatus::Kept); | |
| 687 | let limit = a.limit.unwrap_or(20).clamp(1, 100) as usize; | |
| 688 | let statement = match &a.repo_ids { | |
| 689 | Some(ids) => self | |
| 690 | .db | |
| 691 | .prepare( | |
| 692 | "SELECT * FROM memories WHERE workspace = ? AND status = ? | |
| 693 | AND (scope = 'workspace' OR scope_key IN (SELECT value FROM json_each(?))) | |
| 694 | ORDER BY pinned DESC, COALESCE(last_used_at, updated_at) DESC LIMIT 1000", | |
| 695 | ) | |
| 696 | .bind(&[a.workspace.to_lowercase().into(), status.as_str().into(), serde_json::to_string(ids)?.into()])?, | |
| 697 | None => self | |
| 698 | .db | |
| 699 | .prepare( | |
| 700 | "SELECT * FROM memories WHERE workspace = ? AND status = ? | |
| 701 | ORDER BY pinned DESC, COALESCE(last_used_at, updated_at) DESC LIMIT 1000", | |
| 702 | ) | |
| 703 | .bind(&[a.workspace.to_lowercase().into(), status.as_str().into()])?, | |
| 704 | }; | |
| 705 | let words: Vec<String> = a | |
| 706 | .query | |
| 707 | .as_deref() | |
| 708 | .unwrap_or_default() | |
| 709 | .split_whitespace() | |
| 710 | .map(str::to_lowercase) | |
| 711 | .collect(); | |
| 712 | Ok(statement | |
| 713 | .all() | |
| 714 | .await? | |
| 715 | .results::<crate::memory::MemoryRow>()? | |
| 716 | .into_iter() | |
| 717 | .map(Memory::from) | |
| 718 | .filter(|memory| { | |
| 719 | let text = memory.text.to_lowercase(); | |
| 720 | words.iter().all(|word| text.contains(word.as_str())) | |
| 721 | }) | |
| 722 | .take(limit) | |
| 723 | .collect()) | |
| 724 | } | |
| 725 | ||
| 726 | /// What a merged pull request decided, as a candidate. | |
| 727 | fn decision_item(pull: &Pull) -> Option<CaptureItem> { | |
| 728 | let files: Vec<String> = pull.files.iter().map(|file| file.path.clone()).collect(); | |
| 729 | let text = decision(pull.number, &pull.title, pull.body.as_deref(), &files)?; | |
| 730 | Some(CaptureItem { | |
| 731 | scope: MemoryScope::Project, | |
| 732 | repo_id: Some(pull.repo_id.clone()), | |
| 733 | kind: MemoryKind::Decision, | |
| 734 | text, | |
| 735 | confidence: Some(MERGE_CONFIDENCE), | |
| 736 | source: CaptureSource::Pr, | |
| 737 | reference: format!("pull:{}#{}", pull.repo_id, pull.number), | |
| 738 | evidence: Some(clip(&pull.title, MAX_EVIDENCE_CHARS)), | |
| 739 | number: Some(pull.number), | |
| 740 | run_id: None, | |
| 741 | }) | |
| 742 | } | |
| 743 | ||
| 744 | /// What a person's comment on a pull request corrected, as a candidate. | |
| 745 | fn correction_item(pull: &Pull, comment: &Comment) -> Option<CaptureItem> { | |
| 746 | if comment.kind != CommentKind::Comment | |
| 747 | || comment.author.kind == PrincipalKind::Agent | |
| 748 | || NOT_PEOPLE.contains(&comment.author.username.as_str()) | |
| 749 | { | |
| 750 | return None; | |
| 751 | } | |
| 752 | let text = correction(&comment.body, comment.verdict == Some(Verdict::RequestChanges))?; | |
| 753 | Some(CaptureItem { | |
| 754 | scope: MemoryScope::Project, | |
| 755 | repo_id: Some(pull.repo_id.clone()), | |
| 756 | kind: MemoryKind::Convention, | |
| 757 | text, | |
| 758 | confidence: Some(REVIEW_CONFIDENCE), | |
| 759 | source: CaptureSource::Review, | |
| 760 | reference: format!("comment:{}", comment.id), | |
| 761 | evidence: Some(format!( | |
| 762 | "{} on #{}{}: {}", | |
| 763 | comment.author.username, | |
| 764 | pull.number, | |
| 765 | comment.path.as_deref().map(|path| format!(" ({path})")).unwrap_or_default(), | |
| 766 | clip(&comment.body, MAX_EVIDENCE_CHARS - 60) | |
| 767 | )), | |
| 768 | number: Some(pull.number), | |
| 769 | run_id: None, | |
| 770 | }) | |
| 771 | } | |
| 772 | ||
| 773 | async fn workspace_of(&self, repo_id: &str) -> Result<Option<String>> { | |
| 774 | let mut cache = HashMap::new(); | |
| 775 | Ok(self.path_of(&mut cache, repo_id).await?.map(|path| path.namespace.to_lowercase())) | |
| 776 | } | |
| 777 | ||
| 778 | pub(crate) async fn seed_from_pulls(&self, a: SeedFromPullsArgs) -> Result<Captured> { | |
| 779 | let Some(workspace) = self.workspace_of(&a.repo_id).await? else { | |
| 780 | return Ok(Captured::default()); | |
| 781 | }; | |
| 782 | let limit = a.limit.unwrap_or(20).clamp(1, 50); | |
| 783 | let pulls: Vec<Pull> = self | |
| 784 | .db | |
| Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step | 785 | .prepare(format!("SELECT {PULL_COLUMNS} FROM pulls WHERE repo_id = ? AND status = 'merged' ORDER BY merged_at DESC LIMIT ?")) |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 786 | .bind(&[a.repo_id.as_str().into(), limit.into()])? |
| 787 | .all() | |
| 788 | .await? | |
| 789 | .results::<PullRow>()? | |
| 790 | .into_iter() | |
| 791 | .map(Pull::from) | |
| 792 | .collect(); | |
| 793 | let mut items = Vec::new(); | |
| 794 | for pull in &pulls { | |
| 795 | items.extend(Self::decision_item(pull)); | |
| 796 | let comments = self | |
| 797 | .db | |
| 798 | .prepare("SELECT * FROM comments WHERE repo_id = ? AND number = ? ORDER BY id LIMIT 100") | |
| 799 | .bind(&[pull.repo_id.as_str().into(), pull.number.into()])? | |
| 800 | .all() | |
| 801 | .await? | |
| 802 | .results::<CommentRow>()?; | |
| 803 | items.extend( | |
| 804 | comments | |
| 805 | .into_iter() | |
| 806 | .map(Comment::from) | |
| 807 | .filter_map(|comment| Self::correction_item(pull, &comment)), | |
| 808 | ); | |
| 809 | } | |
| 810 | let mut done = Captured::default(); | |
| 811 | for chunk in items.chunks(MAX_CAPTURE) { | |
| 812 | let part = self.capture(&workspace, chunk, "g1t").await?; | |
| 813 | done.added += part.added; | |
| 814 | done.merged += part.merged; | |
| 815 | done.kept += part.kept; | |
| 816 | done.refused += part.refused; | |
| 817 | } | |
| 818 | Ok(done) | |
| 819 | } | |
| 820 | ||
| 821 | /// Captures from the bus: corrections in review, decisions on merge. | |
| 822 | async fn capture_event(&self, event: &Event) -> Result<()> { | |
| 823 | let field = |name: &str| event.data[name].as_str().map(str::to_owned); | |
| 824 | match event.kind.as_str() { | |
| 825 | "pull.merged" => { | |
| 826 | let Some(pull) = (match field("pullId") { | |
| 827 | Some(id) => self.pull_by_id(&id).await?, | |
| 828 | None => None, | |
| 829 | }) else { | |
| 830 | return Ok(()); | |
| 831 | }; | |
| 832 | if pull.status != PullStatus::Merged { | |
| 833 | return Ok(()); | |
| 834 | } | |
| 835 | if let (Some(item), Some(workspace)) = (Self::decision_item(&pull), self.workspace_of(&pull.repo_id).await?) { | |
| 836 | self.capture(&workspace, &[item], "g1t").await?; | |
| 837 | } | |
| 838 | } | |
| 839 | "comment.created" => { | |
| 840 | let (Some(comment_id), Some(pull_id)) = (field("commentId"), field("pullId")) else { | |
| 841 | return Ok(()); | |
| 842 | }; | |
| 843 | let Some(pull) = self.pull_by_id(&pull_id).await? else { | |
| 844 | return Ok(()); | |
| 845 | }; | |
| 846 | let Some(comment) = self | |
| 847 | .db | |
| 848 | .prepare("SELECT * FROM comments WHERE id = ?") | |
| 849 | .bind(&[comment_id.as_str().into()])? | |
| 850 | .first::<CommentRow>(None) | |
| 851 | .await? | |
| 852 | .map(Comment::from) | |
| 853 | else { | |
| 854 | return Ok(()); | |
| 855 | }; | |
| 856 | if let (Some(item), Some(workspace)) = | |
| 857 | (Self::correction_item(&pull, &comment), self.workspace_of(&pull.repo_id).await?) | |
| 858 | { | |
| 859 | self.capture(&workspace, &[item], "g1t").await?; | |
| 860 | } | |
| 861 | } | |
| 862 | _ => {} | |
| 863 | } | |
| 864 | Ok(()) | |
| 865 | } | |
| 866 | } | |
| 867 | ||
| 868 | /// The methods this module serves at `POST /rpc/<method>`. | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 869 | pub(crate) const METHODS: [&str; 8] = [ |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 870 | "capture_memories", |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 871 | "prune_doc_candidates", |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 872 | "report_learned", |
| 873 | "list_candidates", | |
| 874 | "review_memory", | |
| 875 | "memories_by_id", | |
| 876 | "search_memories", | |
| 877 | "seed_from_pulls", | |
| 878 | ]; | |
| 879 | ||
| 880 | pub(crate) async fn dispatch(work: &Work, method: &str, body: serde_json::Value) -> Result<worker::Response> { | |
| 881 | use g1t_kit::{args, reply}; | |
| 882 | match method { | |
| 883 | "capture_memories" => reply(&work.capture_memories(args(body)?).await?), | |
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 884 | "prune_doc_candidates" => reply(&work.prune_doc_candidates(args(body)?).await?), |
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 885 | "report_learned" => reply(&work.report_learned(args(body)?).await?), |
| 886 | "list_candidates" => reply(&work.list_candidates(args(body)?).await?), | |
| 887 | "review_memory" => reply(&work.review_memory(args(body)?).await?), | |
| 888 | "memories_by_id" => reply(&work.memories_by_id(args(body)?).await?), | |
| 889 | "search_memories" => reply(&work.search_memories(args(body)?).await?), | |
| 890 | "seed_from_pulls" => reply(&work.seed_from_pulls(args(body)?).await?), | |
| 891 | _ => worker::Response::error("Unknown method", 404), | |
| 892 | } | |
| 893 | } | |
| 894 | ||
| 895 | /// Captures what an event teaches. Never holds up the queue: a failure is | |
| 896 | /// logged and the event goes on to its other handlers. | |
| 897 | pub(crate) async fn on_event(work: &Work, event: &Event) { | |
| 898 | if let Err(error) = work.capture_event(event).await { | |
| 899 | worker::console_error!("capturing memory from {} {} failed: {error}", event.kind, event.id); | |
| 900 | } | |
| 901 | } | |
| 902 | ||
| 903 | #[cfg(test)] | |
| 904 | mod tests { | |
| 905 | use super::*; | |
| 906 | ||
| 907 | #[test] | |
| 908 | fn a_correction_keeps_the_sentence_that_says_how() { | |
| 909 | let body = "Thanks for this! We use the shared `api` client instead of calling fetch directly. Otherwise looks fine."; | |
| 910 | assert_eq!( | |
| 911 | correction(body, false).as_deref(), | |
| 912 | Some("We use the shared `api` client instead of calling fetch directly.") | |
| 913 | ); | |
| 914 | } | |
| 915 | ||
| 916 | #[test] | |
| 917 | fn questions_and_praise_teach_nothing() { | |
| 918 | assert_eq!(correction("Why not use the cache here?", false), None); | |
| 919 | assert_eq!(correction("LGTM, nice work.", false), None); | |
| 920 | assert_eq!(correction("> never do this\n\nok", false), None); | |
| 921 | assert_eq!(correction("```\nnever run this\n```", false), None); | |
| 922 | } | |
| 923 | ||
| 924 | #[test] | |
| 925 | fn a_request_for_changes_without_a_marker_keeps_its_first_sentence() { | |
| 926 | assert_eq!( | |
| 927 | correction("The migration has to be idempotent for D1 replays.", true).as_deref(), | |
| 928 | Some("The migration has to be idempotent for D1 replays.") | |
| 929 | ); | |
| 930 | assert_eq!(correction("The migration has to be idempotent for D1 replays.", false), None); | |
| 931 | } | |
| 932 | ||
| 933 | #[test] | |
| 934 | fn a_decision_names_why_and_the_files() { | |
| 935 | let files: Vec<String> = (1..=8).map(|n| format!("src/f{n}.rs")).collect(); | |
| 936 | let text = decision( | |
| 937 | 12, | |
| 938 | "Keep the v1 webhook payload", | |
| 939 | Some("## Summary\n\nTwo customers still parse the v1 payload, so it stays alongside v2.\n\nMore detail."), | |
| 940 | &files, | |
| 941 | ) | |
| 942 | .unwrap(); | |
| 943 | assert!(text.starts_with("Decided in #12, \"Keep the v1 webhook payload\": Two customers still parse")); | |
| 944 | assert!(text.contains("src/f1.rs") && text.contains("and 2 more")); | |
| 945 | assert!(!text.contains("More detail")); | |
| 946 | } | |
| 947 | ||
| 948 | #[test] | |
| 949 | fn a_pull_request_that_says_only_its_title_decides_nothing() { | |
| 950 | assert_eq!(decision(3, "Fix typo", Some("Fix typo."), &[]), None); | |
| 951 | assert_eq!(decision(3, "Fix typo", None, &[]), None); | |
| 952 | } | |
| 953 | ||
| 954 | #[test] | |
| 955 | fn sources_are_counted_once_each() { | |
| 956 | assert_eq!(with_source("[\"run:a\"]", "run:a"), ["run:a"]); | |
| 957 | assert_eq!(with_source("[\"run:a\"]", "comment:b"), ["run:a", "comment:b"]); | |
| 958 | assert_eq!(with_source("not json", "doc:x"), ["doc:x"]); | |
| 959 | } | |
| 960 | ||
| Merge memory from docs: only docs on how to work here, whole sentences, no near-duplicates, stale doc suggestions cleared | 961 | fn existing(id: &str, text: &str, sources: &str) -> Existing { |
| 962 | Existing { id: id.into(), status: "candidate".into(), sources: sources.into(), confidence: Some(0.7), text: text.into() } | |
| 963 | } | |
| 964 | ||
| 965 | #[test] | |
| 966 | fn the_same_thing_in_other_words_is_found() { | |
| 967 | let among = [ | |
| 968 | existing("a", "Run cargo test in the crate you changed.", "[]"), | |
| 969 | existing("b", "g1t is a Cargo workspace (apps/api, crates/*, services/actions); `cargo test` runs its tests.", "[]"), | |
| 970 | ]; | |
| 971 | assert_eq!(same_as(&among, "run cargo test in the crate you changed").map(|e| e.id.as_str()), Some("a")); | |
| 972 | assert_eq!( | |
| 973 | same_as(&among, "g1t is a Cargo workspace (apps/*, crates/*, services/*); `cargo test` runs its tests.").map(|e| e.id.as_str()), | |
| 974 | Some("b") | |
| 975 | ); | |
| 976 | assert!(same_as(&among, "Use pnpm, never npm.").is_none()); | |
| 977 | } | |
| 978 | ||
| 979 | #[test] | |
| 980 | fn a_candidate_cut_short_takes_its_doc_s_whole_words() { | |
| 981 | let cut = existing("a", "Work started by an automation cannot retrigger the", "[\"doc:r:CONTRIBUTING.md\"]"); | |
| 982 | let whole = fingerprint("Work started by an automation cannot retrigger the same automation."); | |
| 983 | assert!(reworded_by(&cut, "doc:r:CONTRIBUTING.md", &whole)); | |
| 984 | // Another source saying it is a sighting, not new words. | |
| 985 | assert!(!reworded_by(&cut, "run:x", &whole)); | |
| 986 | // Kept memory keeps the words someone kept. | |
| 987 | let kept = Existing { status: "kept".into(), ..cut.clone() }; | |
| 988 | assert!(!reworded_by(&kept, "doc:r:CONTRIBUTING.md", &whole)); | |
| 989 | // The same words are not new ones. | |
| 990 | assert!(!reworded_by(&cut, "doc:r:CONTRIBUTING.md", &fingerprint(&cut.text))); | |
| 991 | } | |
| 992 | ||
| 993 | #[test] | |
| 994 | fn waiting_doc_candidates_the_docs_no_longer_suggest_are_let_go() { | |
| 995 | let candidates = [ | |
| 996 | // Cut mid-sentence by the old reader; the docs now say it whole. | |
| 997 | existing("cut", "Run the whole suite before you push, every", "[\"doc:r:CONTRIBUTING.md\"]"), | |
| 998 | // From a plan, which is no longer read for memory. | |
| 999 | existing("plan", "Loop protection. Work started by an automation cannot retrigger the", "[\"doc:r:docs/PLAN.md\"]"), | |
| 1000 | // Still suggested. | |
| 1001 | existing("still", "`cargo test` in the crate or service you changed.", "[\"doc:r:CONTRIBUTING.md\"]"), | |
| 1002 | // Seen by a run too: not the docs' alone to take back. | |
| 1003 | existing("run", "Deduplication. The same Sentry issue firing 500 times maps to one", "[\"doc:r:docs/PLAN.md\",\"run:x\"]"), | |
| 1004 | ]; | |
| 1005 | let texts = [ | |
| 1006 | "Run the whole suite before you push, every time, on every branch.".to_owned(), | |
| 1007 | "`cargo test` in the crate or service you changed.".to_owned(), | |
| 1008 | ]; | |
| 1009 | assert_eq!(unsuggested(&candidates, &texts), ["plan"]); | |
| 1010 | assert!(only_from_docs("[\"doc:a\",\"doc:b\"]")); | |
| 1011 | assert!(!only_from_docs("[]")); | |
| 1012 | assert!(!only_from_docs("[\"doc:a\",\"review:b\"]")); | |
| 1013 | } | |
| 1014 | ||
| Agents get guardrails, run credentials, an audit log, a context hub, repository instructions and mentions; security upkeep; snake_case API | 1015 | fn item(text: &str, evidence: Option<&str>) -> CaptureItem { |
| 1016 | CaptureItem { | |
| 1017 | scope: MemoryScope::Project, | |
| 1018 | repo_id: Some("rep_1".into()), | |
| 1019 | kind: MemoryKind::Gotcha, | |
| 1020 | text: text.into(), | |
| 1021 | confidence: Some(0.6), | |
| 1022 | source: CaptureSource::Run, | |
| 1023 | reference: "run:1".into(), | |
| 1024 | evidence: evidence.map(str::to_owned), | |
| 1025 | number: None, | |
| 1026 | run_id: None, | |
| 1027 | } | |
| 1028 | } | |
| 1029 | ||
| 1030 | #[test] | |
| 1031 | fn secrets_are_refused_on_capture_in_text_and_evidence() { | |
| 1032 | // Key-shaped values are joined at run time, as in the contracts' tests. | |
| 1033 | let key = format!("{}{}", "gh", "p_abcdefghijklmnopqrstuvwxyz0123456789"); | |
| 1034 | assert!(checked(&item(&format!("Clone with {key} for the mirror."), None)).is_err()); | |
| 1035 | assert!(checked(&item("The mirror needs a token from the MIRROR_TOKEN secret.", Some(&format!("export T={key}")))).is_err()); | |
| 1036 | let (text, evidence) = checked(&item(" Tests need TZ=UTC. ", Some("cargo test failed on dates"))).unwrap(); | |
| 1037 | assert_eq!(text, "Tests need TZ=UTC."); | |
| 1038 | assert_eq!(evidence.as_deref(), Some("cargo test failed on dates")); | |
| 1039 | assert!(checked(&item("short", None)).is_err()); | |
| 1040 | } | |
| 1041 | } |
This file's history is long; its oldest lines are credited to the oldest commit read.