feat(outcomes): learn accepted and rejected PR patterns by repo

[Khaostica]

Summary

Adds a deterministic, repo-centric PR outcome learner so recommendations reflect what maintainers in a repo actually merge or close — distinct from the existing contributor-centric buildContributorOutcomeHistory.

What changed

• New buildRepoOutcomePatterns (src/signals/engine.ts) buckets a repo's PRs into merged / closed_unmerged / open_active / open_stale and extracts outcome patterns across seven dimensions: path, label, size, linked-issue presence, test evidence, review churn, and author role.
• Maintainer-lane activity (OWNER/MEMBER/COLLABORATOR) is separated from outside-contributor evidence: dimension stats and outsideContributorMergeRate are computed from outside-contributor decided PRs only, with maintainerLaneMergeRate reported separately so maintainer merges never inflate miner-lane evidence.
• Emits successPatterns ("PRs touching src/ merge well here"), riskPatterns ("PRs with no linked issue have high closure risk here"), stale-open-PR risk, and findings (low sample, maintainer separation, stale PRs).
• Feeds success/risk patterns into the contributor decision pack: RepoDecision gains a bounded repoOutcomePatterns summary, and patterns flow into riskReasons / whyThisHelps (risk lines suppressed for maintainer-lane repos).
• New GET /v1/repos/:owner/:repo/outcome-patterns endpoint, RepoOutcomePatterns OpenAPI schema + path, and gittensory_get_repo_outcome_patterns MCP tool.
• Bulk listRepoPullRequestFiles / listRepoPullRequestReviews loaders to analyze the whole repo without per-PR N+1 reads.

Why

Issue #35 (under roadmap #32) requires recommendations and agent next actions to reflect real merge/close outcomes by repo/path/type, keep maintainer-owned activity out of miner-lane evidence, let closed and stale PRs affect risk reasoning, and stay deterministic.

Acceptance criteria → coverage

• "you tend to merge well in this repo/path/type" / "high closure risk" — per-dimension merges_well / high_closure_risk signals with natural-language patterns; asserted in unit fixtures.
• Maintainer-owned activity does not inflate miner-lane evidence — maintainer-lane fixture asserts maintainer merges are excluded from outsideContributorMergeRate, produce no outside success pattern, and raise a maintainer_activity_separated finding.
• Closed and stale open PRs affect risk reasoning — closed-PR and stale-open fixtures assert risk patterns + findings.
• Deterministic — same input twice yields identical output (generatedAt aside); dimensions emitted in a fixed order with stable tie-breaks.

Required test scenarios

Added test/unit/repo-outcome-patterns.test.ts: merged-PR pattern, closed-PR pattern, maintainer-lane, stale-open, deterministic-sorting regression, low-sample, cross-repo isolation, size bucketing, and a forbidden-language assertion. Plus an integration test for the new endpoint and a decision-pack wiring test.

Validation

• npm run typecheck
• npx vitest run (full unit + integration suite) and npx vitest run --config vitest.workers.config.ts
• npx vitest run --coverage — global coverage stays ≥ 95% (lines/statements/functions/branches)
• npm run docs:check, npm run build:mcp, npm audit --audit-level=moderate (0 vulnerabilities)

Safety

• Backend-only change
• No secrets, wallet details, hotkeys, raw trust scores, or private rankings exposed
• Public-surface text audited against forbidden language (wallet, hotkey, raw trust score, payout, reward estimate, farming, private reviewability, public score estimate) with a regression test
• OpenAPI and MCP schemas aligned with behavior
• No changelog change (not a release PR)

Closes #35

[JSONbored]

@Khaostica this needs some work:

Required:

• Resolve conflicts in src/api/routes.ts, src/mcp/server.ts, src/openapi/spec.ts, src/services/decision-pack.ts, and test/unit/openapi.test.ts.
• Move outcome-pattern generation to snapshot/queue-backed data, not request-time broad scans.
• Include files/reviews/checks completeness in the pattern source, or mark incomplete evidence explicitly.
• Ensure all CI passes green when done.

[Khaostica]

@JSONbored All PRs will be updated by 3pm EST tomorrow.

[JSONbored]

@Khaostica This is a useful feature direction, but it is not merge-ready yet.

A few notes:

• Learning accepted/rejected repo outcome patterns is valuable, especially if the output stays private and helps contributors avoid repeating work patterns maintainers have already rejected.
• The shape is broad: this touches API routes, MCP output, OpenAPI schemas, queue processing, decision packs, signal generation, persistence, and new outcome-pattern services. That is fine for this feature, but it raises the review bar because the behavior becomes part of several production surfaces.
• The branch currently conflicts with src/services/decision-pack.ts.
• CI is failing the global branch coverage gate: all tests pass, but branch coverage is 96.89% against the required 97%. The new src/services/repo-outcome-patterns.ts surface is the obvious place to tighten behavioral branch coverage.

Required changes:

• Rebase onto current main and resolve the decision-pack conflict.
• Add targeted tests for the uncovered repo outcome branches instead of lowering or bypassing the coverage gate.
• Re-run validation after the rebase.

Expected validation:

• npm run typecheck
• npm run test:coverage

[JSONbored]

@Khaostica This is ready from my side. The branch is now green and merge-clean, and the latest updates address the prior conflict, coverage, and OpenAPI freshness blockers.

A few notes:

• The repo outcome-pattern surface separates maintainer-lane evidence from outside-contributor evidence, which is the right boundary.
• API, MCP, OpenAPI, queue, decision-pack, persistence, and UI OpenAPI output are updated together.
• The added tests cover service behavior, decision-pack integration, OpenAPI shape, and branch coverage without weakening the gate.

No requested changes from me.

[dosubot]

Documentation Updates

2 document(s) were updated by changes in this PR:

Gittensory Architecture

View Changes

@@ -20,7 +20,7 @@
 
 2. **Frontend Worker** — Lovable/TanStack Start application deployed at `https://gittensory.aethereal.dev/` [[9]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/README.md#L22). This separate Cloudflare Worker under `apps/gittensory-ui` delivers the web interface for miners, maintainers, and administrators.
 
-3. **MCP Package** — `@jsonbored/gittensory-mcp` [[10]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/README.md#L23). The npm package provides local stdio wrapper functionality for coding agents and serves as the first-class base-agent surface. It exposes 20 MCP tools and 13 CLI commands for branch analysis, preflight checks, and agent planning.
+3. **MCP Package** — `@jsonbored/gittensory-mcp` [[10]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/README.md#L23). The npm package provides local stdio wrapper functionality for coding agents and serves as the first-class base-agent surface. It exposes 21 MCP tools and 13 CLI commands for branch analysis, preflight checks, and agent planning.
 
 4. **GitHub App** — Quiet PR inspection [[11]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/README.md#L24). The app provides public-safe sticky comments, maintainer-configured labels, and explicit `@gittensory` commands on installed repos. Default behavior is low-noise: confirmed Gittensor miners receive one sticky comment and a label, while private reviewability and scoring context remains accessible only through the API.
 
@@ -51,7 +51,7 @@
 
 Core signal types are defined in `engine.ts`: `ParticipationLane` enumerates contributor paths (direct_pr, issue_discovery, split, inactive, unknown), `LaneAdvice` provides lane-specific guidance for contributors and maintainers, `CollisionReport` groups collision clusters by risk level (low, medium, high), and `QueueHealth` quantifies burden score, risk level, and open issues/PRs metrics [[15]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/src/signals/engine.ts#L1-L79).
 
-The engine generates eight signal types per registered repo: queue-health, config-quality, label-audit, maintainer-lane, maintainer-cut-readiness, contributor-intake-health, issue-quality, and collision edges [[16]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/src/queue/processors.ts#L394-L471). Supporting modules include `data-quality.ts` for freshness SLO reporting, `reward-risk.ts` for PR reviewability scoring, and `local-branch.ts` for pre-PR diff analysis using metadata-only input.
+The engine generates nine signal types per registered repo: queue-health, config-quality, label-audit, maintainer-lane, maintainer-cut-readiness, contributor-intake-health, issue-quality, contributor-outcome-history, and repo-outcome-patterns [[16]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/src/queue/processors.ts#L394-L471). Supporting modules include `data-quality.ts` for freshness SLO reporting, `reward-risk.ts` for PR reviewability scoring, and `local-branch.ts` for pre-PR diff analysis using metadata-only input.
 
 ### **C. Scoring System (`src/scoring/`)**
 
@@ -95,7 +95,7 @@
 
 ### **H. MCP Package (`packages/gittensory-mcp`)**
 
-The `@jsonbored/gittensory-mcp` package (v0.3.0, Node >= 22) exposes 20 MCP tools in five categories: repo intelligence, PR preflight & scoring, contributor decisions, current branch analysis, and agent operations [[27]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/packages/gittensory-mcp/package.json) [[28]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/packages/gittensory-mcp/bin/gittensory-mcp.js#L156-L458).
+The `@jsonbored/gittensory-mcp` package (v0.3.0, Node >= 22) exposes 21 MCP tools in five categories: repo intelligence, PR preflight & scoring, contributor decisions, current branch analysis, and agent operations [[27]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/packages/gittensory-mcp/package.json) [[28]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/packages/gittensory-mcp/bin/gittensory-mcp.js#L156-L458).
 
 It also provides 13 CLI commands including login, logout, whoami, status, changelog, doctor, init-client, analyze-branch, preflight, and agent plan/status/explain/packet. Core dependencies are `@modelcontextprotocol/sdk` 1.26.0 and `zod` ^3.25.76 [[29]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/packages/gittensory-mcp/package.json#L40-L42).
 
@@ -147,7 +147,7 @@
 
 ### Signal Snapshot Storage
 
-Signals are computed per-repo and persisted as timestamped snapshots with unique IDs [[16]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/src/queue/processors.ts#L394-L471). Eight signal types are generated per registered repository: `queue-health`, `config-quality`, `label-audit`, `maintainer-lane`, `maintainer-cut-readiness`, `contributor-intake-health`, `issue-quality`, and collision edges [[45]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/src/queue/processors.ts#L415-L471). Each snapshot includes a `generatedAt` timestamp and is stored in the `signalSnapshots` table. Freshness states—fresh, stale, partial, degraded—track data quality [[46]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/src/types.ts#L442-L457).
+Signals are computed per-repo and persisted as timestamped snapshots with unique IDs [[16]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/src/queue/processors.ts#L394-L471). Nine signal types are generated per registered repository: `queue-health`, `config-quality`, `label-audit`, `maintainer-lane`, `maintainer-cut-readiness`, `contributor-intake-health`, `issue-quality`, `contributor-outcome-history`, and `repo-outcome-patterns` [[45]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/src/queue/processors.ts#L415-L471). Each snapshot includes a `generatedAt` timestamp and is stored in the `signalSnapshots` table. Freshness states—fresh, stale, partial, degraded—track data quality [[46]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/src/types.ts#L442-L457).
 
 ### Scoring Data Flow
 
@@ -224,7 +224,7 @@
 └── README.md
 ```
 
-The package exposes 20 MCP tools via the Model Context Protocol and 13 CLI commands including login, logout, whoami, status, doctor, analyze-branch, preflight, and agent subcommands [[28]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/packages/gittensory-mcp/bin/gittensory-mcp.js#L156-L458).
+The package exposes 21 MCP tools via the Model Context Protocol and 13 CLI commands including login, logout, whoami, status, doctor, analyze-branch, preflight, and agent subcommands [[28]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/packages/gittensory-mcp/bin/gittensory-mcp.js#L156-L458).
 
 ### Configuration
 
@@ -274,7 +274,7 @@
 ### Communication Paths
 
 **MCP communication**  
-Coding agents (Claude, Cursor, Codex) or CLI users interact with the Worker API via `@jsonbored/gittensory-mcp` [[63]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/packages/gittensory-mcp/package.json#L2-L3). The MCP package exposes 20 tools and 13 CLI commands over stdio or remote transport using `@modelcontextprotocol/sdk` [[29]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/packages/gittensory-mcp/package.json#L40-L42). Each tool calls a specific Worker API endpoint at `gittensory-api.aethereal.dev` [[64]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/wrangler.jsonc#L32), transmitting metadata-only payloads. Authentication uses GitHub Device Flow, embedding session tokens in API requests.
+Coding agents (Claude, Cursor, Codex) or CLI users interact with the Worker API via `@jsonbored/gittensory-mcp` [[63]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/packages/gittensory-mcp/package.json#L2-L3). The MCP package exposes 21 tools and 13 CLI commands over stdio or remote transport using `@modelcontextprotocol/sdk` [[29]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/packages/gittensory-mcp/package.json#L40-L42). Each tool calls a specific Worker API endpoint at `gittensory-api.aethereal.dev` [[64]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/wrangler.jsonc#L32), transmitting metadata-only payloads. Authentication uses GitHub Device Flow, embedding session tokens in API requests.
 
 **GitHub App communication**  
 GitHub events (PR opened/updated/commented, issue activity) trigger webhook POST requests to the Worker [[65]](https://github.com/JSONbored/gittensory/blob/40c019c82d87568c232fbe7fabf50212e983075a/src/index.ts#L12). The handler enqueues a job for asynchronous processing. `processors.ts` routes data events to `github/backfill.ts` for state synchronization and command events to `github/commands.ts`, which parses @gittensory mentions and dispatches to appropriate services. Public-safe responses are written via `github/comments.ts`.

MCP & Agent Integration

View Changes

@@ -10,13 +10,13 @@
 
 The **Model Context Protocol (MCP)** is a standardized interface that allows coding agents and AI tools to connect with external data sources and services in a structured, tool-call-based way. Rather than ad-hoc API integrations, MCP defines a common schema for tool registration, input validation, and structured JSON responses — enabling any MCP-compatible client (Claude Desktop, Cursor, Codex, etc.) to interact with a server like Gittensory without custom wiring.
 
-`@jsonbored/gittensory-mcp` implements MCP using `@modelcontextprotocol/sdk` v1.26.0 [[4]](https://app.dosu.dev/7d8810aa-91c6-4074-849b-92fab65de734/documents/09d20768-c6a2-4ac0-9000-4395a9728dc9#L323-L334). It registers 20 tools across five capability areas and exposes them over standard stdio transport, making Gittensory's intelligence layer accessible to any MCP-compatible client.
+`@jsonbored/gittensory-mcp` implements MCP using `@modelcontextprotocol/sdk` v1.26.0 [[4]](https://app.dosu.dev/7d8810aa-91c6-4074-849b-92fab65de734/documents/09d20768-c6a2-4ac0-9000-4395a9728dc9#L323-L334). It registers 21 tools across five capability areas and exposes them over standard stdio transport, making Gittensory's intelligence layer accessible to any MCP-compatible client.
 
 ### What it provides
 
 The package exposes two surfaces for different audiences:
 
-- **20 MCP tools** — structured tool definitions callable by AI agents via the MCP protocol, covering repo intelligence, PR preflight, scoring, contributor decisions, current-branch analysis, and agent planning [[5]](https://app.dosu.dev/7d8810aa-91c6-4074-849b-92fab65de734/documents/09d20768-c6a2-4ac0-9000-4395a9728dc9#L92-L98)
+- **21 MCP tools** — structured tool definitions callable by AI agents via the MCP protocol, covering repo intelligence, PR preflight, scoring, contributor decisions, current-branch analysis, and agent planning [[5]](https://app.dosu.dev/7d8810aa-91c6-4074-849b-92fab65de734/documents/09d20768-c6a2-4ac0-9000-4395a9728dc9#L92-L98)
 - **13 CLI commands** — human-friendly equivalents of the same capabilities, usable directly from a terminal or CI script
 
 It is designed for three audiences:
@@ -123,7 +123,7 @@
 gittensory-mcp --stdio
 ```
 
-This starts an `McpServer` named `gittensory-local` that communicates via `StdioServerTransport` — reading JSON-RPC messages from stdin and writing responses to stdout [[14]](https://github.com/JSONbored/gittensory/blob/1e7506f4c5226d0b9b5da8fcd219309d84bacfcd/packages/gittensory-mcp/bin/gittensory-mcp.js#L151-L154) [[15]](https://github.com/JSONbored/gittensory/blob/1e7506f4c5226d0b9b5da8fcd219309d84bacfcd/packages/gittensory-mcp/bin/gittensory-mcp.js#L459). The server registers all 20 MCP tools and remains alive until the client process exits.
+This starts an `McpServer` named `gittensory-local` that communicates via `StdioServerTransport` — reading JSON-RPC messages from stdin and writing responses to stdout [[14]](https://github.com/JSONbored/gittensory/blob/1e7506f4c5226d0b9b5da8fcd219309d84bacfcd/packages/gittensory-mcp/bin/gittensory-mcp.js#L151-L154) [[15]](https://github.com/JSONbored/gittensory/blob/1e7506f4c5226d0b9b5da8fcd219309d84bacfcd/packages/gittensory-mcp/bin/gittensory-mcp.js#L459). The server registers all 21 MCP tools and remains alive until the client process exits.
 
 The `--stdio` flag is what MCP clients use to spawn and communicate with the server. From the client's perspective, Gittensory appears as a local process that accepts tool calls and returns structured JSON results.
 
@@ -154,7 +154,7 @@
 
 ## What Context It Exposes to Agents
 
-The 20 MCP tools are organized into five capability categories. Together they give agents a complete picture of contribution health — from raw repo signals to scored, ranked next actions — without requiring source code upload.
+The 21 MCP tools are organized into five capability categories. Together they give agents a complete picture of contribution health — from raw repo signals to scored, ranked next actions — without requiring source code upload.
 
 ### Repo Intelligence
 
@@ -163,6 +163,7 @@
 | `gittensory_get_repo_context` | Returns the canonical repo intelligence bundle from the Gittensory API — queue health, lane configuration, scoring model details, and registered maintainer policies [[18]](https://github.com/JSONbored/gittensory/blob/1e7506f4c5226d0b9b5da8fcd219309d84bacfcd/packages/gittensory-mcp/bin/gittensory-mcp.js#L156-L166) |
 | `gittensory_get_registry_changes` | Returns the latest cached Gittensor registry change report — validator set changes, stake shifts, network topology updates [[19]](https://github.com/JSONbored/gittensory/blob/1e7506f4c5226d0b9b5da8fcd219309d84bacfcd/packages/gittensory-mcp/bin/gittensory-mcp.js#L203-L210) |
 | `gittensory_local_status` | Returns the local MCP status, inferred git repo metadata, and privacy defaults — useful for confirming the server is properly configured and what repo it has detected [[20]](https://github.com/JSONbored/gittensory/blob/1e7506f4c5226d0b9b5da8fcd219309d84bacfcd/packages/gittensory-mcp/bin/gittensory-mcp.js#L257-L288) |
+| `gittensory_get_repo_outcome_patterns` | Retrieves learned PR outcome patterns for a repository, showing what tends to get merged vs closed based on historical data across dimensions like paths, labels, size, linked issues, tests, body length, and review discussion. This tool provides insights into what maintainers in a specific repository actually merge or close, distinct from the contributor-centric outcome history |
 
 ### PR Preflight & Scoring
 
@@ -546,7 +547,7 @@
 
 ### MCP Tools Reference
 
-All 20 tools use Zod-validated input schemas [[53]](https://github.com/JSONbored/gittensory/blob/1e7506f4c5226d0b9b5da8fcd219309d84bacfcd/packages/gittensory-mcp/bin/gittensory-mcp.js#L28-L144) and return a structured response containing a human-readable `summary` string and a JSON `data` object.
+All 21 tools use Zod-validated input schemas [[53]](https://github.com/JSONbored/gittensory/blob/1e7506f4c5226d0b9b5da8fcd219309d84bacfcd/packages/gittensory-mcp/bin/gittensory-mcp.js#L28-L144) and return a structured response containing a human-readable `summary` string and a JSON `data` object.
 
 #### Category 1: Repo Intelligence
 

^{How did I do? Any feedback?}