Repo-pilot

## RepoPilot A **repo-intelligence engine** built for IBM Bob 2.0 that automates two expensive developer workflows from one shared pipeline: --- ### 🔍 What it does #### 1. PR / Release Mode When a developer opens a pull request, RepoPilot runs **4 AI agents in parallel** and produces a scored **Release Readiness Report**: | Agent | What it checks | |---|---| | `diff-analyst` | Breaking API changes — renamed fields, new env vars | | `test-gap-agent` | New endpoints with zero test coverage → writes test stubs | | `doc-sync-agent` | README/CHANGELOG out of sync with code changes | | `compliance-agent` | Checklist items: secrets, env vars, setup docs | #### 2. Onboarding Mode For a new developer joining a codebase, RepoPilot generates an **Onboarding Brief** containing: - Architecture map (Mermaid diagram) - "Where things live" table - Domain glossary - Day-0 setup checklist - Safest files to start with (ranked by git churn) - Self-check quiz --- ### 🧠 The core idea Both modes **share intelligence** — the onboarding agent reuses the compliance-agent's checklist and the diff-analyst's file-churn risk scan instead of recomputing them. This is the "shared repo-intelligence engine" concept. ``` git diff ──► diff-analyst ──┐ test-gap-agent ├──► release-report.md (PR mode) doc-sync-agent ├──► onboarding-brief.md (Onboarding mode) compliance-agent─┘ (shared state between both) ``` --- ### 🎯 Problem it solves | Problem | Current cost | RepoPilot | |---|---|---| | PR review for breaking changes | 25–40 min manually | ~0.2 seconds | | New contributor first commit | 3–5 days to absorb codebase | Instant onboarding brief | --- ### 🏗️ Structure ``` repo-pilot/ demo.py ← single entry point, run this orchestrator/ run.py ← CLI, runs agents in parallel (ThreadPoolExecutor) agents/ diff_analyst.py ← detects field renames, new routes, new env vars test_gap.py ← finds untested endpoints, writes test stubs doc_sync.py ← checks README/CHANGELOG drift compliance.py ← validates checklist, secrets, env docs onboarding.py ← builds the onboarding brief report_builder.py ← scoring + Markdown report generation bob_backend.py ← swap point to use live IBM Bob Shell instead sample-project/ ← demo Flask "Tasks API" with planted bugs bob-agents/*.md ← role-definition prompts for IBM Bob custom modes ``` --- ### 🔌 IBM Bob 2.0 integration The rule-based agents in `orchestrator/agents/` are a **working reference implementation** — they run right now with zero dependencies. The [`bob_backend.py`](repo-pilot/orchestrator/bob_backend.py) is the **swap point**: replace any agent call with `run_bob_agent(...)` and the rest of the pipeline (scoring, reports, parallelism) stays unchanged because both backends return the same JSON shape.