Skip to content

CHANGELOG.md

Latest commit

 

History

History
56 lines (44 loc) · 2.38 KB

CHANGELOG.md

File metadata and controls

56 lines (44 loc) · 2.38 KB

Changelog

All notable changes, organized by development day.

Day 0 – Project skeleton

  • Initialized repository and Python package layout
  • Added MCP “hello” example to verify SDK wiring

Day 1 – GitHub ingest + diff cache

  • Implemented ingest_pr tool using PyGitHub and unidiff
  • Added SQLite caching (24 h TTL) for PR diffs and stats

Day 2 – LLM analysis & risk scoring

  • Created local LLM smells stage with CodeLlama-13B (Ollama)
  • Integrated cloud LLM (Groq Llama-3 8K) for summary and risk score
  • Exposed analyze_pr tool

Day 3 – Inline locator & embeddings

  • Built locate util that maps static rule “smells” to file hunks
  • Added ChromaDB-backed hunk embeddings to improve line-number accuracy
  • Wrapped all GitHub calls to avoid duplicate requests

Day 4 – CLI & CI risk gate

  • Developed reviewgenie CLI with subcommands: ping, ingest, analyze, inline, check
  • Added check tool to fail CI when risk_score > threshold
  • Updated GitHub Actions CI workflow to run pytest + risk gate

Day 5 – Stub test generator

  • Implemented generate_tests tool to draft pytest stubs and open a PR
  • Created branch, added files, and submitted pull request via PyGitHub

Day 6 – Stability fixes & CI passing

  • Fixed Duplicate-ID bug in Chroma upsert logic
  • Corrected review-comment arg order for PyGitHub
  • Finalized CI so all tests pass (ping, ingest, analyze, inline, check)

Day 7 – Evaluation & benchmarks

  • Added bench.py to replay 10 historical PRs
  • Measures latency (ingest/analyze/inline) and accepted-comment counts
  • Produces bench/benchmarks.md with per-PR stats and averages
  • Introduced --dry-run for inline to avoid write errors

Day 8 – Secrets, back-off, tracing

  • Moved secrets to codeview_mcp/secret.py (env-vars or OS keyring)
  • Wrapped all GitHub API calls in gh_call() with exponential back-off
  • Instrumented tools with OpenTelemetry spans via @traced

Day 9 – Full docs & schema

  • Published complete docs/ suite: Architecture, Quickstart, Usage, Config, Contributing, API schema
  • Added scripts/export_schema.py to generate docs/API_SCHEMA.json
  • Polished README.md and finalized changelog entries

[1.3.0] – 2025-05-31

Changed

  • Switched Groq default model from llama3-8b-8192 → llama-3.1-8b-instant
  • Added RG_CLOUD_MODEL env-var to override at runtime