Skip to content

README.md

Latest commit

 

History

History
140 lines (104 loc) · 4.7 KB

README.md

File metadata and controls

140 lines (104 loc) · 4.7 KB
Raw

speedreview

AI code review in under 5 seconds. Powered by jCodeMunch + Groq.

Your code reviewer is slower than your linter.

Uses jCodeMunch's token-efficient code retrieval (AST parsing, symbol analysis, blast radius) combined with Groq's ultra-fast inference (280-1000 tok/s) to post a structured review as a PR comment — before your CI checks even start.

Example Output

## speedreview (3.2s)

### Summary
This PR adds retry logic to the HTTP client with exponential backoff.

### Issues Found
- **[High]** `retry_count` has no upper bound — infinite loop risk (src/client.py:42)
- **[Medium]** Missing timeout on the retry delay — could block event loop (src/client.py:58)

### Impact Analysis
3 downstream callers affected: `fetch_user()`, `fetch_repo()`, `sync_data()`.
No breaking signature changes detected.

---
*Powered by [jCodeMunch](https://github.com/jgravelle/jcodemunch-mcp) + [Groq](https://groq.com) | Review completed in 3.2s*

Setup

1. Get a Groq API key

Sign up at console.groq.com and create an API key.

2. Add the secret to your repo

Go to Settings > Secrets and variables > Actions and add GROQ_API_KEY.

3. Create the workflow

# .github/workflows/speedreview.yml
name: speedreview
on: [pull_request]

permissions:
  pull-requests: write
  contents: read

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0
      - uses: jgravelle/jcodemunch-mcp/speedreview@v1.108.22
        with:
          groq_api_key: ${{ secrets.GROQ_API_KEY }}

Pinning for production

The example above pins the action to a specific package release tag (@v1.108.22), which is the recommended baseline. For workflows under stricter supply-chain review, pin to the commit SHA the tag points to instead:

      - uses: jgravelle/jcodemunch-mcp/speedreview@<full-40-char-sha>
        with:
          groq_api_key: ${{ secrets.GROQ_API_KEY }}

Resolve the SHA with git ls-remote https://github.com/jgravelle/jcodemunch-mcp refs/tags/v1.108.22.

SHA pinning makes the consumed action immutable: if a tag is ever moved or a breaking change ships on main, the workflow keeps running the same code until you intentionally bump the SHA. The @main form (older docs) is no longer recommended for production workflows.

The action also pins its installed Python packages by default (jcodemunch-mcp==1.108.20, openai>=1.50,<2). Override those defaults via the jcodemunch_version and openai_version inputs if your environment requires a different pin.

Configuration

Input Default Description
groq_api_key (required) Groq API key
model llama-3.3-70b-versatile Groq model for generating the review
severity_threshold low Minimum severity to include: low, medium, high
max_comment_length 4000 Maximum PR comment length in characters
token_budget 8000 Token budget for jCodeMunch context retrieval
base_ref (auto-detect) Base ref to diff against
jcodemunch_version ==1.108.22 PyPI version specifier for jcodemunch-mcp
openai_version >=1.50,<2 PyPI version specifier for the openai SDK

Model options

Model Speed Best for
llama-3.3-70b-versatile 280 tps Best review quality (default)
openai/gpt-oss-120b 500 tps Complex reasoning
openai/gpt-oss-20b 1000 tps Maximum speed
llama-3.1-8b-instant 560 tps Rate-limit friendly

How It Works

PR opened/updated
     |
     v
1. git diff -> changed files + hunks
2. jCodeMunch indexes the repo (cached between runs)
3. get_changed_symbols -> what changed at the symbol level
4. get_blast_radius -> downstream impact analysis
5. get_ranked_context -> relevant surrounding code (token-budgeted)
6. Groq inference -> structured review (sub-2s)
7. Post as PR comment
     |
     v
Review appears in < 5 seconds (cache hit)
  • jCodeMunch runs locally in the Action — no network hop, works on private repos
  • Index is cached via GitHub Actions cache — subsequent PRs skip indexing
  • Groq is the only external call — one API request, sub-2s inference
  • Updates existing comment on force-push — no comment spam

What It Reviews

Focuses on: bugs, security vulnerabilities, performance problems, missing error handling, breaking API changes.

Ignores: style, formatting, naming, docs, test coverage.

Cost

  • Groq API: ~$0.001-0.01 per review depending on diff size and model
  • GitHub Actions: Standard runner minutes (typically 10-30s per review)