InferSpect is an LLM proxy and observability platform designed to centralize infrastructure for managing, monitoring, and validating LLM interactions.

Note: This project is currently under development.

This repository uses an AI-powered development workflow with three specialized providers:

• Jules (Specifications & Planning):

  • Comment @jules spec on an issue or PR to generate technical specifications and requirements analysis
  • Comment @jules plan on an issue or PR to request system design and architecture planning
  • Uses Google Gemini 1.5 Pro for comprehensive analysis
  • Generates detailed specifications, implementation strategies, technology recommendations, and risk assessments
  • Requires JULES_API_KEY secret - See setup guide
  • Powered by jules-specs and jules-planner packages

• Claude (Documentation Refresh):

  • Auto-Review: Non-draft Pull Requests are automatically reviewed by Claude upon opening, synchronization, or when marked ready for review
  • On-Demand: Comment @claude in your Pull Request to request specific documentation updates
  • Claude keeps Markdown and planning docs current while flagging any issues for Cursor to address
  • Note: Draft PRs are excluded from automatic reviews to avoid premature feedback

• Cursor (Verify): Comment @cursor verify on a PR to:

  • Invoke the Cursor Cloud GPT-5.1 Codex agent for a whole-repo security/quality review
  • Receive a Markdown report posted back to the PR with prioritized findings
  • Run the automated verification pipeline (poetry run pytest plus Bandit) to validate fixes

To enable Jules architecture planning, you need to configure the JULES_API_KEY secret:

1. Get a Gemini API Key:

   • Visit Google AI Studio
   • Sign in with your Google account
   • Click "Create API Key"
   • Copy the generated API key

2. Add Secret to GitHub:

   • Go to your repository Settings
   • Navigate to "Secrets and variables" → "Actions"
   • Click "New repository secret"
   • Name: JULES_API_KEY
   • Value: Paste your Gemini API key
   • Click "Add secret"

3. Test the Integration:

   • Create a test issue or PR
   • Comment @jules spec to generate technical specifications
   • Comment @jules plan to generate a comprehensive architecture plan

Note: The Gemini API has usage limits. Check the pricing page for details.

To enable the Cursor AI agent step inside @cursor verify, configure CURSOR_CLOUD_API_KEY:

1. Create an API Key
   • Visit https://cursor.com and open the Cloud dashboard
   • Generate a new API key with access to the Cursor Cloud Agent endpoints
2. Add Repository Secret
   • Go to Settings → Secrets and variables → Actions
   • Add CURSOR_CLOUD_API_KEY with the newly created key
   • (Optional) add CURSOR_CLOUD_BASE_URL if you are targeting a private Cursor Cloud deployment
3. Trigger the Workflow
   • Comment @cursor verify on any pull request
   • The workflow launches GPT-5.1 Codex via Cursor Cloud, posts the agent's Markdown report, then runs pytest + Bandit

If the secret is missing, the workflow skips the agent invocation but still performs the local test suite. Setting the secret is strongly recommended so you receive the automated review before the tests run.