Skip to content

ROADMAP.md

Latest commit

 

History

History
92 lines (61 loc) · 3.77 KB

ROADMAP.md

File metadata and controls

92 lines (61 loc) · 3.77 KB

Roadmap

This file captures the longer-term product direction for GitHub Portfolio Reviewer Agent.

It is intentionally vision-oriented rather than time-based.

Current Foundation

The project already has:

  • a hosted Streamlit app
  • deterministic repository scoring
  • per-repository and portfolio-level LLM analysis
  • persistent SQLite-backed caching for single-instance hosting
  • Markdown and PDF export
  • GitHub OAuth for public-profile convenience
  • baseline GitHub Actions CI

Product Direction

1. Stronger Repository Presentation

Keep improving the recruiter-facing polish of the repository and the product:

  • refine README screenshots and demo assets as the UI evolves
  • keep exported reports visually stronger and more branded
  • improve documentation around setup, testing, and expected outputs
  • add clearer badges and public signals that make the repo easier to evaluate quickly

2. Production-Grade Backend and Hosting Model

The current app is a strong hosted prototype, but the long-term direction is a more explicit application backend:

  • move from a Streamlit-only runtime toward a dedicated backend layer where appropriate
  • introduce a FastAPI service for analysis orchestration, background jobs, and cleaner API boundaries
  • support a more standard production deployment model beyond a single hosted Streamlit instance
  • separate frontend presentation concerns from backend execution and persistence concerns

3. Shared Persistence Instead of Local SQLite

SQLite works well for local development and single-instance hosting, but the scaling path is a hosted/shared data layer:

  • move persistent analysis storage to a hosted database
  • support shared cache/report reuse across instances instead of host-local files only
  • preserve repository freshness validation while making cached reports available across deployments
  • create a cleaner path for usage analytics, audit history, and future team features

Possible future storage direction:

  • hosted Postgres or similar managed relational database for persistent report state
  • optional Redis-style cache layer for fast transient reuse and rate-limit protection

4. Containerized Deployment and More Portable Infrastructure

For a more production-style deployment model, the app should become easier to run outside the current Streamlit-hosted setup:

  • add Docker support for reproducible local and cloud deployment
  • make the runtime environment explicit instead of relying on host defaults
  • support cleaner migration to VPS/cloud/container platforms
  • make browser/PDF dependencies easier to manage consistently across environments

5. Repository RAG / Q&A Mode

One of the clearest future product extensions is a public-repo exploration mode:

  • allow users to ask questions about a repository after analysis
  • build retrieval over repository structure, README content, and selected source files
  • support repository-level Q&A grounded in indexed public code and docs
  • extend the app from static report generation into interactive repository understanding

This would complement the current audit flow rather than replace it.

6. Broader Report Intelligence

Beyond the current recruiter-facing audit, the report system could grow into a richer review layer:

  • compare multiple repositories more explicitly
  • cluster portfolio strengths and recurring weaknesses across projects
  • support alternate report modes for recruiters, maintainers, or self-review
  • add stronger export customization for portfolios, case studies, and public showcases

Guiding Principle

The long-term goal is not just to analyze a GitHub profile once.

It is to turn this project into a more production-grade portfolio intelligence system:

  • hosted
  • shareable
  • reproducible
  • scalable
  • and eventually interactive through deeper repository understanding workflows