ProjectBridge

Analyze your GitHub repos against job descriptions and get concrete project ideas to close skill gaps.

What should I build next to move toward the role I want?

Runs entirely on your machine. No accounts. No uploads.

---

Quickstart

pip install projectbridge

Try the built-in example (no GitHub token or AI provider required):

projectbridge analyze --example

Sample output (trimmed)

{
  "schema_version": "1.2",
  "experience_level": "mid",
  "strengths": [
    { "name": "Python", "category": "language" },
    { "name": "React", "category": "framework" },
    { "name": "Docker", "category": "infrastructure" }
  ],
  "gaps": [
    { "name": "TypeScript", "category": "language" },
    { "name": "Kubernetes", "category": "infrastructure" },
    { "name": "PostgreSQL", "category": "tool" }
  ],
  "recommendations": [
    {
      "title": "Build a REST API with Django",
      "description": "Create a RESTful API using Django REST Framework...",
      "skills_addressed": ["Django", "REST API"],
      "estimated_scope": "medium",
      "skill_context": "Django powers production web apps from startups to enterprises..."
    }
  ],
  "portfolio_insights": [
    {
      "category": "infrastructure",
      "message": "Your portfolio doesn't demonstrate deployment or infrastructure skills..."
    }
  ],
  "interview_topics": [
    {
      "skill": "Kubernetes",
      "topics": ["Pod lifecycle and deployment strategies", "Services, ingress, and networking"]
    }
  ]
}

Run with your own data:

projectbridge analyze --job job.txt --github-user your-username

Export as Markdown:

projectbridge export --job job.txt --github-user your-username

Install from source

git clone https://github.com/akuligowski9/project-bridge
cd project-bridge
pip install -e ./engine

---

What It Does

ProjectBridge reads a job description, scans your GitHub profile, and produces a structured analysis with:

• Strengths — skills you already demonstrate
• Gaps — skills the job requires that your repos don't show
• Recommendations — concrete project ideas to close those gaps

Key capabilities:

• 4 AI providers — OpenAI, Anthropic, Ollama, or no-AI heuristic fallback
• 97+ framework/tool detections — from React to Terraform
• 31 curated project templates — scoped as small, medium, or large
• Experience calibration — infers junior/mid/senior level, adjusts recommendation difficulty
• Portfolio insights — surfaces underrepresented areas (missing infra, domain gaps, imbalance)
• Interview prep — maps 40 gap skills to common interview topics
• Personalized recommendations — references your existing skills as a foundation
• Non-technical JD rejection — catches non-tech job descriptions early with a clear error
• Desktop app — Tauri + Svelte 5 UI for visual exploration
• Structured JSON output — schema v1.2, ready for CI/CD or downstream tools
• File-based caching — skip redundant GitHub API calls
• Privacy-first — tokens stay local, no telemetry, no uploads

---

Architecture

Tauri Desktop App (Svelte 5 + TypeScript)
        ↓
  Python Engine (pip install projectbridge)
        ↓
  Analysis Pipeline
    ├── GitHub repo scanner
    ├── Job description parser
    ├── Skill matcher
    └── Recommendation engine
        ↓
  AI Provider Interface (pluggable)
    ├── OpenAI  ├── Anthropic  ├── Ollama  ├── None (heuristic)
        ↓
  Structured JSON Output (schema v1.2)

---

AI Setup

AI is optional — ProjectBridge works great without it. To enable AI-enhanced recommendations for free, use Ollama (local, no API key, no cost):

# Install Ollama (macOS)
brew install ollama

# Start the server and pull a model
ollama serve
ollama pull llama3.2

Then create projectbridge.config.yaml:

ai:
  provider: ollama

That's it. See the full AI Setup Guide for other providers (OpenAI, Anthropic) and troubleshooting.

Prompts are transparent and editable — see engine/projectbridge/ai/prompts/.

---

Desktop App

The Tauri desktop app provides a visual interface for running analyses and exploring results.

cd app
npm install
npm run dev

Requires the Python engine to be installed (pip install -e ./engine).

---

Privacy

ProjectBridge is intentionally not SaaS.

• GitHub tokens never leave your machine
• Private repositories remain private
• No telemetry or analytics
• All outputs stored locally

---

Contributing

ProjectBridge is designed to be forked and extended. Common areas:

• AI provider integrations
• Language/framework detection
• Recommendation improvements
• Export formats
• UI improvements

See CONTRIBUTING.md.

---

License

MIT License — see LICENSE.