AI Project Tracker

An intelligent AI-powered Project Tracker assistant that processes meeting transcripts, extracts action items, generates status reports, and provides AI-driven project insights.

Features

• 📝 Meeting Transcript Processing - Automatically process meeting transcripts using LLMs
• ✅ Action Item Extraction - Extract structured action items with owner, task, due date, and blockers
• 📊 Status Report Generation - Generate executive summaries, sprint status, risks, and blockers
• 🤖 AI Chat Interface - Ask questions about project memory and get intelligent answers
• 🧠 Vector-based Memory - RAG (Retrieval-Augmented Generation) for context-aware responses
• 💾 PostgreSQL Integration - Persistent storage for meetings and action items with pgvector support
• ⚡ Async API - Fast, non-blocking FastAPI endpoints

Architecture

API Layer (FastAPI)
  ├── /chat → Chat Agent (RAG)
  └── /transcript → Workflow (Action Extraction)
        ↓
LLM Layer (OpenAI via OpenRouter)
  ├── Chat Agent
  ├── Action Agent
  └── Status Agent
        ↓
Data Layer
  ├── Vector Store (PGVector with Ollama embeddings)
  └── PostgreSQL (Meetings, Action Items)

Tech Stack

• Framework: FastAPI
• LLM: OpenAI GPT-4o-mini (via OpenRouter)
• Vector DB: PostgreSQL with pgvector
• Embeddings: Ollama (nomic-embed-text)
• Orchestration: LangGraph
• ORM: SQLAlchemy

Installation

Prerequisites

• Python 3.10+
• PostgreSQL with pgvector extension
• Ollama (for embeddings)
• OpenAI API key

Setup

1. Clone the repository

   git clone <repo-url>
   cd AI-Project-Tracker/api

2. Create virtual environment

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies

   pip install -r requirements.txt

4. Configure environment variables

   cp .env.example .env

   Edit .env with:

   OPENAI_API_KEY=your_api_key
   OPENAI_MODEL=openai/gpt-4o-mini
   DATABASE_URL=postgresql://user:password@localhost:5432/ai_pm
   

5. Setup Database

   # Using docker-compose
   docker-compose up -d
   
   # Then run migrations
   alembic upgrade head

6. Start Ollama (for embeddings)

   ollama run nomic-embed-text

Usage

Start the Server

uvicorn app.main:app --reload

The API will be available at http://localhost:8000

API Endpoints

1. Process Transcript

POST /transcript

{
  "meeting_title": "Sprint Planning",
  "meeting_time": "2024-05-10T10:00:00",
  "participants": ["Alice", "Bob", "Charlie"],
  "transcript": "Today we discussed... action items include..."
}

Response:

{
  "actions": {
    "items": [
      {
        "owner": "Alice",
        "task": "Complete backend API",
        "due_date": "2024-05-17",
        "blocker": "Waiting on design approval"
      }
    ]
  }
}

2. Chat with Project Memory

POST /chat

{
  "question": "What are the current blockers?"
}

Response:

{
  "answer": "Based on project memory, the current blockers are...",
  "retrieved_memories": ["...", "..."]
}

Project Structure

AI-Project-Tracker/api
├── app/
│   ├── main.py                 # FastAPI app entry point
│   ├── llm.py                  # LLM configuration
│   │
│   ├── api/
│   │   ├── transcript_api.py   # Transcript processing endpoint
│   │   └── chat_api.py         # Chat endpoint
│   │
│   ├── agents/
│   │   ├── chat_agent.py       # RAG-based Q&A agent
│   │   ├── action_agent.py     # Action extraction
│   │   ├── status_agent.py     # Status report generation
│   │   ├── memory_agent.py     # Memory ingestion
│   │   └── transcript_agent.py # Transcript processing
│   │
│   ├── graph/
│   │   └── workflow.py         # LangGraph workflow orchestration
│   │
│   ├── memory/
│   │   ├── vector_store.py     # PGVector configuration
│   │   ├── postgres.py         # PostgreSQL setup
│   │   ├── schemas.py          # ORM models
│   │   └── memory_repository.py# Data access layer
│   │
│   ├── prompts/
│   │   ├── action_prompt.py    # Action extraction prompts
│   │   ├── status_prompt.py    # Status generation prompts
│   │   └── summary_prompt.py   # Summary prompts
│   │
│   └── models/
│       └── dto.py              # Request/response schemas
│
├── requirements.txt            # Python dependencies
├── docker-compose.yml          # PostgreSQL + pgvector setup
└── README.md

Key Concepts

RAG (Retrieval-Augmented Generation)

The chat agent uses vector embeddings to retrieve relevant project context before answering questions, ensuring responses are grounded in actual project data.

Agents

• Chat Agent: Answers user questions using RAG
• Action Agent: Extracts structured action items from transcripts
• Status Agent: Generates comprehensive status reports
• Memory Agent: Ingests and vectorizes meeting data

Workflow

LangGraph orchestrates a state machine that processes transcripts through various agents and returns results.

Development

Running Tests

pytest tests/

Code Style

black app/
pylint app/

Database Migrations

# Create migration
alembic revision --autogenerate -m "description"

# Apply migrations
alembic upgrade head

Contributing

1. Create a feature branch
2. Make your changes
3. Submit a pull request

License

MIT

Support

For issues and questions, please open an issue on the repository.