Personal Chat Agent — Portfolio Assistant Backend

A ReAct agent backend for developer portfolio chatbots. Built with LangGraph, RAG and FastAPI.

Live demo: prashantnathv2.netlify.app

Architecture Diagrams

See docs/architecture/

---

What it does

• Answers questions about you using a RAG knowledge base (Pinecone)
• Fetches your live GitHub repositories on demand
• Captures visitor contact details to Supabase
• Streams responses token by token via SSE
• Remembers context within a conversation session

---

Tech Stack

Layer	Technology
Framework	FastAPI
Agent	LangGraph (ReAct pattern)
LLM	OpenAI gpt-4o-mini
Embeddings	OpenAI text-embedding-3-small
Vector Store	Pinecone
Database	Supabase (PostgreSQL)
Memory	LangGraph MemorySaver
Streaming	Server-Sent Events (SSE)

---

Project Structure

app/
├── agent/
│   ├── graph.py          # LangGraph graph assembly
│   ├── nodes.py          # llm_node and tool_node
│   ├── prompt.py         # System prompt
│   └── state.py          # Agent state schema
├── api/
│   ├── api_service/
│   │   └── chat_service.py   # SSE streaming logic
│   └── routes/
│       └── chat_router.py    # /chat endpoint
├── core/
│   └── config.py         # Environment variables settings
├── services/
│   ├── pinecone_service.py
│   └── supabase_service.py
└── tools/
    ├── rag_tool.py
    ├── github_tool.py
    └── user_capture_tool.py
scripts/
└── indexing.py           # Knowledge base ingestion

---

Use This For Your Portfolio

1. Clone the repo

git clone https://github.com/prashant00797/personal-chat-agent
cd personal-chat-agent

2. Set up environment variables

cp .env.example .env

Fill in your values:

OPENAI_API_KEY=
PINECONE_API_KEY=
PINECONE_INDEX_NAME=
PINECONE_HOST=
SUPABASE_URL=
SUPABASE_KEY=
GITHUB_BASE_URL=https://api.github.com/users/YOUR_USERNAME/repos
STREAM_DELAY=
DUMMY_HEALTH_CHECK_BOT_ID=

3. Add your knowledge base

Create a PDF with information about yourself — experience, skills, projects, availability, preferences. Place it at:

knowledge_base/your_name_kb.pdf

4. Run ingestion

uv run python -m scripts.indexing

This embeds your document and pushes vectors to Pinecone.

5. Update the system prompt

Edit app/agent/prompt.py — replace Prashant's details with your own name, contact info and persona.

6. Run the server

uv run uvicorn main:app --reload

API available at http://localhost:8000 Swagger docs at http://localhost:8000/docs

---

API

POST /api/chat

Streams a chat response as SSE.

Request:

{
  "message": "What is your experience with React?",
  "thread_id": "uuid-string"
}

SSE Events:

data: {"type": "tool_call", "tool": "retrieve_relevant_chunks"}
data: {"type": "token", "content": "Prashant"}
data: {"type": "end"}
data: {"type": "error", "message": "..."}

---

Deploy

Tested on Render free tier.

Start command:

uvicorn main:app --host 0.0.0.0 --port $PORT

Add all environment variables in Render dashboard under Environment.

---

Requirements

• Python 3.11+
• uv package manager
• Pinecone account (free tier works)
• Supabase account (free tier works)
• OpenAI API key

---

Frontend

The Next.js chat widget frontend is in a separate repo: my-portfolio-v2