SelfGPT

SelfGPT is a private, production‑ready LangChain + LangSmith chat application. It ships a Next.js UI with a FastAPI backend, supports multiple LLM providers, PDF‑based RAG, and optional live web search via Google Custom Search. LangSmith tracing is built in for observability when enabled.

Tech Stack

Frontend

• Next.js 16 (App Router)
• React 19
• TypeScript
• Tailwind CSS
• Radix UI components

Backend

• FastAPI (Python)
• Requests (provider HTTP calls)
• FAISS + SentenceTransformers (RAG)
• LangChain (community loaders/splitters for PDF ingestion)
• LangSmith (tracing + observability)
• Google Custom Search API (web search)

LLM Providers (OpenAI‑compatible or native)

• Groq
• OpenAI
• DeepSeek
• xAI Grok
• Anthropic
• Gemini
• Mistral

LangChain & LangSmith

• LangChain powers PDF loading, chunking, embeddings, and retrieval for RAG.
• LangSmith provides end‑to‑end tracing for RAG and LLM calls when enabled via environment variables.

Key Features

• Multi‑provider LLM support with model selection
• PDF upload + retrieval‑augmented generation (RAG)
• Optional web search integration (Google CSE)
• Persistent chat history (SQLite)
• Streaming‑style typewriter responses
• Responsive UI (desktop + mobile)
• Security controls: API key auth, rate limiting, upload limits
• LangSmith observability (optional, secure, key‑gated)

Project Structure

app/                     # Next.js app (UI)
components/              # UI components
context/                 # React contexts (chat, settings, auth)
server.py                # FastAPI backend
rag_engine.py            # RAG pipeline (FAISS + embeddings)
storage.py               # SQLite chat persistence
requirements.txt         # Python deps
package.json             # JS deps

Setup

1) Install frontend deps

npm install

2) Install backend deps

python3 -m venv .venv
source .venv/bin/activate
python3 -m pip install -r requirements.txt

Note: LangSmith currently does not publish wheels for Python 3.14. If you want tracing enabled, use Python 3.12 or 3.11 for the backend venv.

Environment Variables

Create .env in the project root.

Required (at least one LLM provider key)

GROQ_API_KEY=...
# or
OPENAI_API_KEY=...
# or
DEEPSEEK_API_KEY=...
# or
XAI_API_KEY=...
# or
ANTHROPIC_API_KEY=...
# or
GEMINI_API_KEY=...
# or
MISTRAL_API_KEY=...

Web Search (Google CSE)

GOOGLE_API_KEY=...
GOOGLE_CSE_ID=...   # cx= value from your CSE

LangSmith (Tracing / Observability)

LANGSMITH_TRACING=true
LANGSMITH_API_KEY=...
LANGSMITH_PROJECT=SelfGPT
LANGSMITH_ENDPOINT=https://api.smith.langchain.com

Backend Defaults / Limits

DEFAULT_PROVIDER=groq
DEFAULT_MODEL=llama-3.3-70b-versatile
DEFAULT_SYSTEM_PROMPT="You are a helpful assistant."

# Security
RATE_LIMIT_PER_MIN=60
MAX_UPLOAD_MB=10
MAX_MESSAGE_LEN=6000
MAX_HISTORY_MESSAGES=24

# CORS (add your frontend origin if using a different host)
CORS_ORIGINS=http://localhost:3000

# SQLite DB path (set to a persistent disk on the backend host)
DB_PATH=/var/data/chatbot.db

# Postgres (use this instead of DB_PATH when you don't have a disk)
DATABASE_URL=postgresql://user:pass@host:5432/dbname

# RAG store path (set to a persistent disk on the backend host)
RAG_STORE_DIR=/var/data/rag_store

Frontend (optional)

NEXT_PUBLIC_API_BASE_URL=http://localhost:8000
NEXT_PUBLIC_LLM_PROVIDER=groq
NEXT_PUBLIC_LLM_MODEL=llama-3.3-70b-versatile
NEXT_PUBLIC_SYSTEM_PROMPT="You are a helpful assistant."
NEXT_PUBLIC_USE_WEB=true

If NEXT_PUBLIC_API_BASE_URL is not set, the Next.js dev server will proxy /api/* to the FastAPI backend at http://localhost:8000.

Run

Development (two servers)

npm run dev:full

This starts:

• FastAPI on http://localhost:8000
• Next.js on http://localhost:3000
• live : https://self-gpt-iota.vercel.app

Security Notes (Production‑Grade Defaults)

• Rate limiting: per‑IP, per minute.
• Upload limits: defaults to 10MB per file.
• Thread ID sanitization: prevents path traversal.
• Safe FAISS loading: no dangerous deserialization.
• LangSmith is opt‑in: tracing only occurs when LANGSMITH_TRACING=true and a valid key is set.

API Endpoints (Backend)

• GET /api/config – app + provider config
• GET /api/models?provider=... – list models
• POST /api/chat – send message
• POST /api/upload – upload PDFs
• GET /api/threads – list threads
• GET /api/threads/{thread_id} – messages
• GET /api/docs/{thread_id} – document metadata

Troubleshooting

• Hydration warning: If you see vsc-initialized in a hydration error, a browser extension is injecting DOM changes. Disable the extension or use Incognito.