█████╗  ██████╗ ███████╗███╗   ██╗████████╗██╗     ███████╗███╗   ██╗███████╗
██╔══██╗██╔════╝ ██╔════╝████╗  ██║╚══██╔══╝██║     ██╔════╝████╗  ██║██╔════╝
███████║██║  ███╗█████╗  ██╔██╗ ██║   ██║   ██║     █████╗  ██╔██╗ ██║███████╗
██╔══██║██║   ██║██╔══╝  ██║╚██╗██║   ██║   ██║     ██╔══╝  ██║╚██╗██║╚════██║
██║  ██║╚██████╔╝███████╗██║ ╚████║   ██║   ███████╗███████╗██║ ╚████║███████║
╚═╝  ╚═╝ ╚═════╝ ╚══════╝╚═╝  ╚═══╝   ╚═╝   ╚══════╝╚══════╝╚═╝  ╚═══╝╚══════╝

Know exactly what your AI agent is doing wrong.

---

⚡ What is AgentLens?

Most developers ship AI agents and have no idea what's happening inside them.

AgentLens fixes that. It logs every conversation, automatically extracts user intent and sentiment using AI, detects failures before users churn, and suggests exactly how to improve your system prompt — all in real time.

User asks → Agent responds → AgentLens logs it
                                    ↓
                          AI extracts intent
                          Detects failures
                          Tracks sentiment
                                    ↓
                          Dashboard shows patterns
                          Prompt Lab fixes them

---

✨ Features

Feature	Description
📡 Conversation Logging	One line of Python to capture every chat
🧠 Intent Extraction	AI auto-labels what every user was trying to do
💥 Failure Detection	Catches where your agent failed or frustrated users
😤 Sentiment Analysis	Positive / Neutral / Negative breakdown
⚡ Prompt Lab	AI analyzes failures and rewrites your system prompt
🔄 Auto-Analysis	Runs every hour — no manual clicks needed
👥 Multi-tenant	Each user gets their own project and API key
🔑 Auth System	Signup / login with Supabase

---

🚀 Quick Start

1. Clone

git clone https://github.com/Shraman123/AgentLens.git
cd AgentLens

2. Backend

cd backend
pip install -r requirements.txt

# Set your keys
export GROQ_API_KEY=gsk_...
export SUPABASE_URL=https://your-project.supabase.co
export SUPABASE_SERVICE_KEY=eyJ...

uvicorn app:app --reload --port 8000

3. Frontend

cd frontend
npm install

REACT_APP_API_URL=http://localhost:8000 npm start

4. Open

http://localhost:3000

---

📡 Integrate in 1 Line

from agentlens import AgentLens

lens = AgentLens(api_key="ak_your_key", base_url="https://your-backend.hf.space")

# Decorator — auto-logs everything
@lens.watch
def my_agent(user_message: str) -> str:
    return call_your_llm(user_message)

# That's it. Every conversation is now logged and analyzed.
my_agent("How do I reset my password?")

Async support

@lens.watch_async
async def my_agent(user_message: str) -> str:
    return await call_your_llm(user_message)

Direct HTTP

curl -X POST https://your-backend.hf.space/log \
  -H "x-api-key: ak_your_key" \
  -H "Content-Type: application/json" \
  -d '{"user_message": "Hello", "agent_response": "Hi there!"}'

---

📋 API Reference

Method	Endpoint	Auth	Description
GET	/health	—	Health check
POST	/auth/signup	—	Create account
POST	/auth/login	—	Sign in
GET	/projects	Bearer	List your projects
POST	/projects	Bearer	Create project
POST	/log	API Key	Log a conversation
POST	/analyze	API Key	Analyze conversations
GET	/dashboard	API Key	Get dashboard stats
GET	/conversations	API Key	List conversations
POST	/suggest-prompt	API Key	Get prompt suggestion

---

🏗 Architecture

┌─────────────────┐     ┌──────────────────┐     ┌─────────────────┐
│   Your AI Agent  │────▶│  AgentLens API   │────▶│    Supabase     │
│                 │     │  (FastAPI + HF)   │     │   (Postgres)    │
└─────────────────┘     └──────────────────┘     └─────────────────┘
                                  │                        │
                                  ▼                        ▼
                         ┌──────────────────┐    ┌─────────────────┐
                         │   Groq / LLM     │    │  React Dashboard │
                         │ (Intent extract) │    │    (Vercel)      │
                         └──────────────────┘    └─────────────────┘

---

🚢 Deploy Free

Service	What	Cost
Hugging Face Spaces	Backend API	Free
Vercel	Frontend dashboard	Free
Supabase	Database + Auth	Free
Groq	AI analysis	Free

Total: $0/month for up to ~50k conversations analyzed.

Deploy Backend (Hugging Face)

1. Fork this repo
2. Create a new Space → Docker
3. Add secrets: GROQ_API_KEY, SUPABASE_URL, SUPABASE_SERVICE_KEY
4. Push backend/ files to the Space

Deploy Frontend (Vercel)

1. Import repo to Vercel
2. Set root directory: frontend
3. Add env vars: REACT_APP_API_URL, REACT_APP_API_KEY
4. Deploy

---

🗺 Roadmap

• Conversation logging
• AI intent extraction
• Failure detection
• Sentiment analysis
• Prompt Lab
• Auto-analysis scheduler
• Multi-tenant auth
• Email alerts on failure spikes
• Slack / Discord notifications
• OpenTelemetry ingestion
• Prompt version history
• A/B testing for prompts
• CSV export

---

💰 Cost to Run

Conversations analyzed	Monthly cost
10,000	~$0 (Groq free tier)
100,000	~$1-2
1,000,000	~$10-20

---

🛠 Stack

• Backend — FastAPI, Supabase, APScheduler
• AI — Groq (llama-3.3-70b), free tier
• Frontend — React, Recharts, React Query, TanStack
• Auth — Supabase Auth
• Hosting — Hugging Face Spaces + Vercel

---

📄 License

MIT — free to use, modify, and build on.

---

Built in public · Star ⭐ if this helped you

Live Demo · API Docs · Landing Page