### Cut Your AI API Costs by 40-70%

TokenFlow is a self-hosted reverse proxy that sits between your applications and AI providers, automatically optimizing token costs through **semantic caching, smart model routing, and prompt compression**.

### The Problem
AI API costs are spiraling out of control. Companies spend thousands monthly on OpenAI, Anthropic, and other providers — often paying full price for duplicate or similar requests.

### The Solution
TokenFlow acts as an intelligent middleware. Change one URL in your code, and every request is automatically:
- **Cached semantically** — similar prompts return cached responses (FREE)
- **Compressed** — redundant tokens removed before sending
- **Routed optimally** — simple tasks sent to cheaper models automatically
- **Tracked** — full cost analytics with budget alerts

**8 AI Provider Support**
OpenAI, Anthropic Claude, Google Gemini, Mistral, DeepSeek, Groq, Cohere, and Ollama (local models).

**OpenAI-Compatible API**
Drop-in replacement. Just change your `base_url` — existing code works unchanged with Python, JavaScript, cURL, PHP, LangChain, Vercel AI SDK.

**Semantic Caching**
Not just exact matches — finds similar prompts using cosine similarity. Configurable threshold and TTL.

**Smart Model Routing**
Define rules: "If prompt < 500 tokens, use gpt-4o-mini" or "If prompt contains 'code', use deepseek-coder."

**Prompt Compression**
Automatically removes redundant whitespace, deduplicates content, and summarizes long conversation histories.

**Budget Protection**
Set daily/weekly/monthly spending limits with automatic model downgrade when exceeded.

**Real-Time Analytics Dashboard**
Live request feed, cost charts, savings breakdown, provider distribution, latency tracking.

**API Playground**
Test requests through the proxy and see caching, routing, and compression in action.

**Admin Panel**
User management, API key management, model configuration, system settings.

- Node.js 18 or higher
- npm or yarn
- No database server needed (SQLite included)
- Optional: PostgreSQL for production
- Optional: AI provider API keys (OpenAI, Anthropic, etc.)

npm install
npx prisma db push
npx tsx prisma/seed.ts
npm run dev
```
Open http://localhost:3000 — gateway is running! Change your AI SDK base_url to http://localhost:3000/api/v1