OpenAI Proxy with Usage Tracking

A transparent proxy that sits between your app and OpenAI's API — adding usage tracking, cost allocation, caching, rate limiting, and a full admin dashboard.

Proxy Features:

• Drop-in replacement — same API interface as OpenAI
• Works with chat completions, embeddings, images, and audio
• Streaming support (SSE passthrough)
• Request/response logging with full payloads
• Latency tracking per model and endpoint

Usage Tracking:

• Per-user token consumption tracking
• Cost calculation using OpenAI pricing tiers
• Daily/weekly/monthly usage reports
• Budget alerts when users approach limits
• Department/team allocation for enterprise billing

Rate Limiting:

• Per-user requests per minute/hour/day limits
• Per-model rate limiting
• Burst allowance with token bucket algorithm
• Queue overflow handling with retry-after headers
• Priority lanes for premium users

Response Caching:

• Semantic caching — match similar prompts to cached responses
• Exact match caching with configurable TTL
• Cache hit rate dashboard
• Estimated savings from cache hits
• Per-model cache configuration

Admin Dashboard:

• Real-time usage graphs
• Top users by token consumption
• Cost breakdown by model and user
• API key management with scoped permissions
• Audit log with full request history
• Export to CSV for accounting

Tech Stack:

• Node.js + TypeScript (Fastify)
• Redis for caching and rate limiting
• PostgreSQL for usage data and audit logs
• React admin dashboard
• Docker Compose with Redis + Postgres
• Prometheus metrics endpoint