AI API
llmproxy
A lightweight, high-performance LLM proxy for caching, failover, cost tracking, and seamless integration between local and cloud AI providers.
llmproxy
What is llmproxy?
llmproxy is an open-source Flask server that emulates the HTTP APIs of Ollama, OpenAI, and llama.cpp, forwarding requests to NVIDIA's OpenAI-compatible API for seamless integration without client-side changes.
llmproxy vs Similar AI Tools
| Pricing Model | Free | Custom Pricing | Free | Free, Freemium |
| Free Credits | ||||
| Key Features |
|
|
|
|
| Pros |
|
|
|
|
| Cons |
|
|
|
|
| Best For |
|
|
|
|
How to use llmproxy?
- 1Configure your NVIDIA API key in .env file.
- 2Run with Docker Compose: docker compose up -d.
- 3Test with curl: curl http://localhost:11434/.
llmproxy Key Features
- Emulates Ollama, OpenAI, and llama.cpp APIs
- Transparent forwarding to NVIDIA's OpenAI-compatible API
- Optional response caching with configurable TTL and size
- Automatic failover and retry on transient upstream errors
- Live /stats dashboard for metrics and process monitoring
- Inbound authentication support
- Multi-model discovery
- Streaming support for chat and completions
llmproxy Use Cases
- Integrate local LLM tools with NVIDIA cloud-hosted models without client modifications
- Monitor and track API costs and usage via stats dashboard
- Reduce latency and API calls with response caching for non-streaming requests
llmproxy Pricing & Free Credits
llmproxy currently operates on a Free model.
This tool is completely free to use
llmproxy Pros & Cons
Pros
- Lightweight and easy to deploy via Docker
- Caches responses to reduce API calls and latency
- Automatic failover and retries improve reliability
- Provides cost tracking and live metrics
- Works with existing clients without changes
Cons
- Only forwards to NVIDIA's API; no other cloud provider support
- Requires a valid NVIDIA API key
- Caching only for non-streaming responses
What is llmproxy best for?
- Developers integrating NVIDIA LLMs into existing workflows
- Users of Open WebUI, curl, or SDKs wanting to leverage NVIDIA models
- Teams needing cost tracking and caching for LLM API calls