LLMSee

LLMSee is a cross-platform lightweight proxy for logging OpenAI-compatible API calls.

Features

• Platform Compatibility: A single executable that runs on Linux, MacOS, Windows, or via Docker.
• Built-in Web Interface: Offers simple web UI to monitor logged requests in real time.
• Supports Multiple Providers: Supports transparent proxying to multiple LLM providers.

Quick Start

1. Create a ~/.config/llmsee.json configuration file with your provider details.

2. Run llmsee to start the server.

3. Access the LLMSee Web UI at: http://localhost:5050/ui/.

4. Configure your application:

• API_BASE_URL=http://localhost:5050/v1
• OPENAPI_API_KEY=none

6. Start using your app and see real-time updates in the LLMSee Web UI

Configuration

The configuration file should be set up as follows:

{
	"host": "localhost",
	"port": 5050,
	"pageSize": 15,
	"providers": {
		"ollama": {
			"baseUrl": "http://localhost:11434/v1",
			"apikey": ""
		},
		"openai": {
			"baseUrl": "https://api.openai.com/v1",
			"apikey": "<OpenAI API Key>"
		},
		"deepseek": {
			"baseUrl": "https://api.deepseek.com",
			"apikey": "<Deepseek API Key>"
		},
		"gemini": {
			"baseUrl": "https://generativelanguage.googleapis.com/v1beta/openai",
			"apikey": "<Gemini API Key>"
		},
		"groq": {
			"baseUrl": "https://api.groq.com/openai/v1",
			"apikey": "<Groq API Key>"
		},
		"openrouter": {
			"baseUrl": "https://openrouter.ai/api/v1",
			"apikey": "<openrouter API key>"
		},
		"provider...": ...etc...
	}
}

Building

To build the project, you will need Go installed:

make

To build the Docker image, run this command:

make docker

TODOs

• Update documentation
• Clean up and refactor (cmd/, internal/)
• Add search
• Add database auto-trimming (max N records)

Contributing

Contributions are welcome! Please fork the repository and submit a pull request.

License

LLMSee is made by @yz778 and licensed under the MIT License. See LICENSE for more details.