Ollama Proxy 'Vessel' Released

The open-source proxy Vessel has been released to log and visualize prompts, token counts, response speeds, and other metrics for agents using Ollama and compatible APIs. This tool stores all requests in a local SQLite file and allows for searching, filtering, and per-agent tag management via a web UI.

The core feature is Request Replay. It allows captured requests to be re-executed on different models to compare the two responses and performance metrics side by side. It reads Ollama's native API statistics to accurately measure token generation speed (t/s) and Time to First Token (TTFT), and automatically detects model loading delays (cold starts).

Key Features

• Compatibility: Works in any environment with an API, including Ollama, OpenAI/Anthropic compatible endpoints, LM Studio, llama.cpp, and vLLM.
• Built-in MCP Server: Supports the Model Context Protocol (MCP) server, allowing coding agents to query data in real time.
• Easy Deployment: Provided as a single binary file with no additional dependencies, accounts, or telemetry features. It is released under the MIT license.