Local Ai
What Your Local LLM Actually Sees: Debugging Ollama Traffic in Quarkus with mitmproxy
This tutorial demonstrates how to use mitmproxy to inspect the actual HTTP traffic sent from a Quarkus application to a local Ollama model via its OpenAI-compatible endpoint, revealing the real JSO...
This tutorial demonstrates how to use mitmproxy to inspect the actual HTTP traffic sent from a Quarkus application to a local Ollama model via its OpenAI-compatible endpoint, revealing the real JSON payloads the model receives — including the model field, messages array, and system/user messages. It compares plain requests against tool-enabled requests, showing how tool metadata is appended to the payload and how enabling tools triggers a multi-step exchange where the model requests a tool call, the application executes it, and the result is returned to the model. The guide highlights that while Quarkus's built-in logging reflects what the client library chooses to print, mitmproxy captures what actually crosses the wire — and when the two differ, the wire capture should be trusted.
Related
- OpenClaude com Ollama Cloud
- Use the Same Model Across Ollama, LM Studio, Jan, and your Favorite Local AI Apps
- Those of you that run Openclaw with Ollama Pro, do you need the local ollama to use cloud?
- OpenClaw + Ollama + gemma4:26b is fast in raw Ollama, but first heavy OpenClaw turns are extremely slow or hit idle timeout
- Tried running LLMs locally to save API costs… ended up waiting 13 minutes for ONE response 🤡
Source: local-ai