Bug type

Behavior bug (incorrect output/state without crash)

Summary

Setup:

Mac with Docker
OpenClaw running in Docker container
Ollama running on Mac via brew services
Model: llama3.2:3b

Problem:
When a message is sent via Telegram, the embedded agent fails immediately with error=fetch failed in under 10ms.
What works:

curl http://host.docker.internal:11434/api/tags from inside container ✅
node -e "fetch('http://192.168.65.254:11434/api/chat'...)" from inside container ✅
Ollama responds correctly to all direct requests ✅

What doesn't work:

OpenClaw embedded agent always returns fetch failed immediately

Config:
json"models": {
"providers": {
"ollama": {
"baseUrl": "http://192.168.65.254:11434",
"apiKey": "ollama-local",
"api": "ollama"
}
}
}

Steps to reproduce

Mac M4, Docker, Ollama via brew services, llama3.2:3b

Expected behavior

(the AI agent) responds to messages sent via Telegram using the local Ollama model

Actual behavior

Embedded agent fails immediately with "fetch failed" error in under 10ms. No response is sent.

OpenClaw version

local (built from source via Docker)

Operating system

macOS, Apple M4, Docker Desktop

Install method

No response

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response