Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

OpenClaw agent times out when calling Ollama qwen2.5:7b model, despite direct curl to Ollama API succeeding in 7-8 seconds.

Steps to reproduce

1. docker compose up -d
2. docker compose exec openclaw-gateway openclaw agent --session-id "test" --message "1+1="
3. Observe timeout error after 120 seconds

Expected behavior

Agent should return model response "1+1=2" within 10-15 seconds, matching direct API response time of 7-8 seconds.

Actual behavior

Agent returns error: "The model did not produce a response before the LLM idle timeout."
Gateway logs show: "[agent] embedded run failover decision: stage=assistant decision=surface_error reason=timeout provider=ollama/qwen2.5:7b"

OpenClaw version

2026.4.9

Operating system

Windows 11 with WSL2 (Ubuntu)

Install method

docker

Model

ollama/qwen2.5:7b

Provider / routing chain

openclaw -> ollama (direct, no proxy)

Additional provider/model setup details

{
"models": {
"mode": "replace",
"providers": {
"ollama": {
"baseUrl": "http://172.20.48.1:11434",
"apiKey": "ollama-local",
"api": "ollama",
"models": [{"id": "qwen2.5:7b", "name": "Qwen2.5 7B"}]
}
}
}
}

Logs, screenshots, and evidence

Gateway Logs (relevant entries)
text
[agent] embedded run failover decision: stage=assistant decision=surface_error reason=timeout provider=ollama/qwen2.5:7b
The model did not produce a response before the LLM idle timeout.
Direct API Test (SUCCESS)
bash
$ curl -X POST http://172.20.48.1:11434/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen2.5:7b","messages":[{"role":"user","content":"test"}],"stream":false}'
Response: Returns valid model output

Agent Test (FAILURE)
bash
$ openclaw agent --session-id "test" --message "1+1="
Response: The model did not produce a response before the LLM idle timeout.

Verified Working Endpoints
✅ POST /api/generate (native Ollama)

✅ POST /api/chat (Ollama chat API)

✅ POST /v1/chat/completions (OpenAI compatible)

✅ POST /api/chat with tools parameter

Version Regression Test
✅ Working: 2026.3.28-beta.1

❌ Broken: 2026.4.9

Impact and severity

Affected users/systems/channels: All users attempting to use OpenClaw agent with Ollama provider on version 2026.4.9

Severity: Blocks workflow (agent cannot complete any conversation)

Frequency: Always (100% of attempts)

Consequence: Agent cannot respond to any messages when using Ollama provider, making the agent functionality unusable

Additional information

Last known good version: 2026.3.28-beta.1

First known bad version: 2026.4.9

Workaround: Downgrade to version 2026.3.28-beta.1

Additional context:

Direct curl to Ollama works in 7-8 seconds

Network connectivity confirmed (container can reach Ollama)

Model supports tool calling (verified via API test)

Issue persists with timeout values set to 120s, 300s, and 0 (disabled)

Environment variables OPENCLAW_AGENT_MODEL_TIMEOUT do not resolve the issue