openclaw - 💡(How to fix) Fix [Bug]: Embedded agent times out with Ollama (qwen2.5:1.5b) while direct Ollama works [16 comments, 15 participants]

Bug type

Behavior bug (incorrect output/state without crash)

Beta release blocker

No

Summary

OpenClaw (v2026.3.31) successfully connects to a local Ollama model and passes all probes, but chat generation via OpenClaw consistently times out, while direct Ollama API calls succeed normally. This appears to be an issue with the embedded agent runtime or timeout handling, not with Ollama itself.

Steps to reproduce

1. Direct Ollama works curl http://127.0.0.1:11434/api/chat Response: "Hello! It's nice to meet you..." Latency: ~4 seconds
2. OpenClaw API times out curl http://127.0.0.1:18789/v1/chat/completions Response: Request timed out before a response was generated.
3. Model probe succeeds openclaw models status --probe Result: ollama/qwen2.5:1.5b → ok · ~10–15s
4. Embedded agent failure log Profile ollama:default timed out. Trying next account...

embedded_run_failover_decision: failoverReason: "timeout" provider: "ollama" model: "qwen2.5:1.5b" fallbackConfigured: false 5. Timeout behavior inconsistency agents.defaults.timeoutSeconds was set to 120 seconds Actual failure occurs at ~60 seconds Suggests timeout config is not respected in embedded run

Expected behavior

Calling: POST /v1/chat/completions should return a valid model response via OpenClaw using the configured Ollama model.

Actual behavior

The request always returns: Request timed out before a response was generated. Even though: Ollama responds correctly via direct API OpenClaw model probe succeeds Gateway and API endpoints are working

OpenClaw version

2026.3.31

Operating system

Raspberry Pi (Linux 6.12.x)

Install method

Openclaw Onboard

Model

Model: qwen2.5:1.5b

Provider / routing chain

Ollama: local instance (default port 11434)

Additional provider/model setup details

Configuration openclaw.json (simplified + restored versions tested) "agents": { "defaults": { "timeoutSeconds": 120, "model": { "primary": "ollama/qwen2.5:1.5b" } } } Both: full config simplified config → produce the same timeout behavior

Auth profile (auto-generated) ~/.openclaw/agents/main/agent/auth-profiles.json "ollama:default": { "type": "api_key", "provider": "ollama", "key": "ollama-local" }

Logs, screenshots, and evidence

Impact and severity

Impact OpenClaw cannot be used as an OpenAI-compatible gateway for local Ollama models Blocks usage for local AI assistant setups (e.g., Raspberry Pi AI Assistant)

Additional information

Suspected Cause One of the following: Embedded agent runtime timeout not aligned with timeoutSeconds Ollama streaming / response handling not properly awaited Early termination in provider execution layer Known issue with local provider execution path (Ollama)