[Bug]: OpenViking v0.2.8 litellm VLM backend rewrites Z.AI model prefix incorrectly (zai/... → zhipu/zai/...) #784
Description
Bug Description
Summary
When using OpenViking v0.2.8 with Docker and configuring vlm.provider = "litellm" for Z.AI / GLM models, OpenViking appears to rewrite the model name
incorrectly.
LiteLLM docs say Z.AI models should use the zai/ prefix, for example:
zai/glm-4.7zai/glm-4.5
However, in OpenViking v0.2.8, the final model passed to LiteLLM becomes:
zhipu/zai/glm-4.5
which then causes LiteLLM to fail with:
LLM Provider NOT provided
This makes the litellm path for Z.AI / GLM unusable in my deployment.
Environment
- OpenViking version:
v0.2.8 - Deployment: Docker
- OpenClaw plugin connected in remote mode
- Server runs normally on
0.0.0.0:1933
My ov.conf VLM config
{
"vlm": {
"provider": "litellm",
"api_key": "<redacted>",
"model": "zai/glm-4.5",
"api_base": "https://open.bigmodel.cn/api/coding/paas/v4"
}
}
### Steps to Reproduce
1. Start OpenViking with Docker using the config above.
2. Confirm the server is healthy:
```bash
curl http://127.0.0.1:1933/health - Create a tenant user and use a tenant user key for normal session/memory APIs.
- Create a session:
curl -X POST http://127.0.0.1:1933/api/v1/sessions \
-H "X-API-Key: <tenant-user-key>" \
-H "X-OpenViking-Agent: wk-openclaw-main" \
-H "Content-Type: application/json" \
-d '{}' - Add one message into that session:
curl -X POST http://127.0.0.1:1933/api/v1/sessions/<session_id>/messages \
-H "X-API-Key: <tenant-user-key>" \
-H "X-OpenViking-Agent: wk-openclaw-main" \
-H "Content-Type: application/json" \
-d '{"role":"user","content":"用户排查问 题时，习惯先确认权限、鉴权和系统接入是否正常，再去看模型和策略层 问题。"}' - Trigger extract:
curl -X POST http://127.0.0.1:1933/api/v1/sessions/<session_id>/extract \
-H "X-API-Key: <tenant-user-key>" \
-H "X-OpenViking-Agent: wk-openclaw-main" \
-H "Content-Type: application/json" \
-d '{}' - Check Docker logs:
docker logs --tail 200 openviking Expected Behavior
OpenViking should pass a valid Z.AI LiteLLM model name through, e.g.:
- zai/glm-4.5
or otherwise use the correct provider handling expected by current LiteLLM.
Actual Behavior
OpenViking seems to rewrite / resolve the model incorrectly and LiteLLM reports:
litellm.BadRequestError: LLM Provider NOT provided.
You passed model=zhipu/zai/glm-4.5
This happens during memory extraction / semantic processing.
Minimal Reproducible Example
Error Logs
OpenViking Version
0.2.8
Python Version
3.13
Operating System
Linux
Model Backend
None
Additional Context
No response