Repository navigation
[Guide] How to use NVIDIA NIM / API (Nemotron) with PicoClaw on Windows #3446
rahul23145
started this conversation in
Show and tell
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Hey everyone! 👋
If you tried using NVIDIA NIM / NVIDIA API (https://integrate.api.nvidia.com) with PicoClaw as an OpenAI-compatible provider and couldn't get it to connect, here is why it happens and a working solution you can use right now.
Why PicoClaw fails with NVIDIA API out-of-the-box
NVIDIA API strictly expects the provider prefix in
/chat/completions(e.g.,nvidia/nemotron-3-super-120b-a12b). PicoClaw sends it without thenvidia/prefix, which triggers a400 Bad Request./modelsValidation Check:PicoClaw checks
GET /modelson initialization. Differences in the schema returned by NVIDIA cause startup validation to fail.The Fix: Lightweight Local Bridge Proxy
You can run this minimal Python bridge proxy (zero extra dependencies, pure standard library) on
127.0.0.1:18888. It mocks/modelsand automatically prependsnvidia/to incoming chat requests:How to configure PicoClaw
config.json, add the model undermodel_list:{ "model_name": "nvidia-nemotron", "provider": "openai", "model": "nemotron-3-super-120b-a12b", "api_base": "http://127.0.0.1:18888/v1" }.security.yml, add your NVIDIA API key (nvapi-...):python nvidia_proxy.pyand then launch PicoClaw!Hope this helps anyone wanting to use NVIDIA's free inference models with PicoClaw on Windows! 🦞
All reactions