Integrations

Supported APIs

API	Request endpoint
OpenAI Chat Completions	POST /v1/chat/completions
OpenAI Responses	POST /v1/responses
Anthropic Messages	POST /v1/messages

All three support streaming and function tool calls. Image input requires a checkpoint with a vision tower; send data URLs, or enable HTTP(S) image fetching with --allow-image-urls.

Compatibility covers the implemented inference endpoints, not the entire upstream APIs. Responses is stateless: resend the conversation in input; previous_response_id and background requests are unsupported. Built-in tools, grammar-constrained JSON output and forced tool selection are unsupported.

Client base URLs:

• OpenAI (Chat Completions or Responses): http://127.0.0.1:8000/v1
• Anthropic (Messages): http://127.0.0.1:8000

First request

Load a model first. Replace model with its ID from GET /v1/models.

curl http://127.0.0.1:8000/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{"model":"Qwen3.8-Flash-Next","messages":[{"role":"user","content":"Hello!"}],"stream":true}'

Pi

Connect Pi to a running SnowLLM server:

snowllm pi --port 8000
snowllm pi --host 192.168.1.10 --port 8000

This writes the recipe models to ~/.pi/agent/models.json. --host defaults to localhost; use the inference server's port.

For LAN access, serve with --host 0.0.0.0. Model management uses a separate listener; see Web UI.