3. Discover models - co-mind.ai

Before you call chat or embeddings, list what’s actually available on this deployment. Which models appear depends on how your administrator has configured backends — self-hosted (Ollama, vLLM) and cloud (OpenAI, Anthropic, Mistral, Gemini, Groq, SambaNova, Cohere) can be mixed.

List models

curl {BASE_URL}/v1/models \
  -H "Authorization: Bearer $PAT"

{
  "object": "list",
  "data": [
    { "id": "llama3.2:3b", "object": "model", "provider": "ollama" },
    { "id": "nomic-embed-text", "object": "model", "provider": "ollama" },
    { "id": "gpt-4o", "object": "model", "provider": "openai" }
  ]
}

Picking the right one

Chat-class and embedding-class models are separate — chat models won’t produce embeddings and vice versa. The id field is what you’ll pass to /v1/chat/completions, /v1/embeddings, and the knowledge-base chat endpoint.

Save a chat model id and an embedding model id from this list — you’ll use both in the next few steps.

export CHAT_MODEL="llama3.2:3b"
export EMBED_MODEL="nomic-embed-text"

Next: Compute embeddings →