Embeddings

Vector generation, indexing, and similarity search.

Generation only

http

POST /v1/embeddings

json

{
  "input": ["Premier texte", "Deuxieme texte"],
  "model": null,
  "routing_strategy": "auto"
}

json

{
  "embeddings": [[0.0123, -0.0456, ...], [0.0789, ...]],
  "dimensions": 1536,
  "model": "text-embedding-3-small",
  "provider": "openai"
}

Indexing

Generates the vector and stores it in your organization's vector index (pgvector), with free-form metadata.

http

POST /v1/embeddings/index

json

{
  "content": "IleAI est une plateforme unifiee d'acces a des modeles d'IA.",
  "metadata": { "source": "faq", "id_interne": "42" }
}

json

{ "id": "3f7a1c2e-..." }

Indexing always forces text-embedding-3-small (routing_strategy "manual"), regardless of the model used for your generations: everything touching the index must share the same vector dimension.

Similarity search

http

POST /v1/embeddings/search

json

{
  "query": "Comment fonctionne le routage IA ?",
  "top_k": 5
}

json

{
  "results": [
    {
      "id": "3f7a1c2e-...",
      "content": "IleAI est une plateforme unifiee d'acces a des modeles d'IA.",
      "metadata": { "source": "faq", "id_interne": "42" },
      "model": "text-embedding-3-small",
      "distance": 0.1834
    }
  ]
}

distance is a cosine distance — the closer to 0, the more relevant the result.

Providers

OpenAI, Google.