omniroute-embeddings

name	omniroute-embeddings
description	Embeddings via OmniRoute using OpenAI /v1/embeddings format with auto-fallback across text-embedding-3-large, Voyage, Cohere, Gemini embeddings, Jina. Use when the user needs vector embeddings for RAG, similarity search, or clustering.

OmniRoute — Embeddings

Requires OMNIROUTE_URL and OMNIROUTE_KEY. See entry-point SKILL for setup.

Endpoint

• POST $OMNIROUTE_URL/v1/embeddings

Discover

curl $OMNIROUTE_URL/v1/models/embedding | jq '.data[]'

Each entry: { id, owned_by, dimensions, max_input_tokens }.

Example

curl -X POST $OMNIROUTE_URL/v1/embeddings \
  -H "Authorization: Bearer $OMNIROUTE_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "text-embedding-3-large",
    "input": ["first text", "second text"],
    "encoding_format": "float"
  }'

Response: { data:[{ embedding:[...], index }], usage:{ prompt_tokens, total_tokens } }

Batch input

input accepts a string or array of strings (up to provider batch limit, typically 2048 items).

Errors

• 400 input_too_long → input exceeds max_input_tokens for this model
• 400 invalid_encoding_format → use float or base64
• 503 → provider unavailable; try another model in /v1/models/embedding