Embed API (Ollama Compatible)¶

Generate vector embeddings for semantic search and RAG applications with Amazon Bedrock embedding models through the Ollama /api/embed interface.

This is an alternate route to the OpenAI-compatible Embeddings API: both are served by the same embedding backends and models. Served under /ollama by default; the examples below use $BASE, which includes that prefix.

At a glance¶

Drop-in Ollama compatibility — Follows the Ollama /api/embed request and response shape, so an existing Ollama embedding client works by changing the base URL.
Semantic search — Turn one or several texts into dense vectors for similarity search that understands meaning, not just keywords.
Higher RAG quality — Build retrieval pipelines on the same high-quality embedding models served by the OpenAI-compatible Embeddings API.
Private AWS backend — Served entirely by Amazon Bedrock embedding models in your own AWS account — no traffic to third-party endpoints.
Differs from the Ollama API: truncate, keep_alive and options are accepted and ignored, and load_duration is never reported — see Limits and behaviour to know.

Base URL and route prefix

By default, all Ollama-compatible routes are prefixed with /ollama. This means the Embed API is available at /ollama/api/embed instead of /api/embed. You can customize this prefix using the OLLAMA_ROUTES_PREFIX configuration variable documented in HTTP Server and MCP.

The curl examples on this page use a $BASE variable that must include this prefix — set it to your scheme and host followed by OLLAMA_ROUTES_PREFIX:

export BASE="https://your-host/ollama"  # <scheme>://<host> + OLLAMA_ROUTES_PREFIX

curl -X POST "$BASE/api/embed" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "amazon.titan-embed-text-v2:0",
    "input": ["first", "second"]
  }'

Endpoints¶

Endpoint	Method	What It Does	Powered By	MCP Tool
`/api/embed`	`POST`	Embed one or several inputs, in request order	Amazon Bedrock embedding models	`ollama_embed`
`/api/embeddings`	`POST`	Legacy single-prompt embed, deprecated upstream	Amazon Bedrock embedding models	`ollama_embeddings`

Model Names¶

Send the model names GET /api/tags publishes — those are the canonical identifiers this server resolves directly. A trailing :latest is accepted and stripped as a fallback when the exact name is not found. Short aliases accepted on this server's other APIs also work here even though /api/tags does not list them. A name learned from ollama.com is not available through this server and answers 404. /api/embed echoes the model name exactly as the request spelled it, in model; the legacy /api/embeddings returns the vector alone, as it does upstream.

Find compatible models: Call /search_models with route=ollama_embed to discover model IDs that support embeddings.

Feature compatibility¶

Feature	Status	Notes
Input
`input` (string)		Full support
`input` (array)		One vector per entry, returned in request order
`dimensions`		Sets the vector width on models that support dimension reduction
`truncate`		Accepted and ignored
`keep_alive`		Accepted and ignored — models are never resident
`options`		Accepted and ignored — no runner options apply to embeddings
Extra model-specific params		Extra fields are forwarded as additional model request parameters
Output
`embeddings`		One vector per input, in request order
`total_duration`		Real wall-clock time
`prompt_eval_count`		Real input token count
`load_duration`		Never reported — there is no model-loading phase to measure

Legend:

Supported — Fully compatible with the Ollama API
Available on Select Models — Check your model's capabilities
Unsupported — Not available in this implementation
Extra Feature — Enhanced capability beyond the Ollama API

Ollama Legacy Embeddings API (`/api/embeddings`)¶

/api/embeddings is the singular form Ollama itself has deprecated in favor of /api/embed. It takes a single prompt instead of input, embeds it against the same models, and returns only the vector — no token count, no duration:

curl -X POST "$BASE/api/embeddings" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "amazon.titan-embed-text-v2:0",
    "prompt": "first"
  }'

{"embedding": [0.012, -0.034, ...]}

Deprecated

Prefer /api/embed: it embeds several inputs in one call and reports token usage, which /api/embeddings cannot.

How It Works¶

Requests are served by the same embedding backends and models as the OpenAI-compatible Embeddings API — anything supported there through model is reachable here through the same identifier.

Limits and behaviour to know¶

truncate, keep_alive and options are accepted and ignored.
load_duration is never reported: there is no model-loading phase to measure, and a number there would be invented.
/api/embeddings never reports prompt_eval_count or total_duration — it only ever returns the vector, matching Ollama's own legacy response shape.
A model name learned from ollama.com names nothing this server serves and answers 404; send a name GET /api/tags publishes.

Request headers¶

Header	Purpose	Notes
`Authorization`	Gateway API key	`Bearer <key>`, required like every other route

A local Ollama server needs no key; this one does, on every route.

Try it¶

curl -X POST "$BASE/api/embed" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "amazon.titan-embed-text-v2:0",
    "input": ["first", "second"]
  }'

Example response:

{
  "model": "amazon.titan-embed-text-v2:0",
  "embeddings": [[0.012, -0.034, ...], [0.041, 0.007, ...]],
  "total_duration": 214567890,
  "prompt_eval_count": 4
}

Next steps¶

Next: Chat API · Models API · OpenAI Embeddings API · Search Models API