V1 Public Embed Text

Embed text using Gemini Flash embeddings (768-dim).

Returns {embedding, model, dimensionality, task_type}. Enables agent-side HyDE composition — an agent can build a hypothetical-document embedding then pass it back through search_papers(anchor_embedding=...) (v3 surface) for hybrid retrieval.

POST is used (not GET) to accommodate long text payloads (HyDE hypothetical-document strings can be hundreds of chars).

task_type defaults to RETRIEVAL_DOCUMENT (matches paper-side embeddings, the correct space for HyDE flows). Use RETRIEVAL_QUERY for query-side semantic search.

Thin wrapper over the existing Gemini helpers in routers/papers.py — no duplicate Gemini client setup, no new embedding logic.

Requires a free account (any tier) — see _require_account. Each call is charged against the account's monthly quota; anonymous callers get 403 account_required.

Rate limit: 30/minute per account. Returns 502 if the upstream Gemini service fails (timeout / API error).

post/v1/embeddings

Request

Base URL: https://api.scholarfeed.org (the document declares no server; this is the origin it was published from)
URL: https://api.scholarfeed.org/v1/embeddings
Auth: HTTP bearer

Request body

textstring required

Text to embed (1-8000 chars).

task_type'RETRIEVAL_DOCUMENT' | 'RETRIEVAL_QUERY'

Embedding task type. RETRIEVAL_DOCUMENT (default) matches paper-side embeddings for HyDE flows; RETRIEVAL_QUERY matches query-side semantic search.

Response

Successful Response

object required

Changes

No recorded changes to this operation across all 1 revision of this API.