> ## Documentation Index > Fetch the complete documentation index at: https://docs.conare.ai/llms.txt > Use this file to discover all available pages before exploring further. # Create embeddings > Embed up to 96 texts with the same self-hosted model that indexes memories, so your own vector features live in the exact representation Conare retrieves with. - Stateless: nothing is stored, and no `endUserId` is involved. - Vectors are unit-normalized and returned in input order; each input is at most 8,192 characters. Metered as embed tokens (~4 characters per token). - `model` is a pin, not a selector: the endpoint serves exactly one model (named in every response), and any other name is a `400` (`model_not_served`). Pin it whenever vectors must stay compatible with an existing index. ## OpenAPI ````yaml /api/openapi.yaml post /api/v1/embeddings openapi: 3.1.0 info: title: Conare API version: 1.2.0 summary: Per-end-user memory for AI applications. description: > The Conare API gives your application a persistent, personalized memory layer. Production applications authenticate as an organization-owned Integration. Every request names an `endUserId`, and that end user's memory lives in its own isolated ConareDB namespace. The physical tenant is an opaque HMAC derived from the Integration and end-user ID; clients cannot select namespaces or address another tenant. On top of storage the API provides hybrid retrieval (vector + BM25, fused and reranked, sub-second), LLM-synthesized **deep recall**, and **proactive suggestions** grounded in a user's memory. Authentication uses a scoped server-side Integration Bearer key (`cint_...`). The secret is shown once and persisted by Conare only as a SHA-256 digest. Keep it in a server-side secret manager; never ship it to the browser. Legacy personal `cmem_...` keys remain accepted during migration. Every response includes `X-Request-Id`. Send a caller-generated safe value (1–128 characters; first alphanumeric, then `A-Za-z0-9._:-`) to correlate one request through the Conare edge and memory plane; otherwise Conare generates one. Errors use a stable `{ statusCode, code, message, requestId, details? }` envelope and never include internal tenant IDs or backend error text. contact: name: Conare url: https://conare.ai email: artem@conare.ai servers: - url: https://api.conare.ai description: Production (dedicated partner-API host) - url: https://conare.ai description: Production (alias — same API, on the main host) security: - bearerAuth: [] tags: - name: Integration description: Credential, configuration, and backend readiness. - name: Memories description: Write and search a single end user's memory. - name: Recall description: LLM-synthesized personalization primitives. - name: Embeddings description: Embed text with Conare's own retrieval embedding model. - name: Connectors description: >- Let end users connect any of ~200 data sources that flow into their memory. - name: Users description: End-user lifecycle (GDPR). paths: /api/v1/embeddings: post: tags: - Embeddings summary: Create embeddings description: | Embed up to 96 texts with the same self-hosted model that indexes memories, so your own vector features live in the exact representation Conare retrieves with. - Stateless: nothing is stored, and no `endUserId` is involved. - Vectors are unit-normalized and returned in input order; each input is at most 8,192 characters. Metered as embed tokens (~4 characters per token). - `model` is a pin, not a selector: the endpoint serves exactly one model (named in every response), and any other name is a `400` (`model_not_served`). Pin it whenever vectors must stay compatible with an existing index. operationId: createEmbeddings requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/EmbeddingsRequest' examples: default: value: input: - User prefers smaller islands, wants to avoid crowds. - ASA 104 certified, comfortable with bareboat charters. responses: '200': description: One vector per input, in input order. content: application/json: schema: $ref: '#/components/schemas/EmbeddingsResponse' '400': $ref: '#/components/responses/BadRequest' '401': $ref: '#/components/responses/Unauthorized' '403': $ref: '#/components/responses/Forbidden' '503': $ref: '#/components/responses/ServiceUnavailable' components: schemas: EmbeddingsRequest: type: object required: - input properties: input: type: array minItems: 1 maxItems: 96 description: Texts to embed, each at most 8192 characters. items: type: string minLength: 1 maxLength: 8192 model: type: string minLength: 1 description: >- Optional pin on the expected embedding model. The endpoint serves exactly one model; naming any other is a 400 (`model_not_served`), never a silent substitution. EmbeddingsResponse: type: object required: - success - model - dims - embeddings - usage properties: success: type: boolean model: type: string description: Identifier of the serving embedding model. dims: type: integer description: Dimensionality of every returned vector. example: 1024 embeddings: type: array description: One unit-normalized vector per input, in input order. items: type: array items: type: number usage: type: object required: - totalChars - approxTokens properties: totalChars: type: integer description: Total characters across all inputs. approxTokens: type: integer description: Billed embed-token quantity (~4 characters per token). Error: type: object properties: statusCode: type: integer code: type: string description: Stable snake_case machine error code. message: type: string requestId: type: string description: Correlates this response through Conare services and logs. details: type: object additionalProperties: true required: - statusCode - code - message - requestId responses: BadRequest: description: Missing or invalid fields (e.g. bad `endUserId`, empty `query`). content: application/json: schema: $ref: '#/components/schemas/Error' Unauthorized: description: Missing or invalid Bearer key. content: application/json: schema: $ref: '#/components/schemas/Error' Forbidden: description: >- The Integration credential does not include the required operation scope. content: application/json: schema: $ref: '#/components/schemas/Error' ServiceUnavailable: description: Integration namespace routing or the memory backend is not ready. headers: X-Request-Id: $ref: '#/components/headers/RequestId' content: application/json: schema: $ref: '#/components/schemas/Error' headers: RequestId: description: Caller-provided or generated request correlation identifier. schema: type: string securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: cint_ description: | A scoped Integration key, prefixed `cint_`. Send as `Authorization: Bearer cint_...` on every request. Server-side only. Legacy personal `cmem_...` keys remain accepted during migration; team/org-scoped `cmem_...` keys are not valid on this API. ````