> ## Documentation Index > Fetch the complete documentation index at: https://docs.locusmedical.fr/llms.txt > Use this file to discover all available pages before exploring further. # Grounded clinical answer with citations > Return a concise answer grounded ONLY in the clinical knowledge base, with the recommendations cited. Synchronous (non-streamed). ## OpenAPI ````yaml /openapi.json post /v1/answer openapi: 3.1.0 info: title: Locus API description: >- Locus knowledge layer — REST + MCP for clinical knowledge. Two categories of surface: - **query** (read) — `/v1/search`, `/v1/answer`, `/v1/context` (legacy Meili contract serving AlmaPro) and `/v2/retrieve` (knowledge-graph retrieval, MKF 0.1.3). - **ops** — `/health`. The write plane (PDF → Neo4j graph: ingest, summarize, MKU extraction) is NOT exposed over HTTP — it is a batch pipeline driven by the `locus` CLI, which runs where the source PDFs are. version: 0.2.0 servers: - url: https://core.locusmedical.fr description: Production security: [] paths: /v1/answer: post: tags: - query summary: Grounded clinical answer with citations description: >- Return a concise answer grounded ONLY in the clinical knowledge base, with the recommendations cited. Synchronous (non-streamed). operationId: answer_endpoint_v1_answer_post requestBody: content: application/json: schema: $ref: '#/components/schemas/AnswerRequest' required: true responses: '200': description: Successful Response content: application/json: schema: $ref: '#/components/schemas/AnswerResponse' '422': description: Validation Error content: application/json: schema: $ref: '#/components/schemas/HTTPValidationError' security: - lsk: [] components: schemas: AnswerRequest: properties: query: type: string minLength: 1 title: Query description: Free-text clinical question in French. examples: - Quelle antibioprophylaxie pour une chirurgie colorectale ? source_ids: anyOf: - items: type: integer type: array - type: 'null' title: Source Ids description: >- Optional allow-list of source ids to restrict retrieval. Omit (the default) to search all sources. examples: - [] type: object required: - query title: AnswerRequest AnswerResponse: properties: query: type: string title: Query description: Echo of the submitted question. reformulated_question: anyOf: - type: string - type: 'null' title: Reformulated Question description: The question restated clearly by the model. answer: type: string title: Answer description: >- Concise answer grounded ONLY in the retrieved sources. If the knowledge base does not cover the question, a fixed 'not found' sentence. sources: items: $ref: '#/components/schemas/AnswerSource' type: array title: Sources description: Citations actually used — deduplicated by recommendation. steps: items: $ref: '#/components/schemas/Step' type: array title: Steps description: >- The agentic pipeline as input→output steps (search → rerank → answer). Display-only; for UIs that 'show the reasoning'. Safe to ignore. evidence: items: $ref: '#/components/schemas/EvidenceSource' type: array title: Evidence description: >- The sources that grounded the answer, as structured stacks (metadata + their excerpt units), grouped by recommendation. Richer than `sources` (the deduped citation line). Display-only. tokens_in: anyOf: - type: integer - type: 'null' title: Tokens In description: Prompt tokens billed. tokens_out: anyOf: - type: integer - type: 'null' title: Tokens Out description: Completion tokens billed. type: object required: - query - answer - sources title: AnswerResponse HTTPValidationError: properties: detail: items: $ref: '#/components/schemas/ValidationError' type: array title: Detail type: object title: HTTPValidationError AnswerSource: properties: reco_id: type: integer title: Reco Id description: Id of the cited recommendation. reco_title: type: string title: Reco Title description: Title of the cited recommendation. source_name: type: string title: Source Name description: Issuing source, e.g. 'HAS', 'SFAR'. deep_link: anyOf: - type: string - type: 'null' title: Deep Link description: Direct link to the source document. type: object required: - reco_id - reco_title - source_name title: AnswerSource description: A cited recommendation actually used to ground the answer. Step: properties: tool: type: string title: Tool description: The step's tool, e.g. 'search', 'rerank'. in: additionalProperties: true type: object title: In description: What the step was given. examples: - query: antibioprophylaxie out: additionalProperties: true type: object title: Out description: What the step produced. examples: - n_units: 8 type: object required: - tool - in - out title: Step description: |- One agentic step (input → output) in a pipeline. Surfaces the retrieve → rerank → synthesize loop so a UI can 'show the reasoning' (SPEC §4.2). Display-only; carries no business logic. EvidenceSource: properties: id: type: string title: Id description: Stable view id for this stack, e.g. 'src900'. org: type: string title: Org description: Issuing source, e.g. 'HAS', 'SFAR'. year: anyOf: - type: string - type: 'null' title: Year description: Publication year, if known. title: type: string title: Title description: Recommendation title. page: anyOf: - type: integer - type: 'null' title: Page description: Page in the source document, if known. url: anyOf: - type: string - type: 'null' title: Url description: Direct link to the original document. grade: anyOf: - type: string - type: 'null' title: Grade description: Evidence grade (A/B/C…) when the source carries one. units: items: $ref: '#/components/schemas/EvidenceUnit' type: array title: Units description: Structured excerpts from this source, most relevant first. type: object required: - id - org - title - units title: EvidenceSource description: >- A source as a CITABLE STACK: its metadata plus its structured units. The view-oriented projection of the chunks actually selected to ground the answer (post-retrieval + rerank), grouped by recommendation. It lets a UI render real, explorable source piles instead of a flat citation line. ValidationError: properties: loc: items: anyOf: - type: string - type: integer type: array title: Location msg: type: string title: Message type: type: string title: Error Type input: title: Input ctx: type: object title: Context type: object required: - loc - msg - type title: ValidationError EvidenceUnit: properties: text: type: string title: Text description: The excerpt text (a retrieved, reranked chunk). type: type: string title: Type description: Kind of unit, for display (extrait, recommendation…). default: extrait type: object required: - text title: EvidenceUnit description: One structured excerpt that grounded the answer — never a blob. securitySchemes: lsk: type: http description: Locus API key — paste the raw `lsk_…` value (no 'Bearer ' prefix). scheme: bearer ````