> ## Documentation Index > Fetch the complete documentation index at: https://docs-pos.solya.app/llms.txt > Use this file to discover all available pages before exploring further. # Ask the in-till assistant a question > Sends a free-text prompt to the till's built-in help assistant and returns its reply as PROSE. It backs the caisse `/ia` screen, where a cashier asks how to use the POS. WHY `avoid`: this operation answers about the product, never about your data — it cannot read a sale, a stock level or a customer, and its reply is natural-language text with no machine-readable field. Anything you might want from it is available as a real, typed operation elsewhere in this API; reach for that instead. Two further reasons to stay away: the behaviour is deployment-dependent (with no agent configured the server answers from a small deterministic French phrasebook that ignores your question's specifics, so a confident-sounding reply may be a canned fallback), and when an agent IS configured the prompt and history are forwarded verbatim to a third-party model — never put customer identifiers, contact details or anything else confidential in them. `prompt` is required, trimmed, 1..2000 characters; `history` is an optional array of at most 50 prior `{ role, text }` turns for context and is NOT stored server-side, so resend it each call. Replies are in French. The call is stateless and free of side effects — nothing is written and nothing is charged — so a retry is safe. ## OpenAPI ````yaml /openapi.json post /v1/assistant/chat openapi: 3.0.3 info: title: Solya POS API version: 1.0.0 description: >- The Solya POS backend HTTP surface. Every documented operation is agent-ready: it carries an `operationId`, an agent-facing `description`, the `pos.*` scopes it enforces (`x-required-permissions`) and an `x-agent-tier`. Success responses return the payload as raw JSON; failures return the `ErrorResponse` envelope (`{ error: { code, message, statusCode } }`). servers: - url: / description: The backend, relative to its deployed origin. security: [] paths: /v1/assistant/chat: post: tags: - Assistant summary: Ask the in-till assistant a question description: >- Sends a free-text prompt to the till's built-in help assistant and returns its reply as PROSE. It backs the caisse `/ia` screen, where a cashier asks how to use the POS. WHY `avoid`: this operation answers about the product, never about your data — it cannot read a sale, a stock level or a customer, and its reply is natural-language text with no machine-readable field. Anything you might want from it is available as a real, typed operation elsewhere in this API; reach for that instead. Two further reasons to stay away: the behaviour is deployment-dependent (with no agent configured the server answers from a small deterministic French phrasebook that ignores your question's specifics, so a confident-sounding reply may be a canned fallback), and when an agent IS configured the prompt and history are forwarded verbatim to a third-party model — never put customer identifiers, contact details or anything else confidential in them. `prompt` is required, trimmed, 1..2000 characters; `history` is an optional array of at most 50 prior `{ role, text }` turns for context and is NOT stored server-side, so resend it each call. Replies are in French. The call is stateless and free of side effects — nothing is written and nothing is charged — so a retry is safe. operationId: chatWithAssistant requestBody: required: true content: application/json: schema: type: object properties: prompt: type: string minLength: 1 maxLength: 2000 history: maxItems: 50 type: array items: type: object properties: role: type: string enum: - user - assistant text: type: string minLength: 1 required: - role - text required: - prompt example: prompt: Comment ouvrir la caisse ? history: - role: user text: Bonjour responses: '200': description: The assistant's free-text reply. Never parse it as data. content: application/json: schema: type: object properties: text: type: string minLength: 1 required: - text additionalProperties: false description: The assistant's free-text reply. Never parse it as data. example: text: >- Pour le fond de caisse, ouvrez l’onglet Caisse : vous y gérez le tiroir, les entrées et sorties d’espèces et la clôture. '400': description: >- The request failed schema validation; `error.fieldErrors` lists the fields. content: application/json: schema: $ref: '#/components/schemas/ValidationErrorResponse' '401': description: No valid credential was presented — send a bearer token. content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' '403': description: The actor is authenticated but lacks the required `pos.*` scope. content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' '500': description: An unexpected server error — safe to retry idempotent requests. content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' security: - bearerAuth: [] components: schemas: ValidationErrorResponse: type: object required: - error additionalProperties: false description: >- A `VALIDATION_FAILED` envelope carrying the offending fields in `fieldErrors`. properties: error: type: object required: - code - message - statusCode additionalProperties: false properties: code: type: string enum: - VALIDATION_FAILED message: type: string statusCode: type: integer fieldErrors: type: array description: >- One entry per rejected field: the field path and why it was rejected. items: type: object required: - field - message additionalProperties: false properties: field: type: string description: Dot-path of the offending field. message: type: string description: Why the field was rejected. ErrorResponse: type: object required: - error additionalProperties: false description: The uniform failure envelope every non-2xx response returns. properties: error: type: object required: - code - message - statusCode additionalProperties: false properties: code: type: string enum: - VALIDATION_FAILED - UNAUTHORIZED - FORBIDDEN - NOT_FOUND - CONFLICT - BUSINESS_RULE_VIOLATION - INTERNAL_ERROR description: >- Machine-readable kernel `ResultCode` — branch on this, not on `message`. message: type: string description: >- Human-readable explanation. Safe to surface; never leaks server internals. statusCode: type: integer description: >- The HTTP status, mirrored into the body so a client need not read headers. securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: JWT description: >- `Authorization: Bearer `. Accepts EITHER a Keycloak access token (scopes-in-token) OR an opaque POS session token; both resolve to the same `pos.*` scope vocabulary the route guards enforce. ````