SapanBaseDevelopers

Generate a grounded answer

Single-turn "bring your own bot" endpoint: send a user question and get back an AI answer grounded in one knowledge base's published articles, plus the cited sources. Runs the same retrieval and generation pipeline as the HelpBento chat widget, but is stateless — no conversation is created. When no article clears the relevance floor the response is answered: false with no generated answer (your bot's signal to fall back to its own handling) and no AI cost is incurred. Requires the AI Support Agent feature on your plan and counts against your monthly AI usage. An Idempotency-Key makes network retries replay the original answer instead of generating (and paying for) a new one. Scope: answer:read.

POSThttps://api.helpbento.com/v1/answer

Authentication

AuthorizationstringheaderBearer <token>required

An API key, e.g. hb_live_… (live) or hb_test_… (test).

Header parameters

1 param
  • Idempotency-Keystringlength: 0…255

    Optional idempotency key (1–255 chars). Repeating a mutating request with the same key within 24h replays the original response instead of re-executing it; a concurrent in-flight request with the same key returns 409. Use a unique key (e.g. a UUID) per logical operation so network retries are safe.

Request body

application/jsonrequired
  • questionstringlength: 0…2000required

    The end user's question, verbatim.

  • knowledgeBaseIdstring

    Defaults to the company default KB (or a KB-scoped key's single KB).

  • maxSourcesintegerdefault: 5range: 1…10

    Max articles to ground the answer in (and cite).

Responses

  • answeredbooleanrequired

    False when no article cleared the relevance floor — no answer is generated and your bot should fall back.

  • answerstring

    The grounded answer (Markdown). Absent when answered is false.

  • sourcesarray of AnswerSourcerequired
  • modelstringrequired

    The Gemini model resolved from your plan.

  • usageAnswerUsage
Errors