Skip to main content
Creates or updates the current user’s bot. Stage 1 supports one bot per user - the route is an upsert keyed on the session’s userId.

​
Request

POST /api/bots
Content-Type: application/json
Cookie: next-auth.session-token=<jwt>

​
Body

{
  "name": "Career Bot",
  "headline": "Senior platform engineer · open to staff roles",
  "personality": "professional",
  "contextText": "I am a software engineer based in...",
  "suggestedQuestions": [
    "What stack do you work in?",
    "Are you open to remote?"
  ],
  "llmProvider": "anthropic",
  "llmModel": "claude-sonnet-4-5"
}
FieldTypeConstraint
namestring1–100 chars, trimmed
headlinestring (optional)≤ 120 chars
personality"professional" | "creative" | "enthusiastic"
contextTextstring1–50,000 chars, trimmed
suggestedQuestionsstring[]≤ 6 entries, each ≤ 200 chars
llmProvider"anthropic" | "openai" | "google" | "azure"
llmModelstring (optional)≤ 60 chars; provider-specific model id or Azure deployment
The LLM API key is not part of this body and must not be sent here. It lives in the browser’s localStorage and rides the x-llm-api-key header on /api/chat/[botId] requests. See BYO-key flow.

​
Behaviour

The route runs in a single Drizzle transaction:
  1. Update the user’s llm_provider and llm_model columns.
  2. Look up an existing bot for the user.
  3. If present, UPDATE it; return 200 OK.
  4. Otherwise INSERT; return 201 Created.

​
Responses

​
201 Created (first save)

{
  "bot": {
    "id": "00000000-0000-0000-0000-000000000000",
    "userId": "11111111-1111-1111-1111-111111111111",
    "name": "Career Bot",
    "headline": "Senior platform engineer · open to staff roles",
    "personality": "professional",
    "contextText": "I am a software engineer based in...",
    "suggestedQuestions": [
      "What stack do you work in?",
      "Are you open to remote?"
    ],
    "loadingMessages": [
      "Thinking…",
      "Searching memory…",
      "Drafting a response…",
      "Almost ready…"
    ],
    "isActive": true,
    "createdAt": "2026-01-01T00:00:00.000Z",
    "updatedAt": "2026-01-01T00:00:00.000Z"
  }
}
Dates serialize as ISO-8601 YYYY-MM-DDThh:mm:ss.sssZ.

​
200 OK (subsequent saves)

Same shape, updatedAt advances.

​
400 Bad Request - invalid JSON

{ "error": "Invalid JSON body" }

​
400 Bad Request - validation failure

{
  "error": "Validation failed",
  "details": {
    "fieldErrors": {
      "personality": ["Invalid enum value..."]
    },
    "formErrors": []
  }
}

​
401 Unauthorized

{ "error": "Unauthorized" }
When no session cookie is attached or the session is invalid.

​
cURL

curl -X POST https://probot.vercel.app/api/bots \
  -H 'Content-Type: application/json' \
  -b 'next-auth.session-token=<JWT>' \
  -d '{
    "name": "Career Bot",
    "personality": "professional",
    "contextText": "I am a software engineer based in...",
    "suggestedQuestions": ["What stack do you work in?"],
    "llmProvider": "anthropic",
    "llmModel": "claude-sonnet-4-5"
  }'

​
Notes

  • loadingMessages is set server-side at INSERT time from a default JSON literal. It is not part of the request body in Stage 1.
  • isActive defaults to true and is not exposed for client updates yet - bot disable/enable is part of Stage 6.
  • Multi-bot support is planned (Stage 6+).