Skip to main content
POST
/
api
/
campaigns
/
{id}
/
generate-messages
Resolve messages or trigger AI generation
curl --request POST \
  --url https://app.puffle.ai/api/campaigns/{id}/generate-messages \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "action": "resolve"
}
'
{
  "messages": [
    {
      "id": "cccc1111-1111-1111-1111-111111111111",
      "campaign_id": "8c2b2d4e-e29b-41d4-a716-446655440000",
      "prospect_id": "dddd1111-1111-1111-1111-111111111111",
      "node_id": "bbbb1111-1111-1111-1111-111111111111",
      "subject": "Quick question about Acme",
      "body_text": "Hi Sarah, ...",
      "status": "draft",
      "manually_edited": false,
      "created_at": "2026-04-22T12:00:00Z"
    }
  ]
}

​
Overview

One POST, two actions, discriminated by the action field in the body:
  • resolve — synchronously resolves %variables% for every non-AI sendable node across every prospect and upserts campaign_prospect_messages rows as draft. Also copies pre-generated AI examples onto matching prospects (matched by lead name). Manually-edited rows are preserved; template changes flow through to unedited rows.
  • generate_ai — queues the generate-ai-messages Trigger.dev task for a specific AI node. Charges AI_MESSAGE_GENERATION credits per prospect after the task is queued. The pre-flight credit check returns 402 without destroying any existing messages.
Both actions require the campaign to be in draft status.

​
AI agent notes

resolve is safe to re-run. It upserts on (campaign_id, prospect_id, node_id) and excludes manually-edited rows from the upsert — template changes propagate without clobbering user customizations. If the “find edited messages” query itself fails, the server falls back to ignoreDuplicates: true for the whole batch to avoid silently overwriting edits.AI seeding matches on lead name. When resolve seeds AI drafts from generated_examples, it normalizes first_name + last_name and matches against prospects. Duplicate names yield multiple rows (one per prospect); unmatched examples are logged and skipped. After adding new prospects or regenerating the skeleton, re-run resolve to pick up the new matches.generate_ai credit flow. Credits are previewed before any write — a 402 response means no existing messages were deleted and no task was queued. Credits are only charged AFTER the task is successfully enqueued. For single-prospect regeneration (prospect_ids provided), the server deletes the existing draft for that (node, prospect) pair first so the task can produce a fresh version.Related endpoints. For reading the draft outbox or inline-editing a single message, see GET and PATCH on the same path. For building the framework before calling resolve, see generateCampaignContent.

Authorizations

​
Authorization
string
header
required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Path Parameters

​
id
string<uuid>
required

Campaign UUID

Pattern: ^([0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[1-8][0-9a-fA-F]{3}-[89abAB][0-9a-fA-F]{3}-[0-9a-fA-F]{12}|00000000-0000-0000-0000-000000000000|ffffffff-ffff-ffff-ffff-ffffffffffff)$

Body

application/json

Discriminated by action. resolve synchronously re-aggregates draft messages; generate_ai fires a background task and returns immediately.

​
action
enum<string>
required

Resolve %variables% for every non-AI sendable node across every prospect, seeding campaign_prospect_messages rows as draft. Also copies pre-generated AI examples onto matching prospects (by lead name). Safe to re-run — manually-edited drafts are preserved; unedited rows are overwritten so template changes flow through.

Available options:
resolve

Response

Action completed. Shape depends on action — inspect the example that matches your call.

Every message row touched by the resolve. Empty array when the campaign has no sendable nodes, no prospects, or no template changes propagated.

​
messages
object[]
required