iblai-api-agent-dataset

iblai-api-agent-dataset

Manage an agent's training datasets (RAG) through the API: list an agent's training documents, add new resources to its knowledge base, train / untrain and set visibility, configure a retrain schedule, and delete datasets. Use when feeding an agent knowledge.

Auth & conventions

• Base URL: https://api.iblai.app
• Header: Authorization: Api-Token $IBLAI_API_KEY on every request.
• Path vars: {org} = $IBLAI_ORG, {username} = $IBLAI_USERNAME, {mentor} = the agent's unique id (e.g. d17dc729-60fd-4363-81a0-f67d9318b03e), used here as the pathway.
• Host: these endpoints live under …/dm/api/ai-index/….
• Not connected yet? Run /iblai-api-login first to populate IBLAI_ORG, IBLAI_USERNAME, and IBLAI_API_KEY.

Reads

• GET https://api.iblai.app/dm/api/ai-index/orgs/{org}/users/{username}/documents/pathways/{mentor}/?limit=5&offset={n}&search={q} — list training docs. Poll this every 2s while any document is pending.
• GET https://api.iblai.app/dm/api/ai-index/documents/{document_id}/settings/ — retrain schedule.
• GET https://api.github.com/repos/{owner}/{repo}/branches — list GitHub branches for a repo resource (external, no auth).

Writes

• POST https://api.iblai.app/dm/api/ai-index/orgs/{org}/users/{username}/documents/train/ — add a training resource (multipart/form-data); type varies:
  • File:

    {
      "file": "File (required)",
      "pathway": "{mentor}",
      "type": "file|<ext>",
      "user_image_description": "string"
    }
    

  • URL / YouTube / Blackboard:

    {
      "type": "url|youtube|blackboard",
      "pathway": "{mentor}",
      "url": "string (required)"
    }
    

  • Website crawl:

    {
      "type": "webcrawler",
      "pathway": "{mentor}",
      "url": "string",
      "crawler_max_depth": "number",
      "crawler_max_pages_limit": "number",
      "crawler_match_patterns": "string[]",
      "crawler_pattern_type": "glob|regex"
    }
    

  • GitHub:

    {
      "url": "repo url",
      "branch": "string",
      "pathway": "{mentor}",
      "type": "github"
    }
    

  • custom_metadata (optional, works with every type above) — a flat JSON object of tags stored on the document, later usable as a hard retrieval filter at chat time via document_filter (see /iblai-api-agent-session). Send it as a nested object on a JSON body, or — because train/ is multipart/form-data — as a JSON-encoded string form field:

    { "custom_metadata": { "stateCode": "CA", "productGroup": "LICENSING", "year": 2026 } }
    

    Rules (rejected with a validation error otherwise): keys must be flat and alphanumeric/underscore (^\w+$, no __); values must be scalars (string, number, or boolean) — no nested objects, arrays, or null. Stored on the document as metadata.custom_metadata and echoed back by the list endpoint above. Leave a tag off documents that should be exempt from a filter on that key — a document_filter only excludes documents that carry the key with a different value, so untagged/generic material always survives (see /iblai-api-agent-session ## Schema).
• PUT https://api.iblai.app/dm/api/ai-index/documents/{document_id}/ — train / untrain + visibility (+ retag):

  {
    "pathway": "{mentor}",
    "url": "string",
    "train": "boolean",
    "access": "public|private",
    "custom_metadata": { "stateCode": "CA" }
  }
  

  custom_metadata here replaces the document's stored tags (same validation as train/); omit it to leave existing tags unchanged.
• POST https://api.iblai.app/dm/api/ai-index/documents/{document_id}/settings/ — set retrain schedule:

  {
    "retrain_interval_days": "number (required)"
  }
  

• DELETE https://api.iblai.app/dm/api/ai-index/documents/{document_id}/ — delete a dataset. Destructive — confirm with the user first.