Sending Attachments via the Inbound Webhook

The inbound web-hook accepts your JSON payload plus an optional top-level _attachments array. Everything else in the body is delivered to the AOP as trigger data; each _attachments entry becomes a standard attachment the AOP can access and parse (e.g., with the Document Parsing tool).

Two ways to send a file

Each _attachments entry must contain either a url or a data_b64 field, plus filename and mime.

Option A — URL reference

Use when the file is already hosted somewhere Leena's servers can reach.

{
  "ticket_number": "INC0012345",
  "short_description": "Printer not working",
  "_attachments": [
    {
      "url": "https://files.example.com/error-screenshot.jpg",
      "filename": "error-screenshot.jpg",
      "mime": "image/jpeg"
    }
  ]
}

• The URL is passed through as-is and fetched only when the AOP's parsing step runs — it must be accessible without an interactive login at that time (public or pre-signed link).
• Links that require an authenticated session (e.g., raw ServiceNow attachment URLs) will fail at parse time — use Option B instead.

Option B — Inline base64 (recommended for ServiceNow and other closed systems)

{
  "ticket_number": "INC0012345",
  "short_description": "Printer not working",
  "_attachments": [
    {
      "data_b64": "<base64-encoded file content>",
      "filename": "error-screenshot.jpg",
      "mime": "image/jpeg"
    }
  ]
}

• Leena decodes the content, stores it in Leena AI's secure file storage, and provides the AOP a short-lived internal URL. Nothing on your side needs to be publicly exposed.

Limits

Supported file types for parsing

The Document Parsing skill routes by the filename extension, so it must match the real file type:

• Images: .png, .jpg, .jpeg, .gif (auto-converted/compressed to JPEG internally)
• Documents: .pdf (up to 200 MB via URL), .docx / .doc (up to 3 MB)

Other formats (WEBP, BMP, TIFF, HEIC, etc.) are rejected at parse time with an "Unsupported document format" error.

Error behavior

Using web-hook attachments in an AOP

When the web-hook fires, the AOP receives two things separately:

• Trigger data — your JSON body (with _attachments stripped out), available to reference in instructions (e.g., ticket number, description).
• Attachments — each entry surfaced to the agent as Name / Type / URL, ready to be passed to skills. Base64 uploads appear as Leena-hosted URLs; URL entries appear as-is.

To read an attachment's content, the AOP (or its AI Colleague) must have the Document Parsing skill enabled. It takes two inputs — the attachment URL and a natural-language query — and parses + answers in a single step.

Sample AOP instructions

## Step 1 "check_attachments": Check for attached images
If the trigger includes one or more image attachments (JPEG/PNG), run Step 2
for each image. If there are no attachments, skip to Step 3.

## Step 2 "parse_image": Extract details from the image
Use the Document Parsing skill:
- document: the attachment URL exactly as provided
- query: "Extract all visible text, error messages, codes, dates and any
  equipment/asset identifiers from this image. Note any illegible parts."
If parsing fails (unsupported format or file too large), do not retry more
than once; record the failure reason and continue to Step 3.

## Step 3 "summarize": Combine and act
Combine the ticket fields from the trigger data with the extracted image
content, then continue the workflow (e.g., post a work note, categorize,
or escalate).

Authoring tips

• Be specific in the parsing query — name the exact fields to extract. A vague query returns a vague summary.
• Use the attachment URL exactly as provided; don't ask the agent to rewrite or shorten it.
• Always add a failure branch (as in Step 2): parse errors are returned to the agent rather than stopping the run, so tell it what to do when a file can't be read.
• If multiple attachments may arrive, say explicitly whether to process each one or only the first.

Quick checklist

1. Put ticket/business fields at the top level of the JSON; files go in _attachments.
2. Prefer base64 for files from systems that require login; keep each file ≤ 5 MB.
3. Set filename with the correct extension and an accurate mime.
4. Handle the 400/413 responses on your side (the run won't start); check Run History for the reason.