Knowledge Base document processing always fails — missing S3 bucket env var in Trigger.dev

[crcorbett]

Knowledge Base document processing always fails — missing S3 bucket env var in Trigger.dev

Summary

Documents uploaded to the Knowledge Base via the API are stored in S3 but never get processed. The Trigger.dev process-knowledge-base-document task completes but returns an error because the APP_AWS_KNOWLEDGE_BASE_BUCKET environment variable is not set in the Trigger.dev runtime.

All documents end up with processingStatus: "failed" and are never indexed for AI features.

Reproduction

1. Upload a document to the Knowledge Base

# Base64 encode a PDF
FILE_DATA=$(base64 -i some-document.pdf)

curl -X POST https://api.trycomp.ai/v1/knowledge-base/documents/upload \
  -H "X-API-Key: $API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"organizationId\": \"$ORG_ID\",
    \"fileName\": \"some-document.pdf\",
    \"fileType\": \"application/pdf\",
    \"fileData\": \"$FILE_DATA\",
    \"description\": \"Test document\"
  }"

The upload succeeds and returns an id, name, and s3Key.

2. Trigger processing

curl -X POST https://api.trycomp.ai/v1/knowledge-base/documents/process \
  -H "X-API-Key: $API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"organizationId\": \"$ORG_ID\",
    \"documentIds\": [\"kbd_xxx\"]
  }"

Returns 201 with a Trigger.dev runId and publicAccessToken.

3. Check the Trigger.dev run

curl https://api.trigger.dev/api/v3/runs/$RUN_ID \
  -H "Authorization: Bearer $PUBLIC_ACCESS_TOKEN"

Expected result

The run should process the document (extract text, generate embeddings, index for AI search) and set processingStatus: "completed".

Actual result

The Trigger.dev run completes but with a failure output:

{
  "status": "COMPLETED",
  "isSuccess": true,
  "output": {
    "success": false,
    "documentId": "kbd_xxx",
    "error": "Knowledge base bucket is not configured. Please set APP_AWS_KNOWLEDGE_BASE_BUCKET environment variable in Trigger.dev."
  }
}

Note: Trigger.dev reports isSuccess: true because the task function itself didn't throw — it caught the error and returned a failure result. The document's processingStatus is set to "failed".

4. Verify the document status

curl "https://api.trycomp.ai/v1/knowledge-base/documents?organizationId=$ORG_ID" \
  -H "X-API-Key: $API_KEY"

Every document shows "processingStatus": "failed".

Root cause

The Trigger.dev worker environment is missing the APP_AWS_KNOWLEDGE_BASE_BUCKET environment variable. The documents are successfully uploaded to S3 (they have valid s3Key values), but when the processing task tries to read them back from S3, it doesn't know which bucket to look in.

Fix

Set the APP_AWS_KNOWLEDGE_BASE_BUCKET environment variable in your Trigger.dev project settings to match the S3 bucket used for Knowledge Base document uploads.

After fixing, re-process affected documents:

curl -X POST https://api.trycomp.ai/v1/knowledge-base/documents/process \
  -H "X-API-Key: $API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"organizationId\": \"$ORG_ID\",
    \"documentIds\": [\"kbd_id1\", \"kbd_id2\", ...]
  }"

Impact

• Knowledge Base is completely non-functional. No uploaded documents can be processed or indexed.
• AI features that depend on the Knowledge Base (questionnaire answering, policy chat, SOA auto-fill) cannot reference any uploaded documents.
• This affects all organizations, not just ours — the env var is a global Trigger.dev configuration.
• Documents in pending status (never had processing triggered) are also affected and will fail when processing is eventually triggered.

Our data

• 152 documents uploaded (95 showing failed, 57 still pending)
• File types: PDF, DOCX, XLSX, PNG
• All have valid s3Key values — the upload step works fine, only processing is broken

Environment

• API: https://api.trycomp.ai/v1
• Trigger.dev task: process-knowledge-base-document (version 20260227.2)
• Date discovered: 2026-02-27

[linear]

COMP-280