Document Pipelines

Automate document processing with intelligent AI-powered pipelines.

Availability: Pro, Team, Enterprise tiers

What are Document Pipelines?

Document Pipelines are automated workflows that process documents when they’re uploaded or on a schedule. Pipelines chain together AI operations to classify, extract data, route documents, and notify your team automatically.

Think of pipelines as: Automated assistants that handle repetitive document tasks 24/7.

Pipeline Components

Triggers

What starts the pipeline:

Trigger Description Example
On Upload Runs when documents are uploaded Process invoices immediately
Scheduled Runs on a schedule Daily compliance scan
Manual Run on-demand One-time batch processing
Folder-Based Upload to specific folder Finance folder triggers invoice pipeline

Actions

What the pipeline does:

Action Description Credits
Classify Identify document type 0.5
Extract Pull key data (amounts, dates, names) 1.0
Summarize Generate document summary 1.5
Tag Apply intelligent tags 0.5
Route Move to appropriate folder 0
Notify Alert team members 0
Export Generate report 3.0

Conditions

Add logic to your pipeline:

  • If document type is Invoice → Extract financial data
  • If amount > $10,000 → Flag for review
  • If contains PII → Apply security tag

Common Pipeline Templates

Invoice Processing Pipeline 

Trigger: On Upload to /Invoices
1. Classify document (confirm it's an invoice)
2. Extract: vendor, amount, due date, invoice number
3. Route to /Invoices/{vendor}/{year}
4. If amount > $5000: Notify finance@company.com
5. Tag with: invoice, {vendor}, {year}

Contract Review Pipeline 

Trigger: On Upload to /Contracts
1. Classify contract type (NDA, MSA, SOW, etc.)
2. Extract: parties, effective date, term, value
3. Check for: auto-renewal, non-compete, liability caps
4. Tag with extracted metadata
5. If missing signatures: Notify legal@company.com

Compliance Check Pipeline 

Trigger: Scheduled (daily at 2am)
1. Scan all documents uploaded in last 24h
2. Check for: SSN, credit cards, medical info
3. If PII found: Apply "Contains PII" tag
4. Generate daily compliance report
5. Notify compliance@company.com

Creating a Pipeline

Via the UI

  1. Navigate to Actions > Pipelines
  2. Click “New Pipeline”
  3. Select a template or start from scratch
  4. Configure trigger, actions, and conditions
  5. Test with a sample document
  6. Activate the pipeline

Via API 

curl -X POST https://api.archivus.app/api/v1/pipelines \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Invoice Processing",
    "trigger": {
      "type": "upload",
      "folder_id": "folder_invoices"
    },
    "steps": [
      {
        "action": "classify",
        "config": {"expected_type": "invoice"}
      },
      {
        "action": "extract",
        "config": {"fields": ["vendor", "amount", "due_date"]}
      },
      {
        "action": "route",
        "config": {"pattern": "/Invoices/{vendor}/{year}"}
      }
    ],
    "active": true
  }'

Pipeline Execution

Processing Flow

  1. Document uploaded → Trigger fires
  2. Pipeline queued → Added to background processing queue
  3. Steps execute → Each action runs in sequence
  4. Conditions evaluated → Branch based on results
  5. Complete → Results logged, notifications sent

Monitoring

View pipeline activity in Actions > Pipelines > Activity Log:

  • Execution history with timestamps
  • Success/failure status for each step
  • Credit usage per execution
  • Error details for debugging

Error Handling

Pipelines handle errors gracefully:

  • Retry - Transient failures retry 3 times
  • Skip - Failed optional steps can be skipped
  • Stop - Critical failures stop the pipeline
  • Notify - Errors can trigger notifications

Credit Usage

Pipelines deduct AI credits based on actions used:

Action Credits per Document
Classify 0.5
Extract 1.0
Summarize 1.5
Tag 0.5
Route 0 (free)
Notify 0 (free)
Export Report 3.0

Example: Invoice Processing pipeline with classify + extract + tag = 2.0 credits per document

Best Practices

Pipeline Design

  1. Start simple - Begin with 2-3 actions, add complexity later
  2. Test thoroughly - Use sample documents before activating
  3. Use conditions - Don’t process everything the same way
  4. Monitor costs - Track credit usage and optimize

Folder Organization

Organize folders for efficient pipeline triggers:

/Inbox
  /Invoices      → Invoice pipeline
  /Contracts     → Contract pipeline
  /HR            → HR document pipeline
/Processed       → Pipeline output destination

Notification Strategy

  • High-value items - Notify immediately (large invoices, contracts)
  • Daily digest - Batch low-priority notifications
  • Errors only - Alert on failures, not successes

API Reference

Create Pipeline

POST /api/v1/pipelines

List Pipelines

GET /api/v1/pipelines

Get Pipeline

GET /api/v1/pipelines/:id

Update Pipeline

PATCH /api/v1/pipelines/:id

Delete Pipeline

DELETE /api/v1/pipelines/:id

Execute Pipeline Manually

POST /api/v1/pipelines/:id/execute

Get Execution History

GET /api/v1/pipelines/:id/executions

Get Pipeline Templates

GET /api/v1/pipelines/templates

Limitations

Tier Active Pipelines Executions/Month
Pro 5 500
Team 20 5,000
Enterprise Unlimited Unlimited

FAQ

Q: Can pipelines process existing documents? A: Yes! Use the Manual trigger or “Reprocess” button to run pipelines on existing documents.

Q: What happens if I run out of credits mid-pipeline? A: The pipeline pauses. You can resume after adding credits or the next billing cycle.

Q: Can I chain multiple pipelines? A: Not directly, but a pipeline’s “Route” action can move documents to a folder with another pipeline trigger.

Q: Are pipeline executions logged for audit? A: Yes. Team and Enterprise tiers have full audit logging of all pipeline activity.

Q: Can I duplicate a pipeline? A: Yes. Use “Duplicate” in the pipeline menu to create a copy for modification.

Need help setting up a pipeline? Ask Archie: “Help me create an invoice processing pipeline”