Workflow Builder

Overview

The Workflow Builder is the central workspace where you design, test, and manage your automation flows. It combines a prompt-based interface with a visual block builder for flexibility and ease of use.

Section	Description	Key Actions
Left Sidebar Chat	Prompt-based input area where you can describe or update your workflow in natural language.	- Write a prompt to create a new workflow - Modify existing workflows via text commands
Center Canvas	Visual workspace for designing workflows using a drag-and-drop interface.	- Add, connect, and arrange blocks - Define the flow of data between steps
Top Navigation Bar	Provides quick access to essential controls and tabs.	- Switch between Editor and History - Save, Run Flow, or Publish workflows
Block Library Tabs	A categorized list of available blocks to build workflows.	- Integrations: File upload, Email, API ingestion - Transform: Data extraction, Python code, AI categorization - Learn: View examples or help articles
Run Flow Button	Executes the workflow for testing purposes.	- Test workflow logic - View results in History tab
Publish Button	Deploys the workflow to production for live usage.	- Make workflow accessible via API or automation triggers
Workflow Menu (⋯)	Access workflow management utilities.	- Copy Workflow ID - Duplicate, Download, or Upload workflows
History Tab	Displays run history for debugging and monitoring.	- View run logs, status, and timestamps
Version Control	Shows current workflow version and allows rollbacks.	- Manage versions

✅ Tip: Use the left sidebar for quick generation and the canvas for fine-tuning — combining both gives you speed and control.

Adding Workflow Blocks

Open the Block Library tab on the canvas to add building blocks.
Each block represents a specific task in your workflow. Common block categories include:

Workflow Blocks

Category	Block Name	Description	Example Use Case
Integrations → Input Blocks	File Upload	Ingest documents via Docsumo upload.	Users manually upload PDFs or images for extraction.
	Email Document Upload	Ingest attachments from incoming emails.	Automatically process invoices or receipts received over email.
	API Ingestion	Ingest documents via API calls.	Integrate with third-party tools or back-end systems to push documents automatically.
Transform → Extraction Blocks	Data Extraction	Extract structured data from documents based on a defined model.	Extract fields like “Invoice Number” and “Amount” from invoices.
Transform → Pre-Processing Blocks	Merge JSON	Combine multiple incoming data streams into a single JSON object.	Merge outputs from multiple extraction models before further processing.
	Python Code	Generate custom logic based on input data context.	Write code to clean, normalize, or enrich data before passing to the next block.
Transform → Post-Processing Blocks	Python Code	Run code-based logic after extraction to transform data further.	Format extracted values, add computed fields, or validate results.
	Categorize with AI	Automatically classify data columns into preset categories using AI.	Tag document types or categorize transaction records.
Transform → Classification Blocks	Document Type Classifier	Automatically classify incoming documents into specific types.	Detect if a document is an Invoice, PO, or Receipt.
Transform → Human in the Loop	Human in the Loop	Pause the workflow and allow a human to review or decide.	Require manual approval for low-confidence extractions.
Transform → Conditional Blocks	Router	Add IF–ELSE / conditional logic to branch the workflow.	Send invoices above ₹50,000 for approval; others directly to export.

✅ Tip: You can mix and match blocks from different categories to design complex, automated document workflows — from ingestion to validation to export.

Connecting Blocks

You can connect blocks by dragging an arrow from one block’s output to another block’s input.
Each connection defines data flow — i.e., how results from one step feed into the next.

✅ Tip: Always start your flow with an input block (e.g., file upload or API trigger) and end with an output block (e.g., export to API or database).

Configuring a Block

Click the edit option on any block in the canvas to open its configuration panel on the right side.

You can:

Define the input format (e.g., JSON, extracted fields, table rows)
Write custom Python code or prompt-based logic
Choose AI models for classification or categorization
Add validation rules and fallback conditions

Every block has input/output mapping — ensure field names and formats align between connected blocks.

Saving & Testing Your Workflow

Once your workflow is ready, click Run Flow to test it.
Review the output for accuracy and flow correctness.
Once verified, click Publish to activate it for production.

Your published workflow can now be triggered automatically via API, integrations, or human input.

Best Practices

Start simple — begin with 2–3 core blocks and expand gradually.
Use versioning to test major changes without affecting production.
Keep naming consistent for all blocks and fields for easy debugging.
Use “Human in the Loop” for exceptions or uncertain AI predictions.