How zipstack/unstract works
Unstract turns unstructured documents into structured JSON using natural-language extraction prompts. The user-facing workflow starts in the web app, where teams configure Prompt Studio profiles and deploy extraction through an API or ETL pipeline; workers process files and return results. The README also documents MCP and n8n integrations, and the tree shows corresponding backend components. Several backend and worker links below are documented capabilities rather than inspected call paths, so those edges cite the README; sampled implementation details are limited.
Main entry points: pipeline_v2 ETL pipeline · workers Processing workers · useMainAppRoutes.jsx Web application · prompt_studio Prompt Studio
Components 18
User experience
- Web application
frontend/src/routes/useMainAppRoutes.jsx - Pipeline console
frontend/src/components/pipelines-or-deployments/pipelines/Pipelines.jsx
Extraction and processing
- Prompt Studio
backend/prompt_studio - Output management
backend/prompt_studio/prompt_studio_output_manager_v2/output_manager_helper.py - File connectors
unstract/connectors/src/unstract/connectors/filesystems - File filtering
workers/shared/processing/filter_pipeline.py - Processing workers
workers - LLM providers
unstract/connectors - Structured JSON
Deployment and integration
- API deployment
backend/api_v2 - ETL pipeline
backend/pipeline_v2 - MCP server
backend/mcp_server - n8n integration
Platform services
- Accounts and auth
backend/account_v2 - Account records
backend/account_v2/models.py - Pipeline records
backend/pipeline_v2/models.py - Platform services
backend/backend
Other components
- Team user
Connections 23
- Team user → Web application: configures
README.md - Web application → Prompt Studio: defines prompts
README.md - Web application → Pipeline console: manages deployments
frontend/src/routes/useMainAppRoutes.jsx - Pipeline console → ETL pipeline: manages pipelines
- Prompt Studio → Output management: formats outputs
backend/prompt_studio/prompt_studio_output_manager_v2/output_manager_helper.py - File connectors → Processing workers: supplies files
unstract/connectors/src/unstract/connectors/filesystems/google_cloud_storage/google_cloud_storage.py - Processing workers → File filtering: filters batches
workers/shared/processing/filter_pipeline.py - Processing workers → Prompt Studio: runs extraction
backend/prompt_studio/prompt_studio_output_manager_v2/output_manager_helper.py - Prompt Studio → LLM providers: queries models
- Output management → Structured JSON: produces JSON
README.md - Team user → API deployment: submits documents
README.md - API deployment → Processing workers: dispatches processing
- Processing workers → Structured JSON: returns results
README.md - Team user → ETL pipeline: configures ETL
README.md - ETL pipeline → File connectors: pulls documents
README.md - ETL pipeline → Processing workers: processes files
backend/pipeline_v2/pipeline_processor.py - ETL pipeline → File connectors: loads results
README.md - Team user → MCP server: connects agent
README.md - Team user → n8n integration: uses automation
README.md - Accounts and auth → Account records: persists accounts
backend/account_v2/models.py - ETL pipeline → Pipeline records: tracks pipeline
backend/pipeline_v2/pipeline_processor.py - Web application → Accounts and auth: authenticates
- API deployment → Platform services: uses services