TurfAI Platform — Feature Reference

Synced from the TurfAI source on 2026-06-21.

Last Updated: February 25, 2026

---

1. Visual Workflow Builder

Design automation workflows on a drag-and-drop canvas powered by React Flow.

• Drag-and-drop node canvas with auto-layout
• Searchable node palette with 13+ task types
• Dynamic configuration panel — forms adapt to each task type
• Real-time workflow validation before execution
• Execution timeline — step-by-step view of past runs with inputs/outputs per node
• Workflow status lifecycle: draft, active, archived
• Ownership model: system workflows (shared) and user workflows (private)
• Clone system workflows to create personal copies via instantiation

---

2. Triggers

Four ways to start a workflow.

Webhook Triggers

• Accept POST requests with JSON or multipart/form-data
• File upload support with configurable max size and MIME type filters
• Auto-generated curl sample code from webhook schema
• Secret key authentication with timing-safe comparison
• Per-webhook event type tagging
• Trigger count and last-triggered timestamp

Scheduled Triggers

• Cron expression scheduling (daily, weekly, custom)
• Timezone-aware execution
• Manual trigger on demand
• Pause / resume toggle
• Skip-if-running guard to prevent overlapping executions
• Next-run prediction display

Event Sources (Pub/Sub)

• Google Drive folder watching — trigger on file upload/change
• Event subscriptions — bind workflows to event sources with filters
• Event type filtering and custom criteria
• Event logs with processing status
• Automatic watch renewal for Drive API
• Event volume statistics

Manual Execution

• Run any workflow with custom JSON inputs from the UI or API

---

3. AI & Document Processing Tasks

Document Classification

• Categorize documents by type, sentiment, or topic
• Multi-classification with confidence scores
• Custom category definitions

Data Extraction

• Extract structured fields (names, dates, amounts) from documents
• Prompt-based extraction using the prompt library
• Define output JSON schema for consistent results
• Confidence scoring per field

LLM Text Generation

• Free-form AI text generation with any prompt
• Model selection: Vertex AI, OpenAI, Anthropic
• Reference saved prompts or write inline
• Structured JSON or plain text output

Summarization

• Generate executive summaries of documents
• Summary types: brief, detailed, bullet-point
• Configurable maximum word count

Combined Classification & Extraction

• Classify then extract in a single task node
• Conditional extraction based on classification result

OCR Text Extraction

• Extract text from scanned PDFs and images
• Automatic language detection
• Integrated into ingestion workflows

---

4. Integration Tasks

Google Drive

• Fetch files by ID or search by name
• Import Drive files into Document Management System
• Automatic upload to Google Cloud Storage
• Metadata preservation (name, type, size)
• Uses user's OAuth permissions

Email (SMTP)

• Send emails with HTML body templates
• Dynamic {{variable}} substitution in to, subject, and body
• Multiple recipients (to, CC, BCC)
• Custom sender address
• Delivery status in execution logs

Gmail

• Read messages from connected Gmail account
• Reply to specific messages with thread preservation
• Label filtering
• Attachment access

REST API

• HTTP methods: GET, POST, PUT, PATCH, DELETE
• Template variables in URL, headers, and body
• Custom headers (Authorization, Content-Type, etc.)
• JSONPath output mapping to extract specific fields as workflow variables
• Configurable timeout

Document Generation

• Create DOCX files from templates with data merge
• Variable substitution from workflow context
• Generated files available for download or email attachment

Web Scraping (Metadata Extraction)

• Fetch and extract content from web pages by URL
• HTML-to-markdown conversion
• Open Graph metadata extraction
• Used in chatbot URL ingestion workflow

---

5. Control Flow

Decision Nodes

• Conditional branching with true/false paths
• Operators: ==, !=, >, <, >=, <=, in, contains, matches (regex)
• JSONPath references to previous task outputs
• Custom edge labels for true/false paths
• Paths can converge downstream

Wait / Poll Tasks

• Poll an endpoint until a condition is met
• Configurable poll interval and timeout
• Success/failure value matching
• Elapsed time tracking

For-Each Loops

• Iterate over arrays with per-item task execution
• Accumulator to collect results across iterations
• Current item accessible via template variables

---

6. Variable System

Input/Output Mapping

• {{variable_name}} syntax throughout all task configs
• Input mapping — resolve variables from workflow context before execution
• Output mapping — extract fields from task results using JSONPath
• Webhook payloads auto-generate variables from form fields
• File uploads produce: {field}_document_id, {field}_file_url, {field}_mime_type

Built-in Context Variables

• {{user_id}} — workflow owner
• {{execution_id}} — current execution
• {{DMS_URL}} — backend base URL
• All webhook form fields available by name

---

7. Knowledge Base & RAG

Document Indexing

• Enable RAG on any document to index for semantic search
• Async processing with status tracking (queued → processing → completed)
• Force reprocess option for re-indexing
• Chunk count and embedding model metadata
• Collection-scoped indexing

Semantic Search (RAG Query)

• Natural language queries across indexed documents
• Top-K results with similarity scoring
• AI-synthesized answers with source citations
• Reranking option for improved relevance
• Configurable similarity threshold

Conversation Sessions

• Persistent multi-turn chat with session IDs
• Session listing with pagination
• Session history loading
• Session deletion
• Title updates for organization

Document Collections

• Group documents into named collections
• Add/remove documents from collections
• Collection metadata (name, description)

---

8. Conversational Chatbot

Chat Interface (Authenticated)

• Full-page chat at /chat with three-panel layout
• Left sidebar: session list with create/delete
• Center: message thread with ChatMessage component
• Right panel: source documents for current response
• Suggestion prompts for empty state
• Optional chatbot scope via ?chatbot=<id> query parameter

Sidebar Chat

• Floating button (bottom-right) opens slide-out chat sheet
• Same useChatSession hook as full page
• Available on every authenticated page

Chatbot Management

• Create named chatbots with unique slugs
• Per-chatbot API key (cb_ prefix, 48 hex chars)
• Branding: primary color, logo URL
• Welcome message configuration
• Allowed origins for CORS (empty = allow all)
• Rate limiting: configurable requests per minute per API key
• Active/inactive toggle
• Query count and last query timestamp
• Settings page with 5 tabs: General, Branding, Security, Embed, Knowledge Base

Public Chat API

• GET /api/chatbots/:slug/config — public, returns branding and welcome message
• POST /api/chatbots/:slug/chat — authenticated via X-Chatbot-Key header
• Queries execute against chatbot owner's indexed documents
• In-memory rate limiting per API key
• CORS preflight handling for cross-origin requests

Embeddable Widget

• Standalone JavaScript widget for any website
• Shadow DOM for CSS isolation from host page
• Floating button opens 380x560px chat overlay
• Session persistence in localStorage keyed by chatbot slug
• Loads branding from public config endpoint
• Source citation badges on responses
• "Powered by TurfAI" footer
• Build: npx vite build --mode widget → dist/widget.js
• Embed: <script src=".../widget.js" data-chatbot="slug" data-api="https://..."></script>
• Programmatic: window.TurfAIChat.init({ chatbotSlug, apiUrl })

KB Assembly View

• Table of all knowledge base sources (documents + URLs)
• Source type badges: "Document" vs "Web URL"
• Status indicators: Indexed, Processing, Pending, Failed
• Reindex action per source
• Summary bar: total sources, document count, URL count, indexed count
• Embedded in chatbot settings Knowledge Base tab

---

9. Workflow Templates

Ten pre-built templates covering common use cases.

• Template preview shows workflow structure before instantiation
• Filter by category, difficulty, tags
• Chatbot templates auto-create a Chatbot instance on instantiation

---

10. Prompt Management

Prompt Library

• Create, edit, delete reusable AI prompts
• System prompts (shared) vs user prompts (private)
• Title, description, category, tags metadata
• Permission-based visibility
• Search and filter by name/category/tags
• Generate RTIO structure from document + classification result

Prompt Lab (Interactive Testing)

• Test extraction prompts against real documents
• Output schema builder for structured JSON results
• Switch between Vertex AI, OpenAI, Anthropic models
• AI-powered prompt enhancement (RTIO polishing)
• Save tested prompts directly to the library

---

11. Document Management

Upload & Storage

• Direct file upload with title, description, tags
• Google Cloud Storage backend
• Format support: PDF, DOCX, TXT, images, Excel
• Auto-conversion: Excel → Markdown
• OCR for scanned documents and images

Document Operations

• List with pagination and filters
• View document details with metadata
• Download original file
• Edit metadata (title, description, tags)
• Delete documents
• Signed URL generation for temporary access

RAG Integration

• Enable/disable RAG indexing per document
• RAG status polling (queued, processing, completed, failed)
• Chunk count and embedding model info
• Force reprocess

---

12. Analytics & Monitoring

Dashboard

• Token usage with USD cost estimate
• Jobs executed count
• Storage usage vs quota (progress bar)
• Recent activity timeline
• System health check

Workflow Analytics

• Per-workflow execution count and success rate
• Average execution time
• Recent errors with details
• Custom time periods: 24h, 7d, 30d

Execution Tracking

• Real-time execution status
• Step-by-step timeline with inputs/outputs
• Node-level status: active, skipped, failed, completed
• Error details and stack traces
• Resume paused workflows with correlation tokens

---

13. Authentication & Integrations

User Authentication

• Email/password login
• JWT token management with auto-refresh
• Protected routes with ProtectedRoute wrapper
• 401 interceptor with auto-redirect to login

Google OAuth 2.0

• Connect Google account for Drive and Gmail access
• OAuth authorization URL generation
• Connection status check
• Revoke access (disconnect)
• Automatic token refresh

Integration Catalog

• Browse available integrations
• Connect user-specific integrations with credentials
• CRUD operations on user integrations

---

14. User Interface

Navigation

• Collapsible sidebar with grouped menu sections
• Groups: Overview, Documents, AI Processing, Prompts, Knowledge Base, Automation, System
• Breadcrumb navigation with route name mapping
• Active route highlighting

UI Component Library

• 49 shadcn/ui components (Radix UI primitives + Tailwind CSS)
• Consistent design language across all pages
• Toast notifications (Sonner) for feedback
• Responsive layout with sidebar provider

Feedback System

• In-app feedback dialog
• Accessible from sidebar footer

---

15. Technology Stack

Frontend

• React 18.3.1 + TypeScript 5.6
• Vite 6.0 (dev server + build)
• React Flow 11.11.4 (workflow canvas)
• TanStack Query 5.83.0 (server state)
• Tailwind CSS 3.4 + shadcn/ui
• Axios 1.7 (HTTP client)
• React Router DOM 6.30 (routing)
• Sonner (toast notifications)

Backend

• Strapi 5.10.3 (Node.js CMS)
• PostgreSQL (database)
• Redis (job queues)
• Python FastAPI (Job Router :8000, Processors :8001, LLM Service :9090, RAG Query :8003)
• Google Cloud Storage (file storage)
• Vector DB (RAG embeddings)
• Nodemailer (SMTP email)
• node-cron (scheduling)

External Services

• Google Cloud: Drive API, Gmail API, Storage, Vision, Vertex AI
• LLM Providers: Vertex AI, OpenAI, Anthropic

---

16. API Surface

22 API domain objects in the frontend client (src/lib/api.ts):