Add a phase transition to history.

Get workflow state for a pipeline.

Get the OpenRouter LLM instance configured for Kimi-k2.

Broadcast workflow state update to connected clients.

Safely send message to websocket.

Search for URLs using ScrapeGraphAI SearchScraper.

Extract domain from URL.

Broadcast message to all connections for a pipeline.

Broadcast a message to all connections for a pipeline.

Get or create enhanced workflow manager instance.

Get or create workflow manager instance.

Process a message through the appropriate agent.

Stream execution updates to connected clients.

Classify the type of data being scraped based on schema.

Get domain-specific knowledge.

Get fallback schema when parsing fails.

Get progress percentage based on current phase.

Get workflow state for a pipeline.

Process a user message and return the response.

Add a message to conversation history.

Add a message to conversation history.

Generate optimized scraping code without Pydantic schemas.

Use AI to generate relevant schema field suggestions based on the search topic and URLs.

Get recent conversation history.

Get conversation history for a pipeline.

Get existing context or create new one.

Update stored context.

Track user modifications.

Process an approval request.

Broadcast message to all connections except sender.

Check if transition to target phase is allowed.

Accept and store a new WebSocket connection.

Create a new workflow state.

Remove a WebSocket connection.

Execute the generated code (placeholder).

Process an approval request.

Process a user message through the agent.

Send a message to a specific WebSocket connection.

Execute SmartScraper with retry logic. Args: website_url: URL to scrape user_prompt: Natural languag

Suggest optimizations based on learned patterns.

Adapt a template pipeline to current context.

Use AI to analyze user intent and determine action needed.

Analyze user intent with full conversation context.

Use AI to analyze and filter URLs based on relevance to the user's prompt.

Apply suggested optimizations to context.

Auto-release edit lock after timeout.

Auto-save draft after delay.

Broadcast workflow state update to connected clients.

Build the LangGraph state machine.

Build the tool-based agent graph.

Build the system prompt for the agent.

Create the LangChain agent with tools.

Find similar pipelines from repository.

Format recent conversation history for context.

Generate simple Python code for scraping.

Generate SmartScraper code (reused from previous implementation).

Generate correct SmartScraper API code based on schema and URLs.

Get the next logical phase in the workflow.

Get all available tools.

Handle messages with direct URLs.

Handle general conversation about schema, code generation, etc.

Handle URL search requests using AI-determined search query.

Load frequently used patterns into memory cache.

Parse the AI response for actions to update context.

Parse search results from various response formats.

Process message based on detected intent.

Process a batch of learning data.

Process message using the original state-based agent.

Process message using the tool-based agent.

Record error patterns for learning.

Schedule auto-save of draft.

Fallback search using web scraping.

Simple fallback check for URL search intent.

Suggest next actions based on current state.

Suggest schema fields based on URLs and description.

Update domain knowledge base.

Update pattern database with new learning.

Update workflow state from agent result.

Update workflow state from agent result.

Update workflow state from tool-based agent results.

Create an approval request.

Create a new workflow state.

Execute scraping for multiple URLs concurrently.

Finish streaming and send complete response.

Get a summary of the workflow state.