> ## Documentation Index > Fetch the complete documentation index at: https://docs.productflo.io/llms.txt > Use this file to discover all available pages before exploring further. # AI Communication > Using WebSockets for AI-powered features # AI Communication via WebSockets ProductFlo provides powerful AI capabilities through its WebSocket API, enabling real-time interaction with the Haitch AI assistant and various AI-powered generation features. This page details how to integrate with these AI capabilities. ## Overview The WebSocket API supports several AI-powered features: Intelligent assistant for product development, engineering, and documentation AI-powered brainstorming and product ideation Create technical documentation, specifications, and reports CAD generation, DFM analysis, and engineering assistance ## Message Structure AI communication uses the standard WebSocket message structure with specific code types: ```json theme={null} { "code": "ai_code_type", // e.g., "haitch_chat", "idea", "documentation", "engineering" "user_id": "user-123", "room_id": "product-456", "data": { // AI-specific message data } } ``` ## Haitch AI Assistant The Haitch AI assistant provides intelligent assistance for product development questions and tasks. ### Sending a Haitch Chat Message ```javascript theme={null} socket.send(JSON.stringify({ "code": "haitch_chat", "user_id": "user-123", "room_id": "product-456", "data": { "query": "What materials would work best for a waterproof enclosure?", "doc_type": "consultation", // Optional: Specifies type of response needed "source_urls": ["https://example.com/materials-guide.pdf"] // Optional: Reference URLs } })); ``` ### Receiving a Haitch Response ```json theme={null} { "status": "success", "code": "chat_response", "user_id": "user-123", "room_id": "product-456", "timestamp": "2024-05-19T10:30:45Z", "data": { "message": "For a waterproof enclosure, I recommend considering these materials:\n\n1. ABS plastic with IP67 rating\n2. Polycarbonate with silicone gaskets\n3. Aluminum with O-ring seals\n\nThe choice depends on your specific requirements for impact resistance, temperature range, and manufacturing methods. Would you like me to elaborate on any of these options?", "sender": "haitch", "message_id": "msg-789", "sources": ["https://example.com/materials-guide.pdf"], "metadata": { "confidence": 0.92, "processing_time": 1.25, "doc_type": "consultation" } } } ``` ### Streaming Responses For longer AI responses, you'll receive the response in chunks as they're generated (streaming is enabled by default): ```javascript theme={null} socket.send(JSON.stringify({ "code": "haitch_chat", "user_id": "user-123", "room_id": "product-456", "data": { "query": "Generate a comprehensive analysis of our thermostat design", "doc_type": "analysis", "request_id": "req-123" // Optional: Track this specific request } })); ``` You'll receive multiple stream chunks with pending status: ```json theme={null} { "status": "pending", "code": "chat_response", "user_id": "user-123", "room_id": "product-456", "timestamp": "2024-05-19T10:30:45Z", "data": { "message": "Here's my analysis of the thermostat design:", "is_complete": false, "sender": "haitch", "message_id": "stream-chunk-1" } } ``` And when complete, you'll receive a final message with success status: ```json theme={null} { "status": "success", "code": "chat_response", "user_id": "user-123", "room_id": "product-456", "timestamp": "2024-05-19T10:30:47Z", "data": { "message": "Here's my analysis of the thermostat design: [complete response]", "sender": "haitch", "message_id": "msg-790", "is_complete": true, "sources": [], "metadata": { "processing_time": 2.5, "doc_type": "analysis" } } } ``` ### Client Implementation for Streaming ```javascript theme={null} let fullResponse = ''; let messageMap = new Map(); // Store chunks by message ID socket.onmessage = (event) => { const data = JSON.parse(event.data); // Only process chat_response messages if (data.code === "chat_response") { // For pending status (streaming chunks) if (data.status === "pending") { // Accumulate the response text fullResponse += data.data.message || ''; // Update UI to show progress updateProgressiveUI(fullResponse); } // For success status (final message) if (data.status === "success" && data.data.is_complete) { // Final message contains the complete response finalResponse = data.data.message; // Update UI with complete response finalizeUI(finalResponse); // Reset for next conversation fullResponse = ''; } } }; ``` ## Document Generation Generate professional documentation from product data and requirements. ### Requesting Document Generation ```javascript theme={null} socket.send(JSON.stringify({ "code": "documentation", "user_id": "user-123", "room_id": "product-456", "data": { "query": "Generate a technical specification for our thermostat", "doc_type": "technical_specification", // Type of document required "source_urls": ["https://example.com/thermostat_requirements.pdf"] // Optional reference URLs } })); ``` ### Receiving Generated Document ```json theme={null} { "status": "success", "code": "documentation_response", "user_id": "user-123", "room_id": "product-456", "timestamp": "2024-05-19T10:31:45Z", "data": { "message": "Technical specification generated successfully", "doc_type": "technical_specification", "files": [ { "file_id": "file-901", "file_name": "thermostat_technical_specification.pdf", "file_url": "https://api.productflo.io/files/thermostat_technical_specification.pdf", "file_type": "application/pdf", "file_size": 245789 } ], "content": "# Technical Specification: Smart Thermostat\n\n## Overview\n...", "sources": ["https://example.com/thermostat_requirements.pdf"], "metadata": { "word_count": 2450, "page_count": 12, "processing_time": 3.2, "sections": ["overview", "specifications", "requirements", "testing"] } } } ``` ### Document Types The system supports various document types: * `technical_specification`: Detailed product specifications * `bom`: Bill of Materials * `dfm_analysis`: Design for Manufacturing analysis * `user_manual`: End-user documentation * `datasheet`: Technical datasheet * `requirements`: Requirements document * `test_plan`: Testing procedures and criteria * `engineering_drawing`: Technical drawings * `compliance_documentation`: Regulatory compliance docs ## Idea Generation Generate product ideas, feature suggestions, and creative solutions. ### Requesting Idea Generation ```javascript theme={null} socket.send(JSON.stringify({ "code": "idea", "user_id": "user-123", "room_id": "product-456", "data": { "query": "Generate innovative ideas for a smart thermostat", "source_urls": ["https://example.com/smart_home_trends.pdf"], "doc_type": "idea_generation" // Optional: Specifies the type of ideation } })); ``` ### Receiving Generated Ideas ```json theme={null} { "status": "success", "code": "idea_response", "user_id": "user-123", "room_id": "product-456", "timestamp": "2024-05-19T10:32:45Z", "data": { "message": "Here are innovative smart thermostat feature ideas:", "content": { "product_name": "Smart EcoTherm", "product_description": "An AI-powered smart thermostat focused on energy efficiency and user comfort", "features": [ { "title": "AI-Powered Usage Learning", "description": "Machine learning algorithm that adapts to user habits and optimizes heating/cooling schedules automatically", "benefits": ["Energy savings", "Improved comfort", "Reduced user configuration"], "implementation_complexity": "medium" }, { "title": "Multi-Zone Sensing", "description": "Low-cost satellite temperature sensors that connect to the main thermostat for more accurate room-by-room temperature control", "benefits": ["Better comfort", "Increased efficiency", "Premium feature"], "implementation_complexity": "high" } ], "market_analysis": { "opportunity": "Strong demand for energy-efficient smart home devices", "competitive_landscape": "Current offerings lack advanced learning capabilities" } }, "sources": ["https://example.com/smart_home_trends.pdf"], "metadata": { "processing_time": 2.8, "feature_count": 2, "doc_type": "idea_generation" } } } ``` ## Engineering AI Features Generate CAD models, perform engineering analysis, and create technical designs. ### Requesting Engineering Analysis ```javascript theme={null} socket.send(JSON.stringify({ "code": "engineering", "user_id": "user-123", "room_id": "product-456", "data": { "query": "Generate a CAD model for a wall-mount bracket for our thermostat with dimensions 100x80x20mm, made of ABS plastic", "doc_type": "cad_generation", // Type of engineering request "source_urls": ["https://example.com/thermostat_dimensions.pdf"] // Optional reference URLs } })); ``` ### Receiving Engineering Response ```json theme={null} { "status": "success", "code": "engineering_response", "user_id": "user-123", "room_id": "product-456", "timestamp": "2024-05-19T10:35:45Z", "data": { "message": "CAD model generated successfully", "content": { "description": "Wall-mount bracket for thermostat", "specifications": { "dimensions": "100x80x20mm", "material": "ABS plastic", "mounting_type": "wall_mount", "estimated_weight": "42g" }, "manufacturing_notes": "Suitable for injection molding with standard draft angles" }, "files": [ { "file_id": "file-902", "file_name": "thermostat_bracket.step", "file_url": "https://api.productflo.io/files/thermostat_bracket.step", "file_type": "application/step", "file_size": 1245789 }, { "file_id": "file-903", "file_name": "thermostat_bracket.stl", "file_url": "https://api.productflo.io/files/thermostat_bracket.stl", "file_type": "application/stl", "file_size": 2345678 }, { "file_id": "file-904", "file_name": "thermostat_bracket_preview.png", "file_url": "https://api.productflo.io/files/thermostat_bracket_preview.png", "file_type": "image/png", "file_size": 245789 } ], "sources": ["https://example.com/thermostat_dimensions.pdf"], "metadata": { "processing_time": 5.2, "doc_type": "cad_generation", "file_count": 3 } } } ``` ### Engineering Request Types The system supports various engineering request types: * `cad_generation`: Generate 3D CAD models * `dfm_analysis`: Design for Manufacturing analysis * `tolerance_analysis`: Component tolerance stack-up analysis * `structural_analysis`: Basic structural integrity checks * `thermal_analysis`: Thermal performance simulation * `electrical_design`: Simple electrical system design * `pcb_layout`: PCB layout suggestions * `manufacturing_process`: Manufacturing process recommendations ## Handling Errors AI operations can sometimes fail due to various reasons. Error responses follow this format: ```json theme={null} { "status": "error", "code": "system", "user_id": "user-123", "room_id": "product-456", "timestamp": "2024-05-19T10:35:45Z", "data": { "message": "Failed to generate CAD model", "details": { "error_code": "CAD_GENERATION_ERROR", "message": "Invalid specifications provided: dimensions must include all three axes" } } } ``` Common AI-specific error codes: * `HAITCH_CHAT_ERROR`: Error in AI assistant processing * `DOCUMENT_GENERATION_ERROR`: Error in document generation * `IDEA_GENERATION_ERROR`: Error in idea generation * `CAD_GENERATION_ERROR`: Error in CAD model generation * `ENGINEERING_ANALYSIS_ERROR`: Error in engineering analysis * `CONTENT_POLICY_VIOLATION`: Request violates content policies * `CONTEXT_LIMIT_EXCEEDED`: Too much context provided ## Canceling Operations Long-running AI operations can be canceled using a cancel message: ```javascript theme={null} socket.send(JSON.stringify({ "code": "system", "user_id": "user-123", "room_id": "product-456", "data": { "message": "cancel", "request_id": "req-123" // The request ID to cancel } })); ``` You'll receive a confirmation: ```json theme={null} { "status": "success", "code": "system", "user_id": "user-123", "room_id": "product-456", "timestamp": "2024-05-19T10:36:45Z", "data": { "message": "Operation canceled", "request_id": "req-123" } } ``` ## Providing Context with Files To enhance AI results, you can attach files for additional context: 1. First, upload files using the `/files` REST API endpoint 2. Then reference the file IDs in your WebSocket message: ```javascript theme={null} socket.send(JSON.stringify({ "code": "haitch_chat", "user_id": "user-123", "room_id": "product-456", "data": { "message": "Analyze this PCB design and suggest improvements", "sender": "user-123", "attached_files": ["file-123", "file-456"], "metadata": { "priority_areas": ["power_efficiency", "thermal_management"] } } })); ``` ## Best Practices 1. **Use Request IDs**: Always include a unique `request_id` for tracking AI operations, especially long-running ones 2. **Enable Streaming**: For responses that might be lengthy, enable streaming for better user experience 3. **Provide Clear Context**: Include relevant product IDs and files to improve AI accuracy 4. **Handle Errors Gracefully**: Implement proper error handling on the client side 5. **Implement Cancellation**: Allow users to cancel operations that are taking too long 6. **Respect Rate Limits**: AI operations are subject to rate limiting - implement retry logic with backoff 7. **Validate Inputs**: Ensure all required fields are properly formatted before sending