LLM Knowledge Assistant: Voice and Text Responses with Pinecone, and Semantic Search

This workflow automates voice-enabled knowledge retrieval using RAG, replacing manual searches across docs that take 20-40 minutes per query for employees. The Voice Webhook receives questions, the Input Validator If node checks format, the Knowledge Agent LangChain Agent queries via OpenAI GPT-4o, the Query Classifier Code node routes by domain (projects/HR/sales), the Knowledge Base Tool VectorStore Pinecone retrieves from namespaces, the Embeddings Engine OpenAI node vectors with text-embedding-3-large, the Response Formatter Set node structures outputs, and Response Handler Respond to Webhook delivers JSON. Error Formatter provides fallbacks. It helps knowledge workers in enterprises (20-80 staff) answering 100+ queries weekly across silos, enabling instant, accurate responses without portals, streamlining operations and reducing frustration.\n\nThis workflow saves 6-10 hours weekly on 100 queries, improving accuracy by 85%. Use cases include employee self-service for HR policies, sales intel access for reps in tech firms. Suitable for mid-sized teams. Requires OpenAI ($0.03/1k tokens), Pinecone ($0.096/GB/month); n8n (free self-hosted or $20/month cloud). Scalable to 500 queries/day with Pro tiers.\n\nInstall n8n via n8n.io or cloud.n8n.io. Get OpenAI key at platform.openai.com (enable chat/embeddings). Create Pinecone index at console.pinecone.io (get key, set namespaces). Set OPENAI_API_KEY, PINECONE_API_KEY env vars. Import JSON; webhook POST 'devhubconnect-voice-rag-endpoint'. Configure Vector Store with index/namespaces.\n\nTest with webhook: {question: 'What's our vacation policy?'} . Verify JSON response. Errors: Empty query (error JSON), invalid key (500—regenerate). Activate webhook. Monitor dashboard weekly. Optimize namespaces; refresh keys quarterly.", "businessValue": "Saves 6-10 hours/week on 100 knowledge queries", "setupTime": "25-35 minutes", "difficulty": "Intermediate", "requirements": ["OpenAI API ($0.03/1k tokens)", "Pinecone API ($0.096/GB/month)", "n8n instance"], "useCase": "Voice-enabled RAG for enterprise knowledge access"