İçeriğe geç
YENİPazar yeri ve trig8n Agent: geliştirdiğinizi satın; planlar ve değişiklikler siz görmeden önce test edilsin. Yeniliklere bak

Kütüphane / RAG ve Bilgi Bankaları

RAG ve Bilgi Bankaları

Alıntılı OpenAI dosya erişimi RAG sohbeti

Bir sohbet mesajı geldiğinde, bir OpenAI asistanı yüklenen dosyalardan bilgi getirir ve yanıt kaynak belgelere atıflar içerir.

Ne yapar

# Make OpenAI Citation for File Retrieval RAG ## Overview Automated workflow: Make OpenAI Citation for File Retrieval RAG. This workflow integrates 11 different services: stickyNote, httpRequest, markdown, code, splitOut. It contains 24 nodes and follows best practices for error handling and security. ## Workflow Details - Total Nodes: 24 - Node Types: 11 - Error Handling: ✅ Implemented - Security: ✅ Hardened (no sensitive data) - Documentation: ✅ Complete ## Node Breakdown - Aggregate: aggregate - Window Buffer Memory: memoryBufferWindow - Sticky Note4: stickyNote - Sticky Note: stickyNote - Create a simple Trigger to have the Chat button within N8N: chatTrigger - OpenAI Assistant with Vector Store: openAi - Sticky Note1: stickyNote - Get ALL Thread Content: httpRequest - Sticky Note2: stickyNote - Split all message iterations from a thread: splitOut - ... and 14 more nodes ## Usage Instructions 1. Configure Credentials: Set up all required API keys and credentials 2. Update Variables: Replace any placeholder values with actual data 3. Test Workflow: Run in test mode to verify functionality 4. Deploy: Activate the workflow for production use ## Security Notes - All sensitive data has been removed or replaced with placeholders - Error handling is implemented for reliability - Follow security best practices when configuring credentials ## Troubleshooting - Check error logs if workflow fails - Verify all credentials are properly configured - Ensure all required services are accessible

Adımlar

  1. Manual Trigger
  2. Aggregate
  3. Window Buffer Memory
  4. Create a simple Trigger to have the Chat button within N8N
  5. OpenAI Assistant with Vector Store
  6. Get ALL Thread Content
  7. Split all message iterations from a thread
  8. Split all content from a single message
  9. Split all citations from a single message
  10. Retrieve file name from a file ID
  11. Regularize output
  12. Optional Markdown to HTML
  13. Finnaly format the output
  14. Error Handler

Toplam 33 adım.

Benzer otomasyonlar