Skip to content

USER_GUIDE.md

Latest commit

 

History

History
206 lines (159 loc) · 5.96 KB

USER_GUIDE.md

File metadata and controls

206 lines (159 loc) · 5.96 KB

👤 FlowTech-AI User Guide

Overview

FlowTech-AI is a comprehensive local AI stack designed for FPV, infrastructure, automation, and documentation tasks. This guide covers how to use the system effectively.

Getting Started

Accessing Services

Service URL Purpose Credentials
OpenWebUI http://localhost:8081 Main AI Interface Auto-configured
n8n http://localhost:5678 Workflow Orchestrator Check .env file
Langfuse http://localhost:3300 AI Observability Check .env file
SearxNG http://localhost:8082 Web Search Public access
Qdrant http://localhost:6333 Vector Database Internal only

Default Credentials

Check your .env file for current credentials:

cat .env | grep -E "(USER|PASSWORD|EMAIL)"

OpenWebUI - Main Interface

Basic Usage

  1. Access: http://localhost:8081
  2. Start Chat: Click "New Chat" to begin conversation
  3. Model Selection: Choose from available Ollama models
  4. Web Search: Enabled by default via SearxNG

Advanced Features

RAG (Retrieval Augmented Generation)

  • Document Upload: Upload PDFs, text files, images
  • Auto-Vectorization: Documents are automatically processed
  • Context Retrieval: AI uses uploaded documents for responses

Pipelines

  • OCR Pipeline: Extract text from images/PDFs
  • Ingestion Pipeline: Process and store documents
  • Custom Functions: Connect to n8n workflows

Configuration

Web Search Setup

  1. Go to Admin Panel > Settings > Web Search
  2. Configure: http://searxng:8080/search
  3. Enable "Use web search"

Model Management

  1. Admin Panel > Models
  2. Add Ollama models: http://192.168.0.2:11434
  3. Recommended models:
    • qwen2.5:7b (primary)
    • llama3.2:3b (fallback)

n8n - Workflow Orchestration

Main Workflows

N8N-OpenWebUI Integration

  1. Import Workflow: SRC/N8N-openwebui-workflow.json
  2. Import Function: SRC/function-N8N Pipe.json
  3. Configure Webhook: http://n8n:5678/webhook/invoke_n8n_agent

Available Workflows

  • PDF Processing: OCR and summarization
  • Infrastructure Analysis: Log analysis and recommendations
  • FPV Documentation: Technical sheet generation
  • Automation Proposals: Workflow suggestions

Using Workflows

  1. Access n8n interface
  2. Select workflow
  3. Configure input parameters
  4. Execute manually or via webhook
  5. Monitor execution in real-time

Langfuse - AI Observability

Key Features

  • Trace Monitoring: Track AI interactions
  • Performance Metrics: Latency and error rates
  • Cost Tracking: Token usage and costs
  • Feedback Collection: User ratings and improvements

Setting Up Projects

  1. Access Langfuse dashboard
  2. Create organization
  3. Create project
  4. Generate API keys
  5. Configure OpenWebUI integration

Monitoring

  • Dashboard: Real-time metrics
  • Traces: Individual AI interactions
  • Analytics: Usage patterns and performance
  • Alerts: Error notifications

Document Management

Document Types

  • docs_public: Public documents, auto-vectorized
  • docs_prive: Private documents, manual approval
  • convos_long: Long conversation storage

Upload Process

  1. Via OpenWebUI: Direct upload in chat
  2. Via n8n: Automated processing workflows
  3. Manual: Copy to AI_Data/docs_*/ directories

Search and Retrieval

  • Vector Search: Semantic document search
  • Keyword Search: Traditional text search
  • Context-Aware: AI uses relevant documents

Multi-Agent Workflows

Agent Types

  • Master Agent: Coordinates other agents
  • Specialist Agents: Domain-specific tasks
  • Document Agent: Handles document processing
  • Search Agent: Web search and information gathering

Workflow Examples

PDF Analysis Workflow

  1. Upload PDF via OpenWebUI
  2. OCR pipeline extracts text
  3. n8n processes content
  4. Specialist agent analyzes
  5. Results returned to UI

Infrastructure Monitoring

  1. Log files uploaded
  2. Analysis agent processes logs
  3. Recommendations generated
  4. Report exported to Nextcloud

Best Practices

Performance Optimization

  • Model Selection: Use appropriate model size
  • Context Management: Limit conversation history
  • Resource Monitoring: Watch RAM and CPU usage
  • Regular Cleanup: Archive old conversations

Security

  • Local Processing: All data stays local
  • Access Control: Use authentication features
  • Backup Strategy: Regular data backups
  • Network Security: Limit external access

Troubleshooting

Common Issues

  • Slow Responses: Check Ollama service
  • Memory Issues: Restart services
  • Search Failures: Verify SearxNG status
  • Workflow Errors: Check n8n logs

Getting Help

  1. Check service logs: docker compose logs [service]
  2. Verify service status: docker compose ps
  3. Review configuration files
  4. Check system resources

Advanced Usage

Custom Pipelines

  1. Define pipeline in OpenWebUI
  2. Connect to n8n webhooks
  3. Configure processing steps
  4. Test and deploy

API Integration

  • OpenWebUI API: REST API for automation
  • n8n Webhooks: Custom integrations
  • Langfuse API: Monitoring integration
  • Qdrant API: Vector operations

Export and Sharing

  • Markdown Export: Conversation summaries
  • PDF Reports: Generated documents
  • JSON Data: Structured data export
  • Nextcloud Sync: Automated backups

Tips and Tricks

Productivity

  • Use Templates: Save common prompts
  • Batch Processing: Process multiple documents
  • Scheduled Tasks: Automate regular workflows
  • Custom Functions: Create reusable components

Integration

  • Discord Bot: Connect via n8n
  • Telegram Integration: Mobile access
  • Webhook Endpoints: External system integration
  • Database Connections: Direct data access

Need Help? Check TROUBLESHOOTING.md or review SETUP.md for technical details.