Privacy-first AI document analysis with Webex Teams integration. No external APIs, full control, runs on your local machine.
A complete, pilot-ready system that lets you:
- 📄 Upload technical documents (FAQs, reports, policies)
- 💬 Ask questions in Webex Teams (or Python/web interface)
- 🤖 Get AI-powered answers based on YOUR documents
- 🔒 Keep everything local - no data sent to OpenAI, Anthropic, etc.
- ⚡ Respond in 5-10 seconds with cited sources
Perfect for: IT teams, network engineers, technical documentation management, compliance-sensitive environments.
Choose your path:
Time: 8-10 hours
Path: Prerequisites Guide → Environment Setup → RAG System
Start here if you're new to command-line tools or containers.
Time: 4-6 hours
Path: Environment Setup → RAG System → Webex Integration
Start here if you're comfortable with basic Unix commands.
| Guide | Purpose | Time | Difficulty |
|---|---|---|---|
| Environment Setup | Install Podman, Ollama, ChromaDB, n8n | 45-60 min | ⭐⭐ |
| RAG System | Set up document loading and querying | 60-90 min | ⭐⭐ |
| Webex Integration | Add enterprise messaging bot | 30-45 min | ⭐⭐⭐ |
| Document | Use When |
|---|---|
| Prerequisites | New to terminal/containers/Python |
| Troubleshooting | Something isn't working |
| What's Next | System working, want to improve it |
| Documentation Standards | Contributing to docs |
┌──────────────────────────────────────┐
│ Your Local Machine │
│ │
│ ┌─────────────┐ ┌──────────────┐ │
│ │ Ollama │ │ Podman │ │
│ │ (Native) │ │ (Containers)│ │
│ │ │ │ │ │
│ │ • LLM │ │ • ChromaDB │ │
│ │ • Embeddings│ │ • n8n │ │
│ └─────────────┘ └──────────────┘ │
│ ↕ ↕ │
│ ┌────────────────────────────┐ │
│ │ Your Documents (Local) │ │
│ │ Network Assessments │ │
│ │ Technical Reports │ │
│ │ Policies & Procedures │ │
│ └────────────────────────────┘ │
└──────────────────────────────────────┘
↕ (via tunnel)
┌──────────────────────────────────────┐
│ Webex Teams Cloud │
│ • Users ask questions │
│ • Bot responds with AI answers │
└──────────────────────────────────────┘
Key Design Principles:
- ✅ Everything local (except Webex messaging)
- ✅ No external AI APIs (Ollama runs locally)
- ✅ Your data stays yours (never leaves your control)
- ✅ Open source components (no vendor lock-in)
Even if you're a beginner, by completing this project you'll understand:
Technical Skills:
- Container orchestration (Podman)
- Vector databases (ChromaDB)
- Local LLM deployment (Ollama)
- Workflow automation (n8n)
- RAG architecture
- Webhook integration
- REST APIs
Practical Knowledge:
- How AI document analysis works
- Privacy-preserving AI deployment
- Enterprise integration patterns
- System troubleshooting
- Production deployment
Career Skills:
- Modern AI/ML deployment
- Infrastructure as code
- DevOps practices
- Technical documentation
Network Engineering:
- Query Cisco network assessments
- Find equipment needing replacement
- Identify security risks
- Budget planning analysis
IT Documentation:
- Search technical runbooks
- Find configuration procedures
- Retrieve troubleshooting steps
- Onboard new team members
Compliance & Policy:
- Query company policies
- Find compliance requirements
- Reference procedures
- Audit documentation access
General Knowledge Base:
- Company wiki alternative
- Technical documentation search
- Team knowledge sharing
- Historical project reference
- Used in the lab: Apple Silicon Mac M4 with 16GB RAM
- Recommended: 24GB+ RAM for better performance
- Storage: 50GB free disk space
- OS This guide is optimized for macOS. Linux adaptation is straightforward; Windows requires WSL2. You can use any AI assistant and provide this current guide and ask for the equivalent in Windows
- Used in the lab: Sonoma (14.0)
- Internet: Required for initial setup and Webex integration
- Optional: Homebrew (package manager) - highly recommended
- Minimum: Basic computer literacy, willingness to learn
- Helpful: Command line experience, basic programming concepts
- Not required: Coding expertise, AI/ML background, DevOps experience
On Mac M4 Pro (16GB RAM):
| Operation | Time | Notes |
|---|---|---|
| Document upload (5 pages) | 30-60 sec | One-time per document |
| First query after startup | 10-15 sec | Model loading (cold start) |
| Subsequent queries | 5-8 sec | Target performance |
| Webex bot response | 5-10 sec | End-to-end with webhook |
Optimization tips:
- Keep system running between queries (avoid cold starts)
- Use SSD storage (faster database access)
Your documents never leave your infrastructure. Perfect for:
- Healthcare (HIPAA compliance)
- Finance (SOC2/PCI requirements)
- Government (data sovereignty)
- Any sensitive corporate information
- Initial setup: $0 (using open source tools and local machine)
- Ongoing costs: $0 (runs on existing hardware)
- Choose your own AI models
- Customize response behavior
- Keep data indefinitely
- No service dependencies
- Works offline
Understanding RAG systems gives you:
- Competitive advantage in AI/ML projects
- Ability to build custom AI solutions
- Knowledge of modern DevOps practices
- Hands-on experience with enterprise AI
- ✅ Step-by-step implementation guides
- ✅ Prerequisites for beginners
- ✅ Advanced features guide
- ✅ Visual references and diagrams
- ✅ Real troubleshooting chronicles
- ✅ Working n8n workflow JSONs
- ✅ Python scripts (load, query)
- ✅ Configuration templates
- ✅ Quick setup scripts
- ✅ Production-tested on macOS M4
- ✅ All commands verified working
- ✅ Common errors documented
- ✅ Performance benchmarked
- ✅ Webex integration validated
- Ollama - Local LLM hosting (llama3.2:3b model)
- ChromaDB - Vector database (v0.4.24)
- Podman - Container runtime (Docker alternative)
- n8n - Workflow automation (visual programming)
- Python - Scripting and automation
- llama3.2:3b - Fast, efficient LLM (2GB RAM footprint)
- nomic-embed-text - Text embeddings (274MB)
- Webex Teams - Enterprise messaging platform
- localhost.run - Secure tunneling (free tier)
- Ollama: Easiest local LLM deployment, no Python dependencies
- llama3.2:3b: Best speed/quality balance for consumer hardware
- ChromaDB: Simple, reliable vector DB with HTTP API
- Podman: Docker-compatible, better security model for macOS
- n8n: Visual workflow design, beginner-friendly, no code required
- ✅ Privacy-first: No external AI API dependencies
- ✅ Beginner-accessible: Complete guides for IT engineers with no AI experience
- ✅ Production-ready: Reliable, tested, documented
- ✅ Enterprise integration: Webex Teams bot
- ✅ Fast responses: Sub-10 second query times
- Simplicity: Use the simplest tool that works
- Documentation: Explain every step, assume no prior knowledge
- Pragmatism: Real solutions from real troubleshooting
- Teachable: Users learn while building
Unlike typical "AI tutorials" that assume coding knowledge:
- Written for IT engineers (network/infrastructure background)
- Uses analogies to familiar concepts (routers, switches, VLANs)
- Every command explained before execution
- Comprehensive troubleshooting from real deployment experience
- pilot-ready, not just installation guide
- Before: Manual document searching: 15-30 minutes per query
- After: AI-powered answers: 5-10 seconds
- ROI: 100x+ time savings on repeated queries
- Before: Static documents, manual search
- After:
- Natural language queries
- Instant answers from any document
- Team-wide access via Webex
- Mobile access (Webex mobile app)
- Multiple query interfaces (CLI, web, messaging)
- Before: Knowledge locked in PDFs and Word documents
- After:
- Searchable knowledge base
- Always available (24/7 with cloud deployment)
- No learning curve for end users
- Scales to entire team
- Historical queries preserved
Completed Features:
- ✅ Core RAG system with vector storage
- ✅ Document loading (txt, doc, docx formats)
- ✅ Python query interface
- ✅ n8n visual workflows (3 complete workflows)
- ✅ Webex bot integration with @mention support
- ✅ Form-based document upload
- ✅ Comprehensive documentation (8 guides, 50K+ words)
Known Limitations:
⚠️ Free tunnel URLs change on restart (use paid ngrok/localhost.run for production)⚠️ macOS-specific instructions (Linux/Windows need adaptation)⚠️ Single collection support (multi-collection is future feature)⚠️ No conversation history (stateless queries, can be added)⚠️ Document format limited to text-based (PDF requires additional setup)
Roadmap: See What's Next
Documentation:
- Improve clarity or fix unclear sections
- Add screenshots/diagrams
- Fix typos or formatting issues
- Share real-world use cases
- Translate to other languages
Code:
- Bug fixes and improvements
- Performance optimizations
- New features (see roadmap)
- Platform adaptations (Linux/Windows)
- Additional document format support
Testing:
- Test on different hardware configurations
- Report issues with detailed logs
- Validate installation guides
- Share feedback on documentation clarity
- Read DOCUMENTATION_STANDARDS.md first
- Test all changes thoroughly on clean system
- Update relevant documentation
- Follow existing code/documentation style
- Provide clear commit messages describing changes
When opening issues, please include:
- Operating system and version
- Hardware specs (CPU model, RAM)
- Which guide/step you're following
- Complete error messages (not screenshots of text)
- What you've already tried
- Output of diagnostic commands
MIT License
What you can do:
- ✅ Use commercially in your organization
- ✅ Modify for your specific needs
- ✅ Distribute modified versions
- ✅ Private use within your company
- ✅ Use in consulting/training
What you must do:
- 📋 Include original license in distributions
- 📋 Include copyright notice
- 📋 State significant changes made
What you cannot do:
- ❌ Hold authors liable for any damages
- ❌ Use trademarks without permission
- ❌ Claim original authorship
See LICENSE file for full legal text.
- Ollama - Making local LLM deployment accessible (ollama.ai)
- ChromaDB - Simple, scalable vector database (trychroma.com)
- llama3.2 - Meta's efficient language model
- n8n - Visual workflow automation (n8n.io)
- Podman - Secure container runtime (podman.io)
This project was built to demonstrate that:
- Privacy-preserving AI is practical for enterprise use
- Local deployment is viable on consumer hardware
- You don't need expensive cloud services for AI pilots
- Open source AI is production-ready
- IT engineers can build AI systems without coding backgrounds
Cisco Live 2026 - Demonstrating practical AI for network engineers
Thanks to everyone who:
- Tested early versions and reported issues
- Provided feedback on documentation clarity
- Shared use cases and deployment stories
- Contributed improvements and fixes
Special thanks to the open source communities behind Ollama, ChromaDB, n8n, and Podman for making this possible.
- 📖 Installation Guides - Complete setup instructions
- 🔧 Troubleshooting - Common issues and solutions
- 💡 What's Next - Advanced features and optimization
- 🎓 Prerequisites - Background knowledge
Before asking for help:
- ✅ Check the Troubleshooting Guide - Most common issues are documented
- ✅ Search existing GitHub issues - Someone may have solved your problem
- ✅ Verify prerequisites - Did you complete all prior steps?
- ✅ Review error messages - They often tell you exactly what's wrong
When you need to ask:
- Open a new issue with details
- Use the issue template (helps us help you faster)
- Include system information and complete error messages
- Describe what you expected vs. what happened
- 💬 GitHub Discussions - Ask questions, share ideas
- 🐛 Report Bugs - Found a problem?
- ✨ Request Features - Ideas for improvements?
If you find this project useful:
- ⭐ Star the repository to show support and help others discover it
- 🍴 Fork for your own use and customize for your organization
- 📢 Share with colleagues who work with sensitive documents
- 💬 Provide feedback to help improve the documentation
- 🤝 Contribute back improvements you make
Let's make privacy-preserving AI accessible to everyone!
🎉 First lab-ready release
Features:
- ✅ Complete documentation suite (8 comprehensive guides)
- ✅ Working RAG system with vector storage
- ✅ Webex Teams integration with bot
- ✅ Three production-ready n8n workflows
- ✅ Python CLI tools for document management
- ✅ Comprehensive troubleshooting guide
Tested & Verified:
- ✅ macOS Sonoma (M4 Pro, 24GB RAM)
- ✅ All installation steps validated
- ✅ Performance benchmarked
- ✅ Common errors documented and resolved
Documentation:
- ✅ 50,000+ words of beginner-friendly documentation
- ✅ IT analogies throughout (routers, switches, VLANs)
- ✅ Every command explained with expected output
- ✅ Real troubleshooting experiences documented
Known Issues:
- Free tunnel URLs require manual update after restart
- macOS-specific (Linux/Windows adaptations needed)
- Single collection limitation (multi-collection in v2.0)
Built with ❤️ for IT engineers who value privacy and control
- ✅ Documents never leave your infrastructure
- ✅ No telemetry or tracking
- ✅ No external AI API calls
- ✅ Complete offline operation (after initial setup)
- ✅ Audit trail optional (you control logging)
- Container isolation prevents cross-contamination
- Local execution limits attack surface
- No cloud credentials required (except optional Webex)
- Regular updates recommended for underlying components
See Security Guide for hardening recommendations.
Ready to build your own private AI assistant?
👉 Start here: Environment Setup Guide
Last Updated: January 2026
Documentation Version: 1.0.0
Project Status: Production Ready