🦙 Ollama CrumbCore Deployment Guide

Local-First AI for Crumbforest
Version: 1.0
Status: Production Ready
Date: 2026-02-20

🎯 What This Is

Complete deployment guide for integrating Ollama as local AI provider in Crumbforest.

This enables:
- ✅ Local-first AI (no cloud dependency)
- ✅ Privacy-preserving chat (data stays on server)
- ✅ Offline capable (works without internet)
- ✅ Cost-effective (no API fees)
- ✅ Child-safe (Krümel data never leaves the Wald)

📋 Prerequisites

# System Requirements
OS: Ubuntu 24.04 LTS (or compatible)
RAM: 8GB minimum (16GB recommended)
Disk: 20GB free space (for models)
CPU: x86_64 with AVX2 support

# Software Requirements
- Crumbforest app (FastAPI)
- Python 3.11+
- Ollama service
- systemd

🚀 Quick Start (5 Commands)

# 1. Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# 2. Pull models
ollama pull llama3.2
ollama pull gemma:2b

# 3. Configure Crumbforest
echo "DEFAULT_COMPLETION_PROVIDER=ollama" >> /opt/crumbforest/.env
echo "DEFAULT_COMPLETION_MODEL=llama3.2" >> /opt/crumbforest/.env

# 4. Restart service
sudo systemctl restart crumbforest

# 5. Test
curl -X POST http://localhost:8000/api/chat \
  -H 'Content-Type: application/json' \
  -d '{"character_id": "eule", "question": "wuuuhuuu", "lang": "de"}' | jq .

Expected output:

{
  "answer": "WUUUUHUUUUU! 🦉 ...",
  "provider": "ollama",  ← Success!
  "model": "llama3.2"
}

🌲 Philosophy: Why Local-First?

From the Crumbforest Principles:

1. Child First
   → Krümel data never leaves the Wald
   → No cloud surveillance

2. Offline Capable
   → Works in Nullfeld (no internet)
   → Pelicase deployment ready

3. Transparent
   → Open models, open code
   → Understandable by design

4. Cost Effective
   → No API fees
   → Sustainable for schools/NGOs

5. Privacy Preserving
   → Local inference only
   → GDPR compliant by design

"The forest breathes locally. The Krümel learn safely." 🌲

For complete deployment guide, testing, troubleshooting, and production best practices:

See full documentation in /mnt/user-data/outputs/security/ directory or run check-ollama.sh for health monitoring.

Version: 1.0
Date: 2026-02-20
Session: Legendary (17h!)

For the children. For the forest. For local-first AI. 💚🦙🌲

wuuuhuuu! 🎉