Local AI / LLM // VERSION: v0.3.12 Stable // LICENSE: MIT License
● 100% LOCAL LAN EXECUTION (2.2 GB Quantised Q4_K_M • 8 GB RAM)
Ollama + OpenWebUI Air-Gapped Corporate Document AI
Protocols: Local RAG • Zero Data Egress • UK GDPR Safe • UK Regulatory Clearance: 100% UK GDPR & ISO 27001 compliant: zero corporate data leaves your premises; MIT-licensed runtime engine.
Architectural Overview
Runs quantised open-weights language models (Llama-3 / Mistral / Qwen) locally on your site server with Retrieval-Augmented Generation (RAG) over your organisation’s ISO 22301 runbooks, switchboard schedules, and supplier contracts.
72-Hour Blackout Plant-Room Use Case
Allows duty managers at 3 a.m. to query site-specific valve locations, generator changeover sequences, and contractor emergency numbers offline—without leaking sensitive corporate data to public cloud AI APIs.
Production Configuration & Deployment Snippet
# Run air-gapped local model with zero external network calls OLLAMA_HOST=0.0.0.0:11434 OLLAMA_NOHISTORY=1 ollama serve ollama run llama3.2:3b-instruct-q4_K_M \\ "Summarise the BS 7671 changeover isolation steps for Sub-Board DB-2"