🧩 End-to-End Workflow: The 100% Free & Local AI Workstation

This recipe shows how to chain Ollama (Local LLM), Whisper (Speech-to-Text), and Open WebUI (ChatGPT Clone) into a fully offline AI workstation.

---

flowchart LR
    Mic[Microphone / Audio File] --> Whisper[OpenAI Whisper Engine]
    Whisper -->|Text Transcription| WebUI[Open WebUI Interface]
    WebUI -->|Inference Request| Ollama[Ollama Local LLM]
    Ollama -->|Stream Response| WebUI

---

Step 1: Start Ollama Model Server

# Pull and start your chosen LLM (e.g. Llama 3.1 8B)
ollama pull llama3.1
ollama run llama3.1

Step 2: Launch Open WebUI with Docker

docker run -d -p 3000:8080 \
  --add-host=host.docker.internal:host-gateway \
  -v open-webui:/app/backend/data \
  --name open-webui \
  --restart always \
  ghcr.io/open-webui/open-webui:main

Step 3: Transcribe Voice / Meetings Offline with Whisper

whisper recorded_meeting.mp3 --model medium --output_format txt

Step 4: Drop Transcript into Open WebUI

Paste the generated .txt directly into your local Open WebUI at http://localhost:3000 with the prompt:

Summarize the key decisions, action items, and owners from this meeting transcript into a bulleted Markdown list: