mirror of https://github.com/kevinveenbirkenbach/computer-playbook.git synced 2025-09-24 11:06:24 +02:00

Files

Kevin Veen-Birkenbach f4cf55b3c8 Open WebUI OIDC & proxy fixes + Ollama preload + async-safe pull

- svc-ai-ollama:
  - Add preload_models (llama3, mistral, nomic-embed-text)
  - Pre-pull task: loop_var=model, async-safe changed_when/failed_when

- sys-svc-proxy (OpenResty):
  - Forward Authorization header
  - Ensure proxy_pass_request_headers on

- web-app-openwebui:
  - ADMIN_EMAIL from users.administrator.email
  - Request RBAC group scope in OAUTH_SCOPES

Ref: ChatGPT support (2025-09-23) — https://chatgpt.com/share/68d20588-2584-800f-aed4-26ce710c69c4

2025-09-23 04:27:46 +02:00

config

Open WebUI OIDC & proxy fixes + Ollama preload + async-safe pull

2025-09-23 04:27:46 +02:00

README.md

Ollama

Description

Ollama is a local model server that runs open LLMs on your hardware and exposes a simple HTTP API. It’s the backbone for privacy-first AI: prompts and data stay on your machines.

Overview

After the first model pull, Ollama serves models to clients like Open WebUI (for chat) and Flowise (for workflows). Models are cached locally for quick reuse and can run fully offline when required.

Features

Run popular open models (chat, code, embeddings) locally
Simple, predictable HTTP API for developers
Local caching to avoid repeated downloads
Works seamlessly with Open WebUI and Flowise
Offline-capable for air-gapped deployments

Further Resources

Ollama — https://ollama.com
Ollama Model Library — https://ollama.com/library