Back to lab
Live playground
Repo-driven demo

Open WebUI Tools

This demo is grounded in the actual repo code and docs. The YouTube transcript flow now uses a small public transcript API and is ready to be paired with an Ollama Cloud-style summarizer or any OpenAI-compatible LLM API.

What this page demonstrates

1. Keyword-based model routing logic from the auto router pipe.

2. Diagram code safety validation inspired by the Python diagram generator.

3. Real YouTube transcript extraction via a small public API, ready for summarization in an Ollama Cloud-style or any OpenAI-compatible API.

Why this is built this way

It is grounded in the actual repository structure and docs.
It fetches real transcript text through a tiny public API, instead of faking transcript output.
It leaves summarization to a configurable LLM endpoint when you want the next step.

Auto router playground

Based on the repo README and tests for the auto router model pipe. This demo shows how a user query maps to a model based on keyword rules.

Selected model: codellama:latest
Matched 3 keywords for codellama:latest.

codellama:latest

Coding tasks

Matched keywords: code, python, debug

gpt-4o

Complex reasoning

Matched keywords: explain

Diagram safety validator

Based on the validation logic inside `diagram_generator.py`. Paste diagram code and see whether the website would treat it as safe enough for a controlled demo context.

Validation passed

The code contains a diagrams import, uses a Diagram context manager, and does not trip blocked imports or dangerous function calls.

YouTube transcript demo

This demo mirrors the Open WebUI transcript tool flow: parse a YouTube URL, fetch transcript text from a tiny public API, and hand the output to an LLM later. If you want to summarize it automatically, point the result at an Ollama Cloud-style or other OpenAI-compatible endpoint.

Parsed video ID: dQw4w9WgXcQ

The public transcript API works with a plain POST request, so it fits the static-export setup of this profile site.