The problem
Content operations were split across materials, editing, models, sources and publishing configuration, requiring repeated manual coordination.
The solution
An internal web workspace supports creating/editing materials, viewing sources, managing several LLM connections and persisting content data.
Implemented features
- Materials management, editing forms and dashboard.
- Multiple OpenAI-compatible LLM connections with Base URL and credentials.
- Model discovery through GET /models and manual Model ID fallback.
- Encrypted API-key storage and basic CMS settings.
Technical approach
Python · FastAPI · SQLAlchemy · SQLite · LLM APIs · Docker Compose. Content, provider connections and persistence are separated into distinct responsibilities.
Demonstration & validation
The repository includes published visuals of the current dashboard, editor, sources and model setup. A manual content workflow is available in the MVP.
Scope & privacy
Automated external ingestion, the complete AI-processing pipeline and CMS publishing are future stages, not implemented features.