Community Contributions
Thank you to everyone who helps make this community better by contributing code, tips, or helping out in the AI Community chat. The following is a collection of learning materials and tools graciously offered to us by community members for everyone to use.
Note
Please note that as we cannot review all contributions in detail, we cannot guarantee their functionality or that they will be kept up to date.
Available Contributions
Below is a comprehensive table of community contributions with key information about each project, a detailed description can be found below:
| Project | Author | Type | Key Features | Use Case | Links |
|---|---|---|---|---|---|
| SAIA-Python | Friedrich Schwarz (Uni Göttingen) | Python Wrapper Library for SAIA | Object-oriented SAIAClient plus standalone functions; covers chat completions (with streaming & tool calling), voice transcription/translation (Whisper), document conversion (Docling/PDF), RAG via ARCANA, model listing, tokenizer utilities (token counting, subword fertility), and rate limit inspection; | Python developers integrating SAIA services into scripts, notebooks, or applications. Install with pip install saia-python | GitHub / Docs |
| SAIA CLI | Egor Kotov (MPI for Demographic Research) | Bash CLI interface for SAIA API. | Single-file script (saia.sh); covers models, rate limits, Docling document conversion, embeddings, audio processing, single-shot chat completions, text/image generation, and ARCANA RAG queries; reads prompts from stdin for pipeline use; requires only curl, jq, and base64; no install needed, download and run | Quick terminal testing, shell pipelines, and lightweight automation without a Python environment | GitHub |
| Docling GWDG Proxy | Marc Schütze | Proxy Service | Lightweight FastAPI service (Docker Compose ready); translates Open WebUI’s docling-serve multipart API format to GWDG’s docling API; returns markdown, JSON, HTML, plain text, and doctags output; configured via env vars (GWDG_API_KEY, GWDG_API_URL); includes health check endpoint; | Open WebUI users who want to use GWDG’s document conversion backend instead of a local docling-serve instance | GitHub |
| ChatAI RAG Tutorial | Luca Tielke (Universität Hamburg) | Training Guide | Step by step blog post (German) covering the complete workflow for building a chatbot with custom document sources using GWDG’s ARCANA service in Chat AI; covers Arcana configuration, document indexing, and querying aimed at users with no prior RAG experience | Learning to create cost-free, privacy friendly custom chatbots with your own sources using Chat AI | Tutorial |
| SAIA OpenCode Plugin | Tobias Weiß (graphwiz-ai) | OpenCode Plugin | Bash script that fetches latest SAIA model list from the API and generates an opencode.json with all models properly categorized (reasoning, coder, vision, medical, research, agentic, large context, general); supports LiteLLM proxy integration; requires only curl and jq | Developers using OpenCode who want to use SAIA models directly in their coding assistant | Codeberg / GitLab |
Project Descriptions
SAIA-Python
SAIA provides self hosted, OpenAI-compatible AI services at GWDG, including chat completions, voice transcription/translation, document conversion, and RAG (ARCANA). This Python wrapper library around the GWDG’s SAIA REST API was developed by Friedrich Schwarz from Uni Göttingen to build a clean, maintainable connection between Python workflows and these services. Friedrich shared the project with GWDG as a potential starting point for collaboration, and has expressed willingness to maintain the package for other users depending on his future position.
The library provides both an object-oriented SAIAClient and equivalent standalone functions for users who prefer a functional style.
Streaming responses and OpenAI SDK compatibility are supported.
API credentials are auto discovered from the SAIA environment variable, a .saia_api file, or a .env file.
Full Sphinx documentation is available online; an interactive demo notebook is included in the repository.
SAIA CLI
A simple, minimal dependency Bash CLI client for the SAIA API. This client provides a minimalistic way to interact with all SAIA service and inference endpoints directly from your terminal, making it ideal for quick testing, shell pipelines, and lightweight automation.
Contributed by Egor Kotov from the Max Planck Institute for Demographic Research as a lightweight alternative for users who do not need a full Python environment. The project was shared with GWDG.
The client is a single file Bash script that requires no installation beyond curl, jq, and base64.
Download the script directly to ~/.local/bin and it is immediately usable.
Supports all major SAIA endpoints: model listing, rate limit checks, Docling PDF to Markdown conversion, embeddings, audio transcription/translation, chat completions with custom system prompts, text generation, image generation, and ARCANA RAG queries.
Accepts prompts via stdin, making it easy to chain into shell pipelines.
Docling GWDG Proxy
A FastAPI based proxy service deployable with Docker Compose.
Open WebUI expects a docling-serve-compatible API, while GWDG exposes its own docling endpoint with a different request/response format.
This proxy bridges the gap by receiving Open WebUI’s multipart file upload, forwarding it to GWDG’s docling API with Bearer authentication, and transforms the response into the docling-serve format Open WebUI expects (returning markdown, JSON, HTML, plain text, and doctags).
Configure via a .env file with your GWDG_API_KEY(SAIA API key).
ChatAI RAG Tutorial
A German language blog post from Universität Hamburg walking through the entire process of setting up a retrieval-augmented chatbot using GWDG’s ARCANA service within Chat AI.
Covers everything from initial Arcana configuration and uploading your own documents to querying the resulting chatbot, with a focus on keeping costs low and data handling privacy friendly.
Suitable for users with no prior RAG experience.
OpenCode SAIA Plugin
An OpenCode plugin developed by Tobias Weiß (graphwiz-ai), based on an original side project by Jaison Lewis at GWDG. It integrates all SAIA (GWDG Chat AI) models directly into OpenCode configurations.
The plugin fetches the latest SAIA model list from the API and automatically generates a properly categorized opencode.json configuration file with models organized by use case: reasoning, coder, vision, medical, research, agentic, large context, and general purpose.
Models are enriched with rich metadata including context window sizes, output limits, reasoning flags, and vision capabilities.
The plugin generates convenient aliases (e.g., saia/best-for-coding, saia/best-for-reasoning) and supports LiteLLM proxy integration for caching and rate limiting.
Installation is cross-platform (Linux/macOS/Windows) and requires only curl and jq.
The plugin is maintained on Codeberg with a mirror on GitLab, and includes JSON schema validation, comprehensive documentation, and troubleshooting guides.