Open WebUI on SovereignEG

Goal: run a ChatGPT-style web chat on your own machine that uses SovereignEG models.

Time: 10 minutes. Needs: Docker.

Open WebUI is a self-hosted chat UI. It can talk to any OpenAI-compatible API, so you point it at SovereignEG and every model your plan allows appears in the model picker.

Step 1 — Start Open WebUI (one command)

docker run -d -p 3000:8080 \
  --name open-webui --restart always \
  -v open-webui:/app/backend/data \
  -e WEBUI_SECRET_KEY="$(openssl rand -hex 32)" \
  -e OPENAI_API_BASE_URL="https://backend.sovereigneg.com/v1" \
  -e OPENAI_API_KEY="sk-YOUR_SOVEREIGNEG_KEY" \
  -e ENABLE_OLLAMA_API=false \
  ghcr.io/open-webui/open-webui:main

PowerShell (Windows) — same thing on one line:

docker run -d -p 3000:8080 --name open-webui --restart always -v open-webui:/app/backend/data -e WEBUI_SECRET_KEY="change-me-to-a-long-random-string" -e OPENAI_API_BASE_URL="https://backend.sovereigneg.com/v1" -e OPENAI_API_KEY="sk-YOUR_SOVEREIGNEG_KEY" -e ENABLE_OLLAMA_API=false ghcr.io/open-webui/open-webui:main

What the variables do:

Step 2 — Open it and create the admin

1. Go to http://localhost:3000.
2. Sign up. The first account becomes the admin.
3. Pick a model from the dropdown at the top (for example SovereignEG/Qwen3.8-27B-FP8) and say hello.

Step 3 — Or add SovereignEG from the UI instead of env vars

Avatar (bottom-left) → Admin Panel → Settings → Connections (newer builds: Settings → Admin → Connections) → Manage OpenAI API Connections → +:

• URL: https://backend.sovereigneg.com/v1
• Key: sk-...
• Click the refresh icon to verify → Save.

The model list is fetched from /v1/models, so you see exactly what your key is allowed to use.

Step 4 — Turn on document chat (RAG) through SovereignEG

By default Open WebUI runs a local SentenceTransformers embedding model (all-MiniLM-L6-v2) inside the container, which costs CPU and up to ~1 GB RAM. Use SovereignEG embeddings instead — lighter, and metered like any other request:

  -e RAG_EMBEDDING_ENGINE=openai \
  -e RAG_OPENAI_API_BASE_URL="https://backend.sovereigneg.com/v1" \
  -e RAG_OPENAI_API_KEY="sk-YOUR_SOVEREIGNEG_KEY" \
  -e RAG_EMBEDDING_MODEL="embeddinggemma-300m" \
  -e RAG_EMBEDDING_BATCH_SIZE=100 \

Add those lines to the docker run command (or set them in Admin Panel → Settings → Documents). Then upload a PDF with the + in the chat box and ask questions about it.

Keep RAG_EMBEDDING_BATCH_SIZE=100. Open WebUI's default is one embeddings request per chunk, all sent at once: a 20-page PDF becomes 60+ requests in the same second, which is more than the default plan's 60 requests per minute, and Open WebUI does not retry a 429, so the upload fails. Batches of 100 turn that into one request.

Step 5 — Lock it down (optional, but do it if others can reach the port)

  -e ENABLE_SIGNUP=false \
  -e ENABLE_PERSISTENT_CONFIG=false \

• ENABLE_SIGNUP=false — no new accounts after yours.
• ENABLE_PERSISTENT_CONFIG=false — env vars stay authoritative; otherwise a setting saved in the UI silently overrides later env changes.

Update / stop / remove

docker pull ghcr.io/open-webui/open-webui:main && docker rm -f open-webui && <same docker run command>
docker stop open-webui          # stop
docker rm -f open-webui         # remove (chats stay in the open-webui volume)

Troubleshooting

Next: OpenCode