Open WebUI with EXAIV: Model Configuration for a Self-Hosted AI Interface
Connect EXAIV as an OpenAI-compatible connection in Open WebUI: UI configuration and Docker environment variable options, model list handling, multi-model setup, and debugging for common connection errors.
Open WebUI with EXAIV: Model Configuration for a Self-Hosted AI Interface
Open WebUI is the most popular self-hosted AI chat interface, and it connects to any OpenAI-compatible endpoint. Point it at EXAIV and you get "your own ChatGPT interface plus a domestic-model backend": conversation history stays on your server, while the models come from EXAIV's catalog (Kimi, GLM, and more). This guide covers both the UI setup and the Docker environment-variable route.
Route one: UI configuration (beginner-friendly)
Open admin settings
Sign in to Open WebUI as an admin; top-right avatar → Admin Settings.Open connection settings
Left sidebar → Connections → find the OpenAI section.Enter URL and key
Set API Base URL tohttps://exaiv.com/api/ai/v1; set API Key to the key from the EXAIV console.Save and pick a model
Save, start a new chat, and choose a model (for examplekimi-k3); if the list is empty, add manually as below.
Model list not showing up?
Open WebUI usually tries to fetch the model list from the endpoint; if it does not arrive, add model IDs manually in the connection's model management — use real IDs from the live /models catalog.
Route two: Docker environment variables
For container deployments, inject connection info directly:
docker run -d -p 3000:8080 \
-e OPENAI_API_BASE_URL="https://exaiv.com/api/ai/v1" \
-e OPENAI_API_KEY="YOUR_EXAIV_KEY" \
-v open-webui:/app/backend/data \
--name open-webui \
ghcr.io/open-webui/open-webui:main
Environment variable names follow official docs
Variable names may change between Open WebUI versions (for example, the historical difference between OPENAI_API_BASE_URL and OPENAI_API_BASE). Check your deployment version's official docs — the names here are the current common form.
Verification and debugging
| Symptom | Check | Fix |
|---|---|---|
| Connection test fails | Base URL complete to /api/ai/v1 | Complete or correct the path |
| 401 | Key correct, no stray spaces | Regenerate and paste |
| Empty model list | Does the endpoint support the list API / add manually | Enter real catalog model IDs |
| Reply timeout | Network path | Check the server's outbound network, retry |
Recommended pairings
| Use | Model | Notes |
|---|---|---|
| Everyday assistant, long summaries | kimi-k3 | Long-context strength |
| Tool calling, multi-turn tasks | glm-5.3-flash | Steady orchestration |
| Lightweight fast chat | A lightweight catalog model | Follow the live /models list |
The self-hosting boundary
Open WebUI self-hosts the interface, but inference still happens on EXAIV's servers — conversation content is sent to the endpoint you configure. If your requirement is "data never leaves the machine room," what you need is a local model deployment (GGUF + Ollama), not a different front end.
FAQ
What is the relationship between Open WebUI and Ollama?
What if the model list will not load?
Can I connect local Ollama and EXAIV at the same time?
What should multi-user deployments watch out for?
Where is pricing?
Related reading
- LobeChat with EXAIV — another chat interface option
- Cherry Studio with EXAIV — a desktop client option
- Kimi K3 and GLM Integration Guide — the full API tutorial
- Qwen GGUF Local Deployment Primer — the fully local route
- Pricing — live pricing and plans