Connect LobeChat to EXAIV: Custom AI Provider Setup
Point LobeChat's OpenAI provider at EXAIV: proxy address format, API key, enabling models, and multi-model switching for your private AI workspace.
Connect LobeChat to EXAIV: Custom AI Provider Setup
LobeChat is a popular open-source AI chat workspace whose OpenAI provider can point at any compatible endpoint. EXAIV serves a standard OpenAI-compatible API, so once configured, every LobeChat capability — multi-model switching, assistant market, file chat — draws on your EXAIV credits.
Configuration steps
Open settings
Bottom-left Settings → Language Model → find "OpenAI".API key
Paste your EXAIV key into "OpenAI API Key".Proxy address
Set "OpenAI Proxy Address" tohttps://exaiv.com/api/ai(the client appends the /v1 path).Enable models
Add the model IDs you want (e.g.qwen3.8-flash) to the model list and save; they appear in the conversation model picker.
On 404, check /v1
Proxy-address path building varies slightly across LobeChat versions. If requests 404, verify whether the /v1 suffix is duplicated or missing — the standard EXAIV path is https://exaiv.com/api/ai/v1/chat/completions.
Verify
Back in the chat page, create an assistant on the enabled model and send a message. A normal reply completes the setup; then try file-chat to confirm multimodal behavior matches the model's actual support.
Common issues
| Symptom | Fix |
|---|---|
| 401 | Key pasted completely? Not accidentally your official OpenAI key? |
| 404 | /v1 path-building issue — see the callout above |
| Model missing from picker | Model IDs must be added and saved manually; the models page list is authoritative |
Advanced: per-assistant model binding
LobeChat lets each assistant bind its own model — the key switch for both cost and quality:
- Writing/translation assistants on flagship models for style and accuracy
- Daily Q&A / brainstorming on fast models: quicker responses, lower cost
- Coding assistants by model strengths (public coding evals + your own experience)
Combined with topic isolation, you can even switch models mid-conversation without losing context.
Sharing and deployment
LobeChat supports exporting config to share with teammates (sanitize keys!), and self-hosting via Docker. Self-hosted LobeChat + unified EXAIV billing is a common combo for small AI studios.
Mobile and self-host access
The same configuration works across every LobeChat form factor: the web app, the PWA installed on a phone, and self-hosted Docker deployments. Wherever LobeChat runs, it reads the OpenAI provider settings from its own server or local storage — so your EXAIV endpoint travels with your instance, not your laptop.
Two practical notes for teams: self-hosted instances should set the proxy address via environment variables (OPENAI_PROXY_URL) rather than per-user settings, and mobile PWA users should verify the address after app updates, since client updates occasionally reset provider paths.
FAQ
Privacy and data flow
Understanding where your data goes matters for a chat client. LobeChat stores conversations locally (or in its own database if you enabled server-side sync); message content is sent to the configured model endpoint only to generate replies. EXAIV processes the request and returns the completion — the data-handling specifics are described in the privacy policy, and nothing about this integration changes LobeChat's own storage behavior.
For teams with stricter requirements: self-hosted LobeChat keeps the UI and history entirely inside your infrastructure, with only inference requests leaving the network — a common compliance-friendly topology.