Supported Providers¶

Lumiverse supports 21 AI providers out of the box. Each provider has its own model catalog, API format, and capabilities.

Provider List¶

Custom Base URLs & Reverse Proxies¶

Every provider in Lumiverse lets you override the default API URL on each connection. This means any provider type can be pointed at a reverse proxy, load balancer, or alternative endpoint — not just the Custom provider.

For example, you could create an OpenAI connection but set its API URL to your proxy at https://my-proxy.example.com/v1. Lumiverse uses OpenAI's API format for the request but sends it to your custom URL.

This is useful for:

• Reverse proxies — Route requests through a proxy for logging, rate limiting, or cost tracking
• Regional endpoints — Use a provider's regional API endpoint instead of the default
• Self-hosted mirrors — Point to your own deployment of an API-compatible service

Using the Custom Provider¶

The Custom provider is for services that aren't covered by the built-in providers but implement the OpenAI-compatible API format. This includes:

• Local models — LM Studio, Ollama, text-generation-webui, KoboldCpp
• Other services — Any endpoint with an OpenAI-compatible chat completions API

To use a custom provider:

1. Create a connection with provider set to Custom
2. Enter the API URL (e.g., http://localhost:5000/v1 for a local model)
3. Enter the Model name as the server expects it
4. Add an API Key if the server requires one

OpenRouter¶

OpenRouter is a popular choice because it gives you access to hundreds of models from many providers through a single API key:

1. Get an API key from openrouter.ai, or click Sign in with OpenRouter on the connection form to do a PKCE OAuth flow that auto-fills the key.
2. Create a connection with provider set to OpenRouter.
3. Use the Models button to browse available models.

The OpenRouter connection form has two extra sections beyond the basics:

The connection card also shows your live OpenRouter credit balance — handy as a usage gauge.

Anthropic Prompt Caching¶

Anthropic connections expose a per-connection prompt caching panel. Caching lets Anthropic reuse identical request prefixes between calls — you pay a small write cost on the cached call and get cheaper, faster reads on every subsequent call until the cache expires.

Settings¶

Automatic and explicit breakpoints can be combined. If you enable caching without ticking any breakpoint, Lumiverse falls back to automatic mode so the toggle never produces an empty cache_control.

When to Use Each Pattern¶

The connection card shows a compact summary (e.g. Cache 5m • auto + system) so you can verify the active configuration at a glance.

Z.AI Coding Plan¶

Z.AI ships two API URLs that share the same authentication but route to different upstreams:

On a Z.AI connection, toggle Use Coding Plan Endpoint to route through /api/coding/paas/v4. Leave it off for normal API keys. Lumiverse rewrites the base URL accordingly — you don't have to edit the API URL field by hand.

Z.AI does not expose an OpenAI-compatible /models endpoint, so Lumiverse ships a built-in model list (glm-5.1, glm-5-turbo, glm-5, glm-4.7 family, glm-4.6, glm-4.5 family, glm-4-32b-0414-128k) and validates your key by sending a minimal chat/completions request rather than a model list call. This is what keeps Coding Plan keys working — they 404 the model list endpoint but accept chat requests fine.

Google Vertex AI¶

Vertex AI is Google's enterprise Gemini endpoint. Lumiverse authenticates with a service account JSON rather than an API key.

1. Create a service account in your GCP project with the Vertex AI User role and download the JSON key.
2. Create a connection with provider set to Google Vertex AI.
3. Paste the entire service account JSON into the API Key field.
4. Pick a Region (e.g. us-central1, europe-west4) in the Vertex metadata section, or leave it on global for the global endpoint.
5. Use the Models button to populate the model list from your project.

Because Vertex routes by project + region, the API URL is derived automatically from your service account and region selection — there's nothing to fill in there yourself.

NanoGPT Usage¶

NanoGPT connections show a subscription usage card directly on the connection profile, similar to the credits readout on OpenRouter. Use it to watch your remaining quota without leaving Lumiverse.

Provider Capabilities¶

Not all providers support all features:

Structured Output¶

Different providers handle structured output differently:

• Google Gemini — Pass responseMimeType and responseSchema in parameters
• OpenAI-compatible — Pass response_format in parameters
• Anthropic — Use tool definitions for structured output