Model Settings

Notice: This page has moved to Settings > Model Settings. Please refer to the Model Settings page for the latest information.

Overview

Starnion is not locked into a single AI model. It supports a wide variety of LLM providers — Gemini, OpenAI, Anthropic Claude, Z.AI, and more — and lets users configure their preferred model and API key. Models can be selected per conversation or linked to a Persona so they are applied automatically.

Supported Providers

Provider	Identifier	Key Models	Notes
Google Gemini	`gemini`	gemini-2.0-flash, gemini-1.5-pro	Default provider, free tier available
OpenAI	`openai`	gpt-4o, gpt-4o-mini, gpt-4-turbo	Broad ecosystem, strong coding performance
Anthropic	`anthropic`	claude-opus-4-5, claude-sonnet-4-5, claude-haiku-3-5	Long context, refined reasoning
Z.AI	`zai`	z1-preview, z1-mini	High-performance reasoning models
Custom	`custom`	(user-specified)	OpenAI-compatible endpoints

Default setting: Newly registered users have Gemini set as the default provider.

How to Set Up an API Key

Navigate to Settings in the web UI.
Select the Models tab.
Select the desired provider and enter your key in the API Key field.
Check the models you want to use (multiple selections allowed).
Click Save.

The API key is automatically validated by the backend as soon as it is saved (see API Key Validation below).

Gemini Setup (Free to Start)

The Gemini API offers a free tier, making it a great place to start.

Go to Google AI Studio.
Log in and click Get API key.
Click Create API key → select a project or create a new one.
Copy the generated key (format: AIza...).
Paste it into Starnion Settings > Models > Gemini and save.

Free tier limits: 15 requests per minute, 1,500 requests per day (as of 2025).

OpenAI Setup

Log in to OpenAI Platform.
Click the profile icon in the upper right → API keys.
Click Create new secret key.
Enter a key name and create it (format: sk-proj-...).
The key is only fully visible immediately after creation — copy it right away.
Paste it into Starnion Settings > Models > OpenAI and save.

The OpenAI API is paid. Before use, register a payment method under billing.

Anthropic Claude Setup

Log in to Anthropic Console.
Click API Keys in the left menu.
Click Create Key and enter a key name.
Copy the generated key (format: sk-ant-...).
Paste it into Starnion Settings > Models > Anthropic and save.

Claude is particularly strong for long document analysis and complex reasoning tasks.

Custom Endpoint Setup

You can connect any server compatible with the OpenAI API (Ollama, LM Studio, vLLM, etc.).

Select Settings > Models > Custom.
Enter the server address in the Base URL field (e.g., http://localhost:11434/v1).
For API Key, you can enter any value for local servers.
Enter the model name directly.

Provider Selection Guide

If you are not sure which model to use when, refer to the table below.

Situation	Recommended Model
Everyday conversation, simple questions	Gemini 2.0 Flash (fast, affordable)
Writing and debugging code	GPT-4o or Claude Sonnet
Long document analysis (100K+ tokens)	Claude Opus or Gemini 1.5 Pro
Minimizing costs	Gemini free tier or GPT-4o-mini
Complex math/reasoning	Claude Opus or GPT-4o

API Key Validation

When you save a key, the Gateway sends a lightweight request to each provider’s API to verify its validity.

Gemini: Checks response via GET /v1beta/models?key=<api_key>
OpenAI: Checks for 200 response via GET /v1/models (Bearer auth)
Anthropic: Checks for 401/403 absence via POST /v1/messages with a minimal request
Z.AI: Checks response via GET /api/paas/v4/models (Bearer auth)

Validation results are displayed in the UI immediately. If validation fails, the key is not saved.

When displayed in the UI, the API key is masked showing only the first 4 and last 4 characters (e.g., AIza...zXYZ). The full key is only used server-side.

Cost Monitoring

LLM APIs incur costs based on token usage. Starnion records token usage for all API calls.

Go to Settings > Usage to view daily and per-model token usage.
Input tokens and output tokens are displayed separately.
Estimated costs are calculated based on the per-model unit price.

For more details, see the Analytics & Usage documentation.

FAQ

Q: Can I set up multiple API keys? A: Yes, you can set one key per provider. For example, you can register both Gemini and OpenAI keys simultaneously and select which to use per conversation.

Q: Can I change the active models without changing the API key? A: Yes. If you leave the API Key field empty when saving, the existing key is retained and only the selected model list is updated.

Q: What happens if I exceed the Gemini free tier limit? A: The API returns a 429 error and the conversation is temporarily suspended. Switch to another provider or try again after a short while.

Q: Can I use a different model for each persona? A: Yes, you can assign a specific provider and model to each persona. Configure this under Settings > Personas.