Models and Registry

Models are defined in indusagi/src/ai/models.generated.ts and loaded into an in-memory registry in indusagi/src/ai/models.ts.

Model Structure

Each Model<TApi> includes:

• id and name
• api and provider
• baseUrl
• reasoning support flag
• input types (text, image)
• cost per million tokens for input, output, cache read, cache write
• contextWindow and maxTokens
• Optional headers and compat

Registry Functions

• getModel(provider, modelId) returns a single model by provider and ID.
• getModels(provider) returns all models for a provider.
• getProviders() returns known provider names.

Cost Calculation

calculateCost(model, usage) mutates and returns usage.cost:

• Computes cost based on model rates and usage tokens.
• Handles cache read and cache write separately.

Reasoning Levels

supportsXhigh(model) returns true for specific OpenAI Codex models. Other providers clamp xhigh to high.

Compatibility Overrides

Model.compat allows overriding behavior for OpenAI-compatible providers. See indusagi/docs/ai/providers.md for the full compatibility fields.

Custom Models

The Web UI supports custom providers and model discovery. See indusagi/docs/webui/storage-and-settings.md and indusagi/docs/webui/sandbox.md.