A reactive notebook for Python — run reproducible experiments, query with SQL, execute as a script, deploy as an app, and version with git. Stored as pure Python. All in a modern, AI-native editor.
LLM Info
This package contains data for the LLM info.
Data
Models and providers are stored in the data directory.
Adding a New LLM Model
If you want to add a new LLM model or provider, you can do so by editing the YAML files in the data directory (models.yml or providers.yml) and running pnpm codegen.
Note: To make it easier for users to choose, keep the number of models to a minimum. Focus on including the latest or recommended models from each provider.
Syncing from models.dev
pnpm sync-models adds the latest models from models.dev to the top of each provider section in models.yml. Existing names, descriptions, and roles are preserved. Machine-sourced metadata is refreshed when upstream supplies it: capabilities, input/output types, release dates, pricing, reasoning options, and token limits. Run pnpm codegen afterwards.
google-vertex maps into the google section, so Anthropic models with @default ids (e.g. claude-opus-5@default) are Vertex-only — note that in their description if you keep them.
pnpm sync-models # all providers, 10 newest each
pnpm sync-models --provider=anthropic # one provider
pnpm sync-models -p openai,google -n 5 # multiple providers, 5 each
pnpm sync-models --metadata-only # refresh existing metadata, add no models
pnpm sync-models --replace # destructive rebuild
Model Metadata
Each provider/model entry can include optional reasoning_options and limits:
reasoning_options:
- type: effort
values: [low, medium, high, xhigh, max]
limits: {context: 1000000, output: 128000}
Reasoning options mirror models.dev: effort lists accepted values (including
null or default when explicitly listed upstream), toggle indicates an
on/off control, and budget_tokens includes optional min and max bounds.
A budget minimum of -1 retains the upstream automatic-budget sentinel.
Token limits are measured in tokens; values are preserved as supplied upstream,
including zero. Omitted metadata is unknown, not an unsupported capability.
Do not copy options between providers or normalize model aliases to enrich metadata: their endpoints can expose different controls and limits.
Use --metadata-only to refresh all machine-sourced metadata in the current catalog without adding models. --metadata-only
cannot be combined with --replace. When upstream omits a field or no exact
provider/model match exists, existing metadata is preserved.