Skip to main content

​
get_small_model

Function
This function is defined in the loader module.
Resolve the auxiliary/“small” model for cheap internal LLM calls. Resolution order:
  1. defaults.small_model (or the equivalent agent.small_model CLI namespace) from the config file, if set.
  2. The supplied primary_model (e.g. the running agent’s model).
  3. defaults.model (or agent.model) from the config file, if set.
  4. fallback (preserves today’s hardcoded behaviour when nothing else is configured).
Both the defaults.* (typed loader) and agent.* (CLI resolver / JSON schema) namespaces are honoured because they read the same config file; a user following schema autocomplete (agent.small_model) is therefore not silently ignored. This keeps internal calls (title generation, summarisation/compaction) on a configured cheap model, and — when unset — on the primary model rather than a hardcoded third-party default.

​
Signature

​
Parameters

​
Optional[str]
The primary model to fall back to when no auxiliary model is configured.
​
Optional[str]
Final fallback used only when neither small_model nor a primary model is available.

​
Returns

​
Optional[str]
The resolved auxiliary model identifier, or fallback.

​
Uses

  • get_default

​
Source

View on GitHub

praisonaiagents/config/loader.py at line 463