Skip to content

feat: add --model flag to override LLM model per review run #83

@otakakot

Description

@otakakot

Problem Statement

When reviewing different types of changes, the optimal model may vary — a quick fix benefits from a fast, lightweight model, while a complex architectural diff may require a more capable one. Currently, the only way to change the model is to edit the config file (llm.model), which requires a manual change before and after each run. This makes it impractical to switch models on a per-run basis.

Proposed Solution

Add a --model flag to the ocr review command that overrides the model specified in the config for a single run:

ocr review --model claude-opus-4-5 --from main
ocr review -c abc123 --model claude-haiku-4-5

When --model is omitted, the behavior is identical to the current behavior (config value is used). The flag takes precedence over llm.model in the config file and does not persist

— it only applies to the current invocation.

Alternatives Considered

  • Manually editing the config file before each run.
  • Using environment variables to override the model.

Affected Area

CLI / Commands

Additional Context

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request
    No fields configured for Feature.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions