feat: (m-decompose) Module Prompt V3

[csbobby]

Module Prompt V3

Description

• Enhance the module execution and extraction robustness.

Implementation Checklist

Protocol Compliance

• parts() returns list of constituent parts (Components or CBlocks)
• format_for_llm() returns TemplateRepresentation or string
• _parse(computed: ModelOutputThunk) parses model output correctly into the specified Component return type

Testing

• Tests added to tests/components/
• New code has 100% coverage
• Ensure existing tests and github automation passes (a maintainer will kick off the github automation when the rest of the PR is populated)

[github-actions]

The PR description has been updated. Please fill out the template for your PR to be reviewed.

[csbobby]

Updated the modular prompts to adapt the recent llms.

[csbobby]

Hi This passes the tests and ready to be reviewed @jakelorocco.

[planetf1]

A few observations — mostly around parsing robustness and test coverage. Also worth checking whether the from e exception-chaining fix applied in _general_instructions.py is needed in any equivalent except blocks in _subtask_constraint_assign.py.

[planetf1]

Hardcoded model mistral-small3.2:latest —can we make it variable?
Is the model used documented in the test docs?
How much memory does it need? Is it reasonable on a smaller system (though if it's needed for results, understandable).

[csbobby]

This is a CLI example shell script and the variable MODEL_ID is configurable as an example here. We assume users will have their own scripts but the model is recommended with proper size and capacities.

[csbobby]

Hi Thedocs/examples/m_decompose/python/ is a tmp folder generated during the runtime (the model id there is generated by the variable model_id). The tmp folder should and will be removed in the next PR but just keeps now to avoid the merge conflict.