src/models/integrated_selector.py is missing — recommendation_engine import fails

[bradsmithmba]

Summary

src/models/recommendation_engine.py imports IntegratedStrategySelector from src.models.integrated_selector, but that module does not exist on disk. Importing recommendation_engine (and collecting its tests) fails.

Location

src/models/recommendation_engine.py:18:

from .integrated_selector import IntegratedStrategySelector

Reproduction

$ pytest tests/models/test_recommendation_engine.py
ModuleNotFoundError: No module named 'src.models.integrated_selector'

$ ls src/models/integrated_selector.py
ls: src/models/integrated_selector.py: No such file or directory

Notes

The file is genuinely absent (not merely untracked). Note that even if it is re-created, the unanchored models/ rule in .gitignore would silently ignore it (see the related .gitignore issue) until that rule is anchored to /models/.

Fix direction

Implement src/models/integrated_selector.py exposing IntegratedStrategySelector with the interface recommendation_engine.py expects, or remove the dependency if the integrated selector was superseded. tests/models/test_recommendation_engine.py documents the expected behavior and should be used as the contract.

[bradsmithmba]

Scope update after investigation — deferring to the maintainer

Digging into this, the gap is larger than a single missing file and is an architecture decision rather than a mechanical fix, so I'm leaving it for you rather than reconstructing your core model speculatively.

What's actually missing

Two modules, not one:

• src/models/integrated_selector.py — IntegratedStrategySelector(strategy_selector, regime_detector) with rank_strategies_with_regime(market_data, regime_features) -> (rankings, regime_type, regime_confidence)
• src/models/strategy_selector.py — StrategySelector(spatial_net, freeze_spatial_net)

What exists and is fine

• src/models/spatial_net.py — SpatialNet (the chess-inspired spatial encoder)
• src/models/regime_detector.py — RegimeDetector
• src/features/regime_detector.py — a 10-line re-export shim (RegimeDetector from models, RegimeType from data.regime_labeler); not a problem

Why this needs you

The two missing modules are the brain that connects SpatialNet + RegimeDetector into ranked strategy recommendations. Reconstructing them faithfully requires design intent that isn't recoverable from what's left in the repo:

• how raw market_data (prices/volume dict) is encoded into the spatial board tensor SpatialNet expects (7x6x16),
• how SpatialNet's 512-dim output maps to the 16 StrategyType probabilities,
• how regime detection is fused with strategy selection.

tests/models/test_recommendation_engine.py defines the call contract (constructors and rank_strategies_with_regime return shape) but not the architecture. A guessed implementation could diverge from your intent and would be harder to unwind than an honest absence.

Consequence

Until these land, src/models/recommendation_engine.py and src/api/strategy_recommender.py cannot be imported, and tests/models/test_recommendation_engine.py cannot be collected. The Kelly None-rendering branch in PR for #17 is guarded but unexercised for the same reason.

Happy to implement to your spec if you can sketch the two modules' intended architecture, or to add a clearly-labeled placeholder that just satisfies the import/test contract if you want the pipeline importable in the meantime. Your call.