Continuous enhancement: Quantamental AI model selection

[ryun6249]

Continuous Enhancement Report

1. Current Status

• Project purpose: FastAPI-served local financial research workstation for market, macro, Quant Lab, Quantamental, ML Forecast, AI Portfolio, and grounded AI briefing workflows.
• Existing structure preserved: yes. Static app/web UI, /api/v1/config, Quantamental API contracts, deterministic default analysis, and existing All/quality/range behavior are preserved.
• Main issues found: Quantamental AI report/Q&A had backend support for model and use_llm, but the UI always sent deterministic requests and did not expose truthful Qwen/Gemma runtime-checked selection.

2. Changes Made

Compatibility

• Added model metadata to existing config model options without removing existing fields.
• Kept deterministic Quantamental analysis as the default initial path.

Data Consistency

• No market data schema or calculation behavior changed.
• Existing global range and quality summaries were preserved.

UI/UX

• Added a Quantamental AI 모델 selector in the analysis report card.
• Shows deterministic default plus Qwen/Gemma runtime-checked options from /api/v1/config.
• Added responsive styling and bundle cache-bust version 20260519-continuous-enhancement-v3.

Visualization

• No chart renderer changes in this slice.
• Browser smoke re-verified Quantamental overview axes and mobile overflow.

AI Briefing

• Explicit AI report/Q&A actions now send use_llm=true and the concrete configured model only when a runtime-checked model is selected.
• Deterministic fallback remains active if local provider execution fails.

Translation

• Added Korean/English status copy that does not claim Qwen/Gemma are installed before runtime verification.

Performance

• No background LLM calls, polling, or additional dashboard fetch loops were added.

Tests / Validation

• Added API/UI contract coverage for runtime-checked model metadata and model selector markers.

3. Changed Files

File	Change	Reason
app/api/routers/system.py	Add concrete model field to UI model options	Avoid inferring model names from labels
app/web/index.html	Add Quantamental AI model selector/status	Let users choose deterministic/Qwen/Gemma truthfully
app/web/app.js	Render model options and wire AI report/Q&A payloads	Apply selection only to explicit AI actions
app/web/styles.css	Responsive selector layout	Avoid desktop/mobile overflow
scripts/check_ui_contract.py	Add UI/JS contract markers	Prevent regression
scripts/ai_portfolio_ui_smoke.py	Bump app bundle version	Match cache-busted UI
tests/test_api_routing_contract.py	Assert runtime-checked model metadata	Verify model selection is not fake
tests/test_ui_routing_contract.py	Assert selector, CSS, and payload wiring markers	Verify static UI contract
docs/CONTINUOUS_ENHANCEMENT_LOG.md	Record analysis, implementation, validation	Keep automation log current

4. Validation Results

Check	Command / Tool	Result	Notes
Lint	git diff --check -- ...	Passed	No whitespace errors in touched files
Build	node --check app/web/app.js; python -m py_compile app/api/routers/system.py scripts/check_ui_contract.py scripts/ai_portfolio_ui_smoke.py	Passed	Repo has no npm/pnpm package build
Test	python -m pytest -q	Passed	691 passed, 9 subtests passed in 140.47s
UI validation	Quantamental + AI Portfolio Playwright/browser smoke	Passed	AI Portfolio passed on standalone retry after parallel Macro timeout
Data validation	python scripts/check_ui_contract.py; Quantamental browser smoke	Passed	Quality/range and Quantamental surfaces still present
AI briefing validation	tests/test_quantamental_api.py, tests/test_quantamental_ui_ai_panel.py, Playwright intercepted POST	Passed	Qwen selection sends use_llm=true, model=qwen2.5:7b

5. Remaining Risks

• Qwen/Gemma options are runtime-checked, not preflight-proven installed models.
• Local provider failure remains intentionally handled by deterministic fallback.
• The worktree had unrelated dirty files before this run; this PR only stages the files listed above.

6. Next Enhancement Suggestions

1. Add a lightweight provider health indicator beside the Quantamental model selector.
2. Add the same explicit model selection pattern to Macro/Forecast AI surfaces where backend contracts support it.
3. Normalize enhancement-log appending to avoid duplicated historical sections during repeated automation runs.

7. Safety Notes

• main/master was not modified directly.
• Auto-merge was not enabled.
• Secrets were not touched.
• Trading/order execution logic was not modified.
• Unsupported facts were not presented as verified.