CMacD/AIC_PHASE1_POC

Phase 1 QC: trust exact lookup matches; honest log for lookup-only attrs

Phase 1: strip '|' from DESCRIPTION (clients reject pipe-delimited submissions)

Phase 1: calibration sklearn>=1.6 fix + analyst-facing log abstraction

Phase 1 (Streamlit): match Attribute_Type case-insensitively in META filter

HF deploy: Phase 1 refactor, NLTK downloaded at runtime (no bundle)

Phase 1: cap XGBoost at 500 classes for very-high-cardinality attributes

Update README.md

Merge pull request #9 from Ciaran-MacDermott/master

Merge pull request #8 from Ciaran-MacDermott/main

Remove redundant custom collapse toggle from Step 2 post-QC export

Merge pull request #7 from Ciaran-MacDermott/master

Merge pull request #6 from Ciaran-MacDermott/main

Restructure Phase 1 ML modules into ml_package/

Merge pull request #5 from Ciaran-MacDermott/master

Update README with deployment notes and repo structure for DevOps handoff

Merge pull request #4 from Ciaran-MacDermott/main

Remove desktop-app and archive files not needed for Docker/Streamlit deployment

Merge pull request #3 from Ciaran-MacDermott/master

Add PL column base name override for multi-model projects

Merge pull request #2 from Ciaran-MacDermott/master

Improve Step 13 and Step 16 log coherence

Step 16: drive TOOL vs MDM QC from Attributes.txt, not META

Improve pipeline log messages for analysts

Pin streamlit-aggrid to 0.3.4.post3 and remove dead import

Merge pull request #1 from Ciaran-MacDermott/master

Update page titles to clarify phase coverage for analysts

Remove os.chdir() from Phase 1 pipeline — prevents future serialisation issues

Switch BM25Okapi to BM25Plus in TextMatch

Show pipeline log only on completion, not during execution

Remove stale log_lines_snapshot parameter from _worker_phase_b

Fix Phase 3 Step 15 null check running on wrong DataFrame

Fix QC wizard edit propagation and Phase 3 Run button disabled after Phase 1 no-zip

Normalise 'NULL VALUE' sentinel before ML feature building

Fix Phase 1 lkp coverage gaps and QC false-positive flags

Fix pipeline log expanders to show full log instead of trimmed display log

Default brand override manufacturer column to RAW_MANUFACTURER

Update Phase 1 input requirements to cover ZIP and individual file modes

Fix truncated QC log download + purple download buttons in Phase 3

Fix brand override SelectboxColumn clearing values on cell navigation

Style download button purple to match brand theme

Revert to native st.download_button — remove Save As iframe approach

Fix AgGrid blank render — remove per-row JS array embedding

Fix blank QC wizard on session timeout or state drift

Remove 'optional' from Attributes/AttributeValues upload labels

Download UX: native button + Save As option side by side

Rename pages for cleaner sidebar labels

UX: Phase 2/3 handoff, brand override stability, log continuity

Phase 3 UX: stage stepper, JS timer, log expanders, elapsed time

Perf: frozenset stopwords, O(n) feature concat, drop sorted() dedup

Rename entry point to 1_AIC_Mapping.py, Phase 2/3 page title to AIC Phase 3