Harden tier bucketing and PDF fallback text

.github/workflows/go.yml

@@ -38,6 +38,25 @@ jobs:
       - name: Build
         run: go build -v ./cmd/gh
 
+  hemmah-python-tests:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Check out code
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.10'
+
+      - name: Install Python dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements/hemmah-test.txt
+
+      - name: Run Hemmah pytest
+        run: pytest tests/test_hemmah_pro.py
+
   integration-tests:
     env:
       GH_TOKEN: ${{ github.token }}

docs/hemmah_example.md

@@ -0,0 +1,63 @@
+# Hemmah Pro IVS 2025: CSV ➜ PDF Example
+
+This is a minimal, reproducible local run that takes a CSV input and produces a PDF report using `hemmah_pro_ivs_2025.py`.
+
+## 1) Install Python dependencies
+
+```bash
+pip install pandas numpy matplotlib scikit-learn shap xgboost ipywidgets ipython fpdf arabic-reshaper python-bidi
+```
+
+## 2) Use the sample CSV
+
+A ready-to-run test dataset is included at:
+
+```
+docs/hemmah_sample.csv
+```
+
+The minimum required columns are:
+- `price` (or any column containing `price/سعر/value/قيمة`)
+- `area` (or any column containing `area/مساحة/size/المساحة`)
+
+## 3) Run the end-to-end script
+
+From the repo root:
+
+```bash
+python - <<'PY'
+from hemmah_pro_ivs_2025 import HemmahDataEngine, HemmahMLEngine, HemmahReportGenerator
+
+engine = HemmahDataEngine()
+engine.load_data("docs/hemmah_sample.csv")
+engine.ivs_quality_check()
+engine.clean_and_engineer()
+model_df, features, target = engine.get_modeling_data()
+
+ml = HemmahMLEngine()
+ml.train_multiple_models(model_df, features, target)
+
+sample_input = model_df[features].head(1)
+result = ml.predict(sample_input)
+
+report = HemmahReportGenerator(engine, ml, result)
+pdf_name = report.generate_pdf("Hemmah_Test_Report.pdf")
+
+print("PDF generated:", pdf_name)
+PY
+```
+
+## 4) Output
+
+You should see training output in the console, and a PDF named:
+
+```
+Hemmah_Test_Report.pdf
+```
+
+in the current working directory.
+
+## Notes on resource loading
+
+- The PDF generator tries to load `Amiri-Regular.ttf` and `Amiri-Bold.ttf` from the current working directory. If they are not present, the report falls back to Latin-1 safe English text to avoid `UnicodeEncodeError` in FPDF.
+- This module does not currently use Jinja2 templates; all report content is built in code.

docs/hemmah_sample.csv

@@ -0,0 +1,11 @@
+price,area,district,sale_date,rooms,age
+321958,871,الروضة,2024-08-17,6,26
+871155,909,الياسمين,2024-10-06,7,25
+331932,615,النرجس,2024-03-31,7,15
+459178,446,الروضة,2024-03-05,4,25
+310268,1035,النرجس,2024-03-20,2,22
+490012,522,الملز,2024-09-11,3,10
+785430,780,الياسمين,2024-11-19,5,12
+612345,640,الروضة,2024-01-28,4,18
+275900,350,الملز,2024-07-03,3,9
+999999,1100,النرجس,2024-12-01,6,8