Add guided judge polish

README.md

@@ -20,6 +20,7 @@ The submission version is feature-frozen around the complete demo loop:
 - Record a transaction with your microphone and transcribe it with faster-whisper.
 - Configure seller context: business name, currency label, low-stock threshold, and language style.
 - Pick common currency presets for INR, USD, EUR, GBP, MXN, and BRL, or enter a custom currency label.
+- Follow a Guided Judge Mode path for health check, seeded data, sale capture, save, dashboard, and reports.
 - Use a first-run “Start in 60 seconds” guide for setup, first transaction, review, and closeout.
 - Type or paste a transaction note.
 - Bulk import multiple pasted notes for review and editing.
@@ -27,6 +28,7 @@ The submission version is feature-frozen around the complete demo loop:
 - Parse common English, Hinglish, Hindi-lite, Gujarati-lite, Spanish, French, and Portuguese seller phrases with deterministic fallback rules.
 - Review a human-friendly transaction card with warning badges before saving.
 - See a language/confidence chip after parsing so multilingual notes are visibly handled.
+- Use price memory to fill missing sale price/amount from the latest saved item sale.
 - Correct transaction type, item, customer, quantity, price, amount, notes, and confidence directly before saving.
 - Keep a correction log for review edits so model/rule mistakes become visible field-test evidence.
 - Choose `Cloud AI first` or `Local fallback only` from the Record screen to make the AI route explicit during demos.
@@ -40,6 +42,8 @@ The submission version is feature-frozen around the complete demo loop:
 - Generate WhatsApp-ready daily business summaries in English, Spanish, French, or Portuguese.
 - Run shortcut commands like `close today`, `show Amit`, or `stock mangoes`.
 - Generate customer follow-up reminders and an inventory reorder list.
+- Use a debt reminder queue sorted by highest customer balance.
+- Use reorder intelligence that combines stock threshold with recent sold quantity.
 - Capture field-test evidence with a seller checklist and anonymized feedback notes.
 - Offload speech transcription and LLM parsing to optional Modal endpoints.
 - Edit or delete saved transactions while keeping customer credit and inventory balances consistent.
@@ -122,7 +126,7 @@ The same examples are available in `sample_data/demo_transactions.txt`.
 
 Capture these three moments for the Space README, demo video, or social post:
 
-- `Record Text & Voice`: Today’s Work quick actions, review card, warning badges, and save receipt.
+- `Record Text & Voice`: Guided Judge Mode, Today’s Work quick actions, review card, warning badges, price memory, and save receipt.
 - `Dashboard`: Insight Coach, metrics, seller-day activity timeline, sales/expense timeline, top-selling item, low-stock table, and outstanding credit.
 - `Field Test`: seller checklist, anonymized “who/tried/changed” notes, and saved evidence status.
 - `Reports & PDF` plus `Ledger`: Daily Closeout, PDF/WhatsApp export, CSV download, and ledger correction.
@@ -198,14 +202,18 @@ curl -X POST https://sagarpat3199--voiceledger-api.modal.run/parse \
 - The Insight Coach turns those dashboard signals into next actions, such as credit follow-up, restocking, and daily closeout.
 - Customer and inventory detail views show transaction history for one customer or item.
 - Smart review warnings flag low confidence, missing fields, duplicate risk, and negative stock before save.
+- Price memory fills missing sale prices from the latest matching saved sale and marks the review with a `Price memory used` badge.
 - Inline review editing lets sellers correct the parsed transaction before the save touches the ledger.
 - Seller setup persists business name, currency label, low-stock threshold, and language style in SQLite.
 - Currency presets make the app easier to demo for INR, USD, EUR, GBP, MXN, BRL, and custom local labels.
 - Field Test persists anonymized seller evidence and a workflow checklist in SQLite.
+- Field Test evidence includes pain point, before/after story, useful moments, and changed-after-feedback notes.
 - The Field Test mistake log records corrected fields before save so product feedback is visible.
 - Local fallback rules include common English, Hinglish, Hindi-lite, Gujarati-lite, Spanish, French, and Portuguese seller notes.
 - Parse status includes a lightweight language/confidence chip for multilingual seller notes.
 - Customer follow-up and inventory reorder helpers generate WhatsApp-ready action messages.
+- Debt reminder queue sorts customers with outstanding balances by highest amount due.
+- Inventory reorder intelligence adds recent sold quantity, low-stock/selling-fast status, and suggested actions.
 - WhatsApp summaries provide short copyable daily recaps in English, Spanish, French, or Portuguese.
 - Voice command shortcuts provide quick access to daily closeout, customer detail, and inventory detail.
 - Bulk import splits pasted notes by line, parses each line, supports review edits, and saves all reviewed transactions.

docs/demo-video-script.md

@@ -8,7 +8,7 @@ VoiceLedger is a voice-first bookkeeping app for informal sellers and home busin
 
 ## 0:10-0:25 — Judge Flow and Health
 
-On the first screen, show the `Judge Demo Flow` panel:
+On the first screen, show `Guided Judge Mode` and the `Judge Demo Flow` panel:
 
 ```text
 1. Seed demo data -> 2. Record/type -> 3. Save -> 4. View dashboard/reports
@@ -47,6 +47,7 @@ Click parse. Show:
 
 - Transcript if using voice
 - Human-friendly transaction review card
+- Price memory badge if a known item is missing price
 - Inline review fields for correcting item, quantity, amount, customer, and notes before saving
 - Warning badges for low confidence, missing fields, duplicate risk, or negative stock when present
 - Parse source/status
@@ -62,7 +63,8 @@ Use the `Sections` navigation to open:
 - `Dashboard` to show sales/profit/credit and the sales/expense timeline
 - `Field Test` to show the seller checklist and anonymized “who/tried/changed” notes
 - `Customer Credit` to show dues, a selected customer detail view, and a follow-up message
-- `Inventory` to show stock, a selected item detail view, and the reorder list
+- `Customer Credit` to show the debt reminder queue sorted by highest balance
+- `Inventory` to show stock, a selected item detail view, and reorder intelligence
 - `Ledger` to show saved transactions
 
 Say: “The model handles messy input, but the app owns the accounting state.”
@@ -77,7 +79,7 @@ Open `Reports & PDF`, run `Daily Closeout`, then generate the PDF, translated Wh
 
 Suggested screenshot/GIF moments:
 
-- Record flow: first-run guide, multilingual examples, review card, inline correction, parse source, warning badges, save receipt.
+- Record flow: Guided Judge Mode, first-run guide, multilingual examples, review card, price memory, inline correction, parse source, warning badges, save receipt.
 - Seller setup: currency/threshold/language style and Command Center.
 - Field Test: seller checklist, feedback notes, and correction log.
 - Submission Story: AI pipeline and small-model fit card.

docs/field-notes.md

@@ -34,7 +34,7 @@ This is a good small-model problem because the output space is constrained: a tr
 1. Seed demo transactions from the first-screen Judge Demo Flow.
 2. Record or type: `Sold 12 mangoes, 20 each`.
 3. Show the parse source, review card, confidence, and warning badges.
-4. Correct one inline review field before saving to show that sellers can fix model or rule-parser mistakes immediately.
+4. Show price memory when a known item is missing price, then correct one inline review field before saving to show that sellers can fix model or rule-parser mistakes immediately.
 5. Save the transaction and show the receipt with bookkeeping side effects.
 6. Show Dashboard totals, the Insight Coach, sales/expense timeline, Ledger row, Customer Credit detail, and Inventory detail.
 7. Open Field Test to show the seller checklist and anonymized feedback notes.
@@ -52,9 +52,10 @@ The hackathon version is intentionally frozen around the complete bookkeeping lo
 - Seller setup for business name, currency, low-stock threshold, and language style.
 - Currency presets and translated WhatsApp summaries for sellers who demo in INR, USD, EUR, GBP, MXN, BRL, English, Spanish, French, or Portuguese contexts.
 - Hinglish/Hindi-lite/Gujarati-lite/Spanish/French/Portuguese fallback examples for common seller notes.
-- First-run onboarding, multilingual examples, language/confidence chips, human-friendly review cards, inline review correction, mistake logging, warning badges, receipts, and save feedback.
-- Command Center, dashboard timeline, seller-day timeline, customer follow-up, inventory reorder list, ledger correction, Daily Closeout, PDF, CSV, and WhatsApp exports.
+- Guided Judge Mode, first-run onboarding, multilingual examples, language/confidence chips, price memory, human-friendly review cards, inline review correction, mistake logging, warning badges, receipts, and save feedback.
+- Command Center, dashboard timeline, seller-day timeline, debt reminder queue, customer follow-up, inventory reorder intelligence, ledger correction, Daily Closeout, PDF, CSV, and WhatsApp exports.
 - Field Test Mode for the seller checklist and anonymized “who/tried/changed” feedback evidence.
+- Richer Field Test Evidence for pain point, before/after workflow, useful moments, and changed-after-feedback notes.
 - Demo Health for backend observability.
 - Submission Story for the AI pipeline and small-model fit rationale.
 
@@ -62,13 +63,13 @@ This freeze keeps the demo reliable and makes the value proposition easier to ju
 
 ## What We Learned
 
-The strongest product behavior is not the model call itself. It is the reliable loop around it: voice input, readable review, warning badges, save receipt, correction, and export. Informal sellers need speed, but they also need a way to spot mistakes and fix them. The correction log makes those mistakes measurable during field tests, while edit/delete, detail drilldowns, Daily Closeout, and CSV export make the app credible as a bookkeeping tool instead of a parser demo.
+The strongest product behavior is not the model call itself. It is the reliable loop around it: voice input, readable review, price memory, warning badges, save receipt, correction, reminders, restocking, and export. Informal sellers need speed, but they also need a way to spot mistakes and fix them. The correction log makes those mistakes measurable during field tests, while edit/delete, detail drilldowns, Daily Closeout, and CSV export make the app credible as a bookkeeping tool instead of a parser demo.
 
 The small-model constraint helped keep the app honest. The model handles the fuzzy human input; the code owns the accounting state.
 
 ## Screenshot Moments
 
-- Record flow: first-run onboarding, multilingual examples, language/confidence chip, voice command shortcuts, review card, inline corrections, warning badges, source/status, save receipt.
+- Record flow: Guided Judge Mode, first-run onboarding, multilingual examples, language/confidence chip, voice command shortcuts, price memory, review card, inline corrections, warning badges, source/status, save receipt.
 - Dashboard: daily totals, Insight Coach, outstanding credit, sales/expense timeline, top seller, low-stock inventory.
 - Field Test: workflow checklist, correction log, and anonymized feedback evidence.
 - Reports and exports: Daily Closeout, PDF download, WhatsApp summary, CSV ledger export.

docs/submission-checklist.md

@@ -12,6 +12,7 @@
 
 - `Sections` navigation opens the correct page and no old Gradio tab row is visible.
 - First screen shows `1. Seed demo data -> 2. Record/type -> 3. Save -> 4. View dashboard/reports`.
+- First screen shows Guided Judge Mode progress states.
 - `Submission Story` shows the AI pipeline strip.
 - `Submission Story` shows the “Why small models fit” card.
 - `Demo Health` shows Modal reachable and backend version.
@@ -22,6 +23,7 @@
 - Local fallback parses `Amit ne 100 dene hai`, `Amit ne 50 diya`, and `50 mango kharida`.
 - Voice parse works and transcript appears.
 - Review card shows type, item, quantity, price, amount, customer, source, and confidence.
+- Price memory fills a missing sale price for a previously sold item and shows `Price memory used`.
 - Parse status shows source plus language/confidence.
 - Smart warnings appear for low confidence, missing fields, duplicate risk, and negative stock.
 - Save shows a receipt with transaction type, amount, and side effects.
@@ -33,15 +35,17 @@
 - Customer Credit section reflects `Amit owes 100` and `Amit paid 40`.
 - Customer detail lookup shows customer transaction history and status.
 - Customer follow-up generates a WhatsApp reminder.
+- Debt reminder queue shows customers with outstanding balances sorted highest first.
 - Inventory reflects purchases and sales.
 - Inventory detail lookup shows bought, sold, current stock, and low-stock status.
 - Inventory reorder list uses the configured threshold.
+- Inventory reorder intelligence shows recent sold quantity, status, and suggested action.
 - Ledger edit/delete works and derived balances rebuild.
 - CSV export downloads from Ledger.
 - Daily Closeout generates PDF, CSV, WhatsApp summary, and a status line.
 - PDF report downloads from Reports.
 - WhatsApp summary generates copyable text in English, Spanish, French, and Portuguese.
-- Field Test correction log shows edited review fields.
+- Field Test evidence saves pain point, before/after story, useful moments, changed-after-feedback notes, and correction log.
 
 ## Prize Alignment
 

tests/test_settings.py

@@ -10,6 +10,11 @@ def test_business_settings_defaults_load_when_empty(tmp_path: Path) -> None:
 
     assert settings["business_name"] == "VoiceLedger Seller"
     assert settings["currency_symbol"] == "₹"
+    assert settings["field_pain_point"]
+    assert settings["field_before"]
+    assert settings["field_after"]
+    assert settings["field_useful_moments"]
+    assert settings["field_changed_after_feedback"]
     assert get_low_stock_threshold(db_path) == 5
 
 

tests/test_ui_audio_flow.py

@@ -204,6 +204,40 @@ def test_parse_note_for_editing_populates_review_fields(monkeypatch) -> None:
     assert result[8] == 240
 
 
+def test_price_memory_fills_missing_sale_price(tmp_path, monkeypatch) -> None:
+    db_path = tmp_path / "voiceledger.sqlite3"
+    gradio_app.add_transaction(gradio_app.local_parse_transaction("Sold 12 mangoes, 20 each"), db_path)
+    monkeypatch.setattr(
+        gradio_app.modal_api,
+        "parse_transaction_result",
+        lambda text, fallback, **kwargs: ParseResult(
+            transaction=fallback(text),
+            source="local",
+            message="Parsed locally with the rule parser.",
+        ),
+    )
+
+    result = gradio_app._parse_note_for_editing("Sold 5 mangoes", db_path)
+
+    assert result[0]["unit_price"] == 20
+    assert result[0]["amount"] == 100
+    assert result[7] == 20
+    assert result[8] == 100
+    assert "Price memory used" in result[3]
+
+
+def test_price_memory_does_not_overwrite_explicit_price(tmp_path) -> None:
+    db_path = tmp_path / "voiceledger.sqlite3"
+    gradio_app.add_transaction(gradio_app.local_parse_transaction("Sold 12 mangoes, 20 each"), db_path)
+    transaction = gradio_app.local_parse_transaction("Sold 5 mangoes, 30 each")
+
+    updated, used = gradio_app.suggest_price_from_history(transaction, db_path)
+
+    assert used is False
+    assert updated.unit_price == 30
+    assert updated.amount == 150
+
+
 def test_apply_review_edits_updates_payload_and_review_card(tmp_path) -> None:
     db_path = tmp_path / "voiceledger.sqlite3"
     parsed = gradio_app.local_parse_transaction("Sold 12 mangoes, 20 each").model_dump()
@@ -287,6 +321,37 @@ def test_customer_followup_and_reorder_helpers(tmp_path) -> None:
     assert "Onions" in message
 
 
+def test_debt_reminder_queue_sorts_positive_balances(tmp_path) -> None:
+    db_path = tmp_path / "voiceledger.sqlite3"
+    gradio_app.add_transaction(gradio_app.local_parse_transaction("Amit owes 100"), db_path)
+    gradio_app.add_transaction(gradio_app.local_parse_transaction("Ramesh owes 250"), db_path)
+    gradio_app.add_transaction(gradio_app.local_parse_transaction("Amit paid 40"), db_path)
+
+    queue = gradio_app.get_debt_reminder_queue(db_path)
+    reminder = gradio_app.generate_debt_reminder("Ramesh", db_path)
+
+    assert list(queue["customer"]) == ["Ramesh", "Amit"]
+    assert "₹250" in queue.iloc[0]["suggested_message"]
+    assert "Ramesh" in reminder
+
+
+def test_reorder_intelligence_flags_low_and_fast_stock(tmp_path) -> None:
+    db_path = tmp_path / "voiceledger.sqlite3"
+    gradio_app.update_business_settings(low_stock_threshold=5, db_path=db_path)
+    gradio_app.add_transaction(gradio_app.local_parse_transaction("Bought 3 onions"), db_path)
+    gradio_app.add_transaction(gradio_app.local_parse_transaction("Bought 20 mangoes"), db_path)
+    gradio_app.add_transaction(gradio_app.local_parse_transaction("Sold 12 mangoes, 20 each"), db_path)
+
+    recommendations = gradio_app.generate_reorder_intelligence(db_path)
+    table, message = gradio_app._generate_reorder_intelligence_for_ui(db_path)
+
+    statuses = dict(zip(recommendations["item"], recommendations["status"], strict=True))
+    assert statuses["onions"] == "Low stock"
+    assert statuses["mangoes"] == "Selling fast"
+    assert "recent_sold" in table.columns
+    assert "Mangoes" in message
+
+
 def test_currency_presets_and_voice_commands(tmp_path) -> None:
     db_path = tmp_path / "voiceledger.sqlite3"
     gradio_app.add_transaction(gradio_app.local_parse_transaction("Amit owes 100"), db_path)
@@ -326,3 +391,40 @@ def test_field_test_evidence_persists_checklist_and_notes(tmp_path) -> None:
     assert "2 checklist item" in status
     assert settings["field_test_who"] == "Local snack seller"
     assert gradio_app._field_test_checklist_values(settings) == ["Record sale", "Export report"]
+
+
+def test_field_test_evidence_persists_real_user_details(tmp_path) -> None:
+    db_path = tmp_path / "voiceledger.sqlite3"
+
+    status = gradio_app._save_field_test_evidence(
+        ["Record sale"],
+        "Local seller",
+        "Voice note",
+        "Old change note",
+        db_path,
+        pain_point="Paper notes get lost",
+        before="Used memory",
+        after="Checks dashboard",
+        useful="Debt reminders",
+        changed_after_feedback="Added price memory",
+    )
+    settings = gradio_app.get_business_settings(db_path)
+
+    assert "Paper notes get lost" in status
+    assert settings["field_pain_point"] == "Paper notes get lost"
+    assert settings["field_before"] == "Used memory"
+    assert settings["field_after"] == "Checks dashboard"
+    assert settings["field_useful_moments"] == "Debt reminders"
+    assert settings["field_changed_after_feedback"] == "Added price memory"
+
+
+def test_guided_demo_status_reflects_progress(tmp_path) -> None:
+    db_path = tmp_path / "voiceledger.sqlite3"
+
+    empty_status = gradio_app.get_guided_demo_status(db_path)
+    gradio_app.add_transaction(gradio_app.local_parse_transaction("Sold 12 mangoes, 20 each"), db_path)
+    progressed_status = gradio_app.get_guided_demo_status(db_path)
+
+    assert "Guided Judge Mode" in empty_status
+    assert "Seed Demo Data" in progressed_status
+    assert "Done" in progressed_status

voiceledger/ledger/settings.py

@@ -24,6 +24,11 @@ DEFAULT_BUSINESS_SETTINGS = {
     "field_test_tried": "Voice notes, typed notes, customer credit, inventory updates, dashboard review, and exports.",
     "field_test_changed": "Added review warnings, save receipts, edit/delete, CSV export, and clearer demo health.",
     "field_test_checklist": "Record sale,Record expense,Customer owes,Customer paid,Bought stock,Review dashboard,Export report",
+    "field_pain_point": "Sales, customer dues, and stock are written in scattered notes during a busy selling day.",
+    "field_before": "The seller checked memory, WhatsApp messages, and paper notes before knowing profit or dues.",
+    "field_after": "VoiceLedger turns short notes into sales, credit, inventory, and closeout reports in one place.",
+    "field_useful_moments": "Voice sale capture, customer balance follow-up, low-stock alerts, and WhatsApp summary.",
+    "field_changed_after_feedback": "Made review cards clearer, added correction logging, and made exports easier to find.",
 }
 
 

voiceledger/ui/gradio_app.py

@@ -79,6 +79,7 @@ CURRENCY_PRESETS = {
     "Brazil - BRL (R$)": "R$",
     "Custom": "",
 }
+PRICE_MEMORY_BADGE = "Price memory used"
 DEMO_NOTES = [
     "Bought 60 mangoes",
     "Sold 12 mangoes, 20 each",
@@ -131,6 +132,8 @@ def create_app(db_path: str | Path | None = None) -> gr.Blocks:
                     )
                 )
                 gr.HTML(_first_run_onboarding_panel())
+                guided_demo_output = gr.HTML(get_guided_demo_status(db_path))
+                refresh_guided_demo_button = gr.Button("Refresh Guided Judge Mode")
                 command_center_output = gr.HTML(_command_center(db_path))
                 gr.HTML(_section_heading("Seller Setup"))
                 with gr.Row():
@@ -199,6 +202,11 @@ def create_app(db_path: str | Path | None = None) -> gr.Blocks:
                     wrap=True,
                     elem_classes="vl-panel",
                 )
+                refresh_guided_demo_button.click(
+                    fn=lambda: get_guided_demo_status(db_path),
+                    inputs=None,
+                    outputs=guided_demo_output,
+                )
                 gr.HTML(
                     _info_panel(
                         "Workflow",
@@ -490,6 +498,32 @@ def create_app(db_path: str | Path | None = None) -> gr.Blocks:
                         lines=3,
                         elem_classes="vl-panel",
                     )
+                with gr.Row():
+                    field_pain_point_input = gr.Textbox(
+                        label="Pain point",
+                        value=initial_settings["field_pain_point"],
+                        lines=3,
+                        elem_classes="vl-panel",
+                    )
+                    field_before_input = gr.Textbox(
+                        label="Before VoiceLedger",
+                        value=initial_settings["field_before"],
+                        lines=3,
+                        elem_classes="vl-panel",
+                    )
+                with gr.Row():
+                    field_after_input = gr.Textbox(
+                        label="After VoiceLedger",
+                        value=initial_settings["field_after"],
+                        lines=3,
+                        elem_classes="vl-panel",
+                    )
+                    field_useful_input = gr.Textbox(
+                        label="Useful moments",
+                        value=initial_settings["field_useful_moments"],
+                        lines=3,
+                        elem_classes="vl-panel",
+                    )
                 save_field_notes_button = gr.Button("Save Field Test Notes")
                 field_notes_status_output = gr.Markdown(elem_classes="vl-status")
                 seed_demo_button = gr.Button("Seed Demo Transactions", variant="primary")
@@ -536,6 +570,38 @@ def create_app(db_path: str | Path | None = None) -> gr.Blocks:
                         lines=4,
                         elem_classes="vl-panel",
                     )
+                with gr.Row():
+                    field_test_pain_point_input = gr.Textbox(
+                        label="Pain point",
+                        value=initial_settings["field_pain_point"],
+                        lines=3,
+                        elem_classes="vl-panel",
+                    )
+                    field_test_before_input = gr.Textbox(
+                        label="Before VoiceLedger",
+                        value=initial_settings["field_before"],
+                        lines=3,
+                        elem_classes="vl-panel",
+                    )
+                with gr.Row():
+                    field_test_after_input = gr.Textbox(
+                        label="After VoiceLedger",
+                        value=initial_settings["field_after"],
+                        lines=3,
+                        elem_classes="vl-panel",
+                    )
+                    field_test_useful_input = gr.Textbox(
+                        label="Useful moments",
+                        value=initial_settings["field_useful_moments"],
+                        lines=3,
+                        elem_classes="vl-panel",
+                    )
+                field_test_changed_after_feedback_input = gr.Textbox(
+                    label="Changed after feedback",
+                    value=initial_settings["field_changed_after_feedback"],
+                    lines=3,
+                    elem_classes="vl-panel",
+                )
                 save_field_test_button = gr.Button("Save Field Test Evidence", variant="primary")
                 field_test_status_output = gr.Markdown(
                     value=_field_test_summary(initial_settings),
@@ -650,6 +716,21 @@ def create_app(db_path: str | Path | None = None) -> gr.Blocks:
                     inputs=followup_customer_name,
                     outputs=followup_output,
                 )
+                gr.HTML(_section_heading("Debt Reminder Queue"))
+                refresh_debt_queue_button = gr.Button("Refresh Reminder Queue")
+                debt_queue_output = gr.Dataframe(
+                    value=get_debt_reminder_queue(db_path),
+                    headers=["customer", "outstanding_balance", "suggested_message"],
+                    label="Customers to follow up",
+                    interactive=False,
+                    wrap=True,
+                    elem_classes="vl-panel",
+                )
+                refresh_debt_queue_button.click(
+                    fn=lambda: get_debt_reminder_queue(db_path),
+                    inputs=None,
+                    outputs=debt_queue_output,
+                )
 
         with gr.Column(visible=False, elem_classes="vl-page-section") as inventory_page:
                 gr.HTML('<div id="vl-page-inventory" class="vl-page-anchor"></div>')
@@ -697,8 +778,8 @@ def create_app(db_path: str | Path | None = None) -> gr.Blocks:
                 gr.HTML(_section_heading("Inventory Reorder List"))
                 reorder_button = gr.Button("Generate Reorder List")
                 reorder_output = gr.Dataframe(
-                    headers=["item", "current_stock", "suggested_action"],
-                    label="Low-stock reorder list",
+                    headers=["item", "current_stock", "recent_sold", "threshold", "status", "suggested_action"],
+                    label="Inventory reorder intelligence",
                     interactive=False,
                     wrap=True,
                     elem_classes="vl-panel",
@@ -710,7 +791,7 @@ def create_app(db_path: str | Path | None = None) -> gr.Blocks:
                     elem_classes="vl-panel",
                 )
                 reorder_button.click(
-                    fn=lambda: generate_reorder_list(db_path),
+                    fn=lambda: _generate_reorder_intelligence_for_ui(db_path),
                     inputs=None,
                     outputs=[reorder_output, reorder_message_output],
                 )
@@ -1049,13 +1130,51 @@ def create_app(db_path: str | Path | None = None) -> gr.Blocks:
             ],
         )
         save_field_notes_button.click(
-            fn=lambda who, tried, changed: _save_field_notes(who, tried, changed, db_path),
-            inputs=[field_who_input, field_tried_input, field_changed_input],
+            fn=lambda who, tried, changed, pain_point, before, after, useful: _save_field_notes(
+                who,
+                tried,
+                changed,
+                db_path,
+                pain_point=pain_point,
+                before=before,
+                after=after,
+                useful=useful,
+            ),
+            inputs=[
+                field_who_input,
+                field_tried_input,
+                field_changed_input,
+                field_pain_point_input,
+                field_before_input,
+                field_after_input,
+                field_useful_input,
+            ],
             outputs=field_notes_status_output,
         )
         save_field_test_button.click(
-            fn=lambda checklist, who, tried, changed: _save_field_test_evidence(checklist, who, tried, changed, db_path),
-            inputs=[field_test_checklist, field_test_who_input, field_test_tried_input, field_test_changed_input],
+            fn=lambda checklist, who, tried, changed, pain_point, before, after, useful, changed_after_feedback: _save_field_test_evidence(
+                checklist,
+                who,
+                tried,
+                changed,
+                db_path,
+                pain_point=pain_point,
+                before=before,
+                after=after,
+                useful=useful,
+                changed_after_feedback=changed_after_feedback,
+            ),
+            inputs=[
+                field_test_checklist,
+                field_test_who_input,
+                field_test_tried_input,
+                field_test_changed_input,
+                field_test_pain_point_input,
+                field_test_before_input,
+                field_test_after_input,
+                field_test_useful_input,
+                field_test_changed_after_feedback_input,
+            ],
             outputs=field_test_status_output,
         )
         record_health_button.click(
@@ -1365,8 +1484,11 @@ def _parse_note(
         force_local=_force_local_mode(ai_mode),
     )
     transaction = result.transaction
+    transaction, price_memory_used = suggest_price_from_history(transaction, db_path)
     payload = transaction.model_dump()
     warnings = _review_warnings(transaction, db_path)
+    if price_memory_used:
+        warnings.append(PRICE_MEMORY_BADGE)
     status = _status_message(transaction, result.message, result.fallback_reason, warnings=warnings)
     return payload, payload, status, _review_card(transaction, result.message, warnings)
 
@@ -1404,8 +1526,11 @@ def _transcribe_and_parse_audio(
         force_local=force_local,
     )
     transaction = parse_result.transaction
+    transaction, price_memory_used = suggest_price_from_history(transaction, db_path)
     payload = transaction.model_dump()
     warnings = _review_warnings(transaction, db_path)
+    if price_memory_used:
+        warnings.append(PRICE_MEMORY_BADGE)
     status = _status_message(
         transaction,
         parse_result.message,
@@ -1702,13 +1827,27 @@ def _save_seller_setup_and_refresh(
     )
 
 
-def _save_field_notes(who: str, tried: str, changed: str, db_path: str | Path | None) -> str:
+def _save_field_notes(
+    who: str,
+    tried: str,
+    changed: str,
+    db_path: str | Path | None,
+    *,
+    pain_point: str | None = None,
+    before: str | None = None,
+    after: str | None = None,
+    useful: str | None = None,
+) -> str:
     """Persist anonymized field-test notes."""
     update_business_settings(
         db_path=db_path,
         field_test_who=who,
         field_test_tried=tried,
         field_test_changed=changed,
+        field_pain_point=pain_point,
+        field_before=before,
+        field_after=after,
+        field_useful_moments=useful,
     )
     return "Saved anonymized field-test notes."
 
@@ -1719,6 +1858,12 @@ def _save_field_test_evidence(
     tried: str,
     changed: str,
     db_path: str | Path | None,
+    *,
+    pain_point: str | None = None,
+    before: str | None = None,
+    after: str | None = None,
+    useful: str | None = None,
+    changed_after_feedback: str | None = None,
 ) -> str:
     """Persist field-test checklist and notes."""
     checklist_value = ",".join(checklist or [])
@@ -1728,6 +1873,11 @@ def _save_field_test_evidence(
         field_test_tried=tried,
         field_test_changed=changed,
         field_test_checklist=checklist_value,
+        field_pain_point=pain_point,
+        field_before=before,
+        field_after=after,
+        field_useful_moments=useful,
+        field_changed_after_feedback=changed_after_feedback,
     )
     return _field_test_summary(settings)
 
@@ -1743,7 +1893,8 @@ def _field_test_summary(settings: dict[str, str]) -> str:
     completed = _field_test_checklist_values(settings)
     return (
         f"Field test evidence saved: {len(completed)} checklist item(s). "
-        f"Who: {settings['field_test_who']} Changed: {settings['field_test_changed']}"
+        f"Who: {settings['field_test_who']} Pain point: {settings['field_pain_point']} "
+        f"Changed: {settings['field_changed_after_feedback']}"
     )
 
 
@@ -1869,6 +2020,147 @@ def _seller_setup_status(settings: dict[str, str]) -> str:
     )
 
 
+def suggest_price_from_history(transaction: Transaction, db_path: str | Path | None) -> tuple[Transaction, bool]:
+    """Fill missing sale price fields from the latest saved sale for the same item."""
+    if (
+        transaction.transaction_type != "sale"
+        or not transaction.item
+        or transaction.quantity is None
+        or transaction.unit_price is not None
+        or transaction.amount is not None
+    ):
+        return transaction, False
+
+    ledger = get_transactions(db_path)
+    if ledger.empty:
+        return transaction, False
+
+    item_name = transaction.item.lower()
+    matches = ledger[
+        (ledger["transaction_type"] == "sale")
+        & (ledger["item"].fillna("").astype(str).str.lower() == item_name)
+    ].copy()
+    if matches.empty:
+        return transaction, False
+
+    matches["unit_price"] = pd.to_numeric(matches["unit_price"], errors="coerce")
+    matches = matches[matches["unit_price"].notna() & (matches["unit_price"] > 0)]
+    if matches.empty:
+        return transaction, False
+
+    unit_price = float(matches.iloc[0]["unit_price"])
+    payload = transaction.model_dump()
+    payload["unit_price"] = unit_price
+    payload["amount"] = round(float(transaction.quantity) * unit_price, 2)
+    payload["confidence"] = max(float(transaction.confidence), 0.78)
+    return Transaction(**payload), True
+
+
+def get_debt_reminder_queue(db_path: str | Path | None = None) -> pd.DataFrame:
+    """Return customers with positive outstanding balances sorted by balance."""
+    balances = get_customer_balances(db_path)
+    columns = ["customer", "outstanding_balance", "suggested_message"]
+    if balances.empty:
+        return pd.DataFrame(columns=columns)
+
+    debt = balances.copy()
+    debt["outstanding_balance"] = pd.to_numeric(debt["outstanding_balance"], errors="coerce").fillna(0)
+    debt = debt[debt["outstanding_balance"] > 0].sort_values("outstanding_balance", ascending=False)
+    if debt.empty:
+        return pd.DataFrame(columns=columns)
+
+    debt["suggested_message"] = debt["customer"].apply(lambda customer: generate_debt_reminder(str(customer), db_path))
+    return debt[columns].reset_index(drop=True)
+
+
+def generate_debt_reminder(customer_name: str, db_path: str | Path | None = None) -> str:
+    """Generate a WhatsApp-ready reminder for one customer's outstanding balance."""
+    return generate_customer_followup(customer_name, db_path)
+
+
+def generate_reorder_intelligence(db_path: str | Path | None = None) -> pd.DataFrame:
+    """Return stock recommendations with sales velocity context."""
+    columns = ["item", "current_stock", "recent_sold", "threshold", "status", "suggested_action"]
+    inventory = get_inventory(db_path)
+    if inventory.empty:
+        return pd.DataFrame(columns=columns)
+
+    threshold = get_low_stock_threshold(db_path)
+    ledger = get_transactions(db_path)
+    stock = inventory.copy()
+    stock["current_stock"] = pd.to_numeric(stock["current_stock"], errors="coerce").fillna(0)
+    sales_by_item: dict[str, float] = {}
+    if not ledger.empty:
+        sale_rows = ledger[ledger["transaction_type"] == "sale"].copy()
+        sale_rows["quantity"] = pd.to_numeric(sale_rows["quantity"], errors="coerce").fillna(0)
+        sale_rows["item_key"] = sale_rows["item"].fillna("").astype(str).str.lower()
+        sales_by_item = sale_rows.groupby("item_key")["quantity"].sum().to_dict()
+
+    records: list[dict[str, Any]] = []
+    for _, row in stock.iterrows():
+        item = str(row["item"])
+        current_stock = float(row["current_stock"])
+        recent_sold = float(sales_by_item.get(item.lower(), 0.0))
+        status = "Stock ok"
+        if current_stock < threshold:
+            status = "Low stock"
+        elif recent_sold > 0 and current_stock <= max(threshold * 2, recent_sold):
+            status = "Selling fast"
+        if status == "Low stock":
+            action = f"Restock {item} now; stock is below {_format_quantity(threshold)}."
+        elif status == "Selling fast":
+            action = f"Plan {item} restock soon; recent sales are {_format_quantity(recent_sold)}."
+        else:
+            action = f"{item} stock is ok."
+        records.append(
+            {
+                "item": item,
+                "current_stock": current_stock,
+                "recent_sold": recent_sold,
+                "threshold": threshold,
+                "status": status,
+                "suggested_action": action,
+            }
+        )
+
+    recommendations = pd.DataFrame.from_records(records, columns=columns)
+    recommendations["_status_rank"] = recommendations["status"].map({"Low stock": 0, "Selling fast": 1, "Stock ok": 2})
+    return recommendations.sort_values(["_status_rank", "current_stock", "item"]).drop(columns=["_status_rank"]).reset_index(drop=True)
+
+
+def get_guided_demo_status(db_path: str | Path | None = None) -> str:
+    """Return compact progress states for the judge demo flow."""
+    ledger = get_transactions(db_path)
+    has_transactions = not ledger.empty
+    has_sale = has_transactions and bool((ledger["transaction_type"] == "sale").any())
+    has_closeout_exports = Path(tempfile.gettempdir(), "voiceledger_transactions.csv").exists()
+    health_ready = initialize_database(db_path).exists()
+    steps = [
+        ("Check Health", health_ready, "Demo Health is reachable from this page."),
+        ("Seed Demo Data", has_transactions, "Adds realistic sales, expenses, credit, and stock."),
+        ("Record/Type Sale", has_sale, "Try Sold 12 mangoes, 20 each."),
+        ("Save", has_transactions, "Receipt, ledger, dashboard, credit, and stock refresh."),
+        ("Open Dashboard", has_transactions, "Metrics and timeline are ready once data exists."),
+        ("Open Reports", has_closeout_exports, "Daily Closeout prepares PDF, WhatsApp, and CSV."),
+    ]
+    items = "".join(
+        f"""
+        <li class="{'vl-guided-done' if done else 'vl-guided-next'}">
+          <strong>{escape(label)}</strong>
+          <span>{'Done' if done else 'Next'} · {escape(help_text)}</span>
+        </li>
+        """
+        for label, done, help_text in steps
+    )
+    return f"""
+    <section class="vl-guided-mode">
+      <h2>Guided Judge Mode</h2>
+      <p>Use this path for a 90-second demo: health, seed, record, save, dashboard, reports.</p>
+      <ol>{items}</ol>
+    </section>
+    """
+
+
 def _last_transaction_label(db_path: str | Path | None) -> str:
     """Return a concise label for the newest transaction."""
     ledger = get_transactions(db_path)
@@ -2048,6 +2340,29 @@ def _run_voice_command(command: str | None, db_path: str | Path | None) -> str:
     return _empty_detail_card("Command result", "Command not recognized. Try close today, show Amit, stock mangoes, or a transaction note.")
 
 
+def _generate_reorder_intelligence_for_ui(db_path: str | Path | None) -> tuple[pd.DataFrame, str]:
+    """Return reorder intelligence and a WhatsApp-ready restock message."""
+    recommendations = generate_reorder_intelligence(db_path)
+    if recommendations.empty:
+        return recommendations, "No inventory recorded yet."
+
+    settings = get_business_settings(db_path)
+    message_lines = [
+        f"{settings['business_name']} Reorder Plan",
+        f"Low-stock threshold: {_format_quantity(get_low_stock_threshold(db_path))}",
+        "",
+    ]
+    action_rows = recommendations[recommendations["status"].isin(["Low stock", "Selling fast"])]
+    if action_rows.empty:
+        message_lines.append("All stock looks ok today.")
+    else:
+        message_lines.extend(
+            f"- {row['item'].title()}: {row['status']} ({_format_quantity(row['current_stock'])} left, {_format_quantity(row['recent_sold'])} sold recently)"
+            for _, row in action_rows.head(8).iterrows()
+        )
+    return recommendations, "\n".join(message_lines)
+
+
 def _generate_daily_summary_report(db_path: str | Path | None) -> tuple[str | None, str]:
     """Generate the Daily Summary PDF for download in Gradio."""
     settings = get_business_settings(db_path)
@@ -2353,6 +2668,8 @@ def _review_readiness(transaction: Transaction, warnings: list[str]) -> str:
 
 def _warning_badge_class(warning: str) -> str:
     """Return a severity-aware warning badge class."""
+    if warning == PRICE_MEMORY_BADGE:
+        return "vl-success-badge"
     if warning in {"Inventory would go negative", "Possible duplicate"}:
         return "vl-warning-badge vl-warning-strong"
     return "vl-warning-badge"

voiceledger/ui/theme.py

@@ -316,6 +316,7 @@ body,
 .vl-closeout-card,
 .vl-pipeline-strip,
 .vl-small-model-card,
+.vl-guided-mode,
 .vl-command-center,
 .vl-seller-timeline,
 .vl-ai-mode-card,
@@ -337,6 +338,7 @@ body,
 .vl-closeout-card *,
 .vl-pipeline-strip *,
 .vl-small-model-card *,
+.vl-guided-mode *,
 .vl-command-center *,
 .vl-seller-timeline *,
 .vl-ai-mode-card *,
@@ -354,6 +356,7 @@ body,
 .vl-closeout-card h2,
 .vl-pipeline-strip h2,
 .vl-small-model-card h2,
+.vl-guided-mode h2,
 .vl-command-center h2,
 .vl-seller-timeline h2,
 .vl-ai-mode-card h2,
@@ -424,6 +427,56 @@ body,
   margin: 4px 0 0 !important;
 }
 
+.vl-guided-mode p {
+  color: #475247 !important;
+  font-size: 14px !important;
+  line-height: 1.4 !important;
+  margin: 4px 0 12px !important;
+}
+
+.vl-guided-mode ol {
+  display: grid !important;
+  gap: 10px !important;
+  grid-template-columns: repeat(auto-fit, minmax(170px, 1fr)) !important;
+  list-style: none !important;
+  margin: 0 !important;
+  padding: 0 !important;
+}
+
+.vl-guided-mode li {
+  background: #f6f7f2 !important;
+  border: 1px solid var(--vl-border) !important;
+  border-radius: 10px !important;
+  padding: 10px !important;
+}
+
+.vl-guided-mode li strong,
+.vl-guided-mode li span {
+  display: block !important;
+}
+
+.vl-guided-mode li strong {
+  font-size: 14px !important;
+  font-weight: 900 !important;
+}
+
+.vl-guided-mode li span {
+  color: #475247 !important;
+  font-size: 12px !important;
+  line-height: 1.35 !important;
+  margin-top: 4px !important;
+}
+
+.vl-guided-done {
+  border-color: #9bd2ad !important;
+  background: #e7f6ed !important;
+}
+
+.vl-guided-next {
+  border-color: #e3bc5b !important;
+  background: #fff9e8 !important;
+}
+
 .vl-warning-row {
   display: flex !important;
   flex-wrap: wrap !important;