feat: persistent pantry on HF Spaces /data, TTS wired, clean pipeline

app.py

@@ -11,15 +11,18 @@ from PIL import Image as PILImage
 
 sys.path.append(os.path.dirname(__file__))
 from utils import (
-    estimate_expiry, format_parsed_items, get_pantry,
-    mark_used, save_to_db, expiry_status
+    estimate_expiry, format_parsed_items,
+    save_to_db, expiry_status, init_db
 )
 
+# Use HF Spaces persistent storage if available
+DB_PATH = "/data/pantry.db" if os.path.exists("/data") else "db/pantry.db"
+
 CHARACTERS = {
-    "Grandma (Ammamma)": "You are Ammamma, a warm but judgemental grandma who loves cooking. Speak in English only. Scold gently about food waste but show love.",
-    "Chef": "You are a sharp professional chef. Direct, technically precise, slightly impatient with bad ingredients, but brilliant at making something from nothing.",
-    "Fitness Coach": "You are an enthusiastic fitness coach. Every recipe must be high protein, low waste, macro-balanced. You relate everything back to gains and clean eating.",
-    "Food Critic": "You are a pompous but secretly warm food critic. You critique the ingredients dramatically before reluctantly producing a brilliant recipe.",
+    "Grandma (Ammamma)": "Warm loving grandma who hates food waste.",
+    "Chef": "Sharp professional chef. Direct and brilliant.",
+    "Fitness Coach": "Enthusiastic fitness coach obsessed with gains.",
+    "Food Critic": "Pompous but secretly warm food critic.",
 }
 
 CUISINES = ["South Indian", "North Indian", "Italian", "Mediterranean", "East Asian", "Surprise me"]
@@ -32,6 +35,56 @@ CHAR_EMOJI = {
 }
 
 
+def get_pantry_display():
+    init_db(DB_PATH)
+    conn = sqlite3.connect(DB_PATH)
+    c = conn.cursor()
+    c.execute("SELECT item_name, quantity, estimated_expiry FROM pantry WHERE used=0 ORDER BY estimated_expiry ASC")
+    rows = c.fetchall()
+    conn.close()
+
+    if not rows:
+        return "Your pantry is empty — scan a receipt to get started."
+
+    red = [r for r in rows if expiry_status(r[2])[1] <= 1]
+    amber = [r for r in rows if 1 < expiry_status(r[2])[1] <= 5]
+    green = [r for r in rows if expiry_status(r[2])[1] > 5]
+
+    lines = []
+    lines.append(f"📦 {len(rows)} items in pantry")
+    lines.append(f"🔴 {len(red)} expiring today  🟡 {len(amber)} expiring soon  🟢 {len(green)} fine")
+    lines.append("─" * 40)
+
+    if red:
+        lines.append("\n🔴 COOK TODAY:")
+        for r in red:
+            lines.append(f"  • {r[0]} (x{r[1]})")
+
+    if amber:
+        lines.append("\n🟡 USE THIS WEEK:")
+        for r in amber:
+            _, days = expiry_status(r[2])
+            lines.append(f"  • {r[0]} (x{r[1]}) — {days} days left")
+
+    if green:
+        lines.append("\n🟢 ALL GOOD:")
+        for r in green:
+            _, days = expiry_status(r[2])
+            lines.append(f"  • {r[0]} (x{r[1]}) — {days} days")
+
+    return "\n".join(lines)
+
+
+def mark_used_local(item_name):
+    init_db(DB_PATH)
+    conn = sqlite3.connect(DB_PATH)
+    c = conn.cursor()
+    c.execute("UPDATE pantry SET used=1 WHERE item_name=? AND used=0", (item_name,))
+    conn.commit()
+    conn.close()
+    return get_pantry_display()
+
+
 def parse_receipt_gradio(image):
     if image is None:
         return "Please upload a receipt image.", []
@@ -55,7 +108,7 @@ def parse_receipt_gradio(image):
                 estimate_expiry(item["name"])
             )
 
-        save_to_db(result["items"], result.get("shop", "unknown"))
+        save_to_db(result["items"], result.get("shop", "unknown"), DB_PATH)
 
         return format_parsed_items(result["items"]), result["items"]
 
@@ -64,55 +117,58 @@ def parse_receipt_gradio(image):
 
 
 def generate_recipe_with_audio(character, cuisine, use_expiring):
-    conn = sqlite3.connect(
-        "/data/pantry.db" if os.path.exists("/data") else "db/pantry.db"
-    )
-    c = conn.cursor()
-    c.execute(
-        "SELECT item_name, quantity, estimated_expiry FROM pantry WHERE used=0 ORDER BY estimated_expiry ASC"
-    )
-    rows = c.fetchall()
-    conn.close()
+    try:
+        init_db(DB_PATH)
+        conn = sqlite3.connect(DB_PATH)
+        c = conn.cursor()
+        c.execute(
+            "SELECT item_name, quantity, estimated_expiry FROM pantry WHERE used=0 ORDER BY estimated_expiry ASC"
+        )
+        rows = c.fetchall()
+        conn.close()
 
-    if not rows:
-        return "Your pantry is empty — scan a receipt first.", None
+        if not rows:
+            return "Your pantry is empty — scan a receipt first.", None
 
-    all_items = [{"name": r[0], "quantity": r[1], "expiry": r[2]} for r in rows]
+        all_items = [{"name": r[0], "quantity": r[1], "expiry": r[2]} for r in rows]
 
-    for item in all_items:
-        _, days = expiry_status(item["expiry"])
-        item["days_left"] = days
+        for item in all_items:
+            _, days = expiry_status(item["expiry"])
+            item["days_left"] = days
 
-    expiring = [i for i in all_items if i["days_left"] <= 5]
-    expiring_names = [i["name"] for i in expiring]
+        expiring = [i for i in all_items if i["days_left"] <= 5]
+        expiring_names = [i["name"] for i in expiring]
 
-    if not expiring_names:
-        expiring_names = [all_items[0]["name"]] if all_items else []
+        if not expiring_names:
+            expiring_names = [all_items[0]["name"]] if all_items else []
 
-    dialogue_fn = modal.Function.from_name("zerowastekitchen", "generate_dialogue")
-    recipe_fn = modal.Function.from_name("zerowastekitchen", "generate_recipe_llm")
-    tts_fn = modal.Function.from_name("zerowastekitchen", "text_to_speech")
+        dialogue_fn = modal.Function.from_name("zerowastekitchen", "generate_dialogue")
+        recipe_fn = modal.Function.from_name("zerowastekitchen", "generate_recipe_llm")
+        tts_fn = modal.Function.from_name("zerowastekitchen", "text_to_speech")
 
-    with concurrent.futures.ThreadPoolExecutor() as executor:
-        dialogue_future = executor.submit(
-            dialogue_fn.remote, character, expiring_names[:5], cuisine
-        )
-        recipe_future = executor.submit(
-            recipe_fn.remote, all_items, expiring, cuisine
-        )
-        dialogue = dialogue_future.result()
-        recipe = recipe_future.result()
+        with concurrent.futures.ThreadPoolExecutor() as executor:
+            dialogue_future = executor.submit(
+                dialogue_fn.remote, character, expiring_names[:5], cuisine
+            )
+            recipe_future = executor.submit(
+                recipe_fn.remote, all_items, expiring, cuisine
+            )
+            dialogue = dialogue_future.result()
+            recipe = recipe_future.result()
+
+        audio_bytes = tts_fn.remote(dialogue, character)
 
-    audio_bytes = tts_fn.remote(dialogue, character)
+        emoji = CHAR_EMOJI.get(character, "🧓")
+        output = f"{emoji} {character} says:\n\"{dialogue}\"\n\n"
+        output += recipe
 
-    emoji = CHAR_EMOJI.get(character, "🧓")
-    output = f"{emoji} {character} says:\n\"{dialogue}\"\n\n"
-    output += recipe
+        buf = io.BytesIO(audio_bytes)
+        audio_np, sr = sf.read(buf)
 
-    buf = io.BytesIO(audio_bytes)
-    audio_np, sr = sf.read(buf)
+        return output, (sr, audio_np)
 
-    return output, (sr, audio_np)
+    except Exception as e:
+        return f"Error: {str(e)}", None
 
 
 with gr.Blocks(title="ZeroWasteKitchen", theme=gr.themes.Soft()) as app:
@@ -153,7 +209,7 @@ with gr.Blocks(title="ZeroWasteKitchen", theme=gr.themes.Soft()) as app:
                 label="Pantry status",
                 lines=15,
                 interactive=False,
-                value=get_pantry()
+                value=get_pantry_display()
             )
             with gr.Row():
                 mark_item = gr.Textbox(
@@ -161,9 +217,9 @@ with gr.Blocks(title="ZeroWasteKitchen", theme=gr.themes.Soft()) as app:
                     placeholder="e.g. CURRIS LEAVES PACKET"
                 )
                 mark_btn = gr.Button("Mark as used", size="sm")
-            refresh_btn.click(fn=get_pantry, outputs=[pantry_output])
+            refresh_btn.click(fn=get_pantry_display, outputs=[pantry_output])
             mark_btn.click(
-                fn=mark_used,
+                fn=mark_used_local,
                 inputs=[mark_item],
                 outputs=[pantry_output]
             )

modal_services.py

@@ -2,37 +2,24 @@ import modal
 from modal import Volume
 from pathlib import Path
 
-# App definition
 app = modal.App("zerowastekitchen")
 
-# Volume for model storage
 model_volume = Volume.from_name("model-cache", create_if_missing=True)
 MODEL_DIR = Path("/models")
 
-# Download image - lightweight, just needs huggingface_hub
 download_image = (
     modal.Image.debian_slim()
     .pip_install("huggingface_hub", "hf_transfer")
     .env({"HF_HUB_ENABLE_HF_TRANSFER": "1"})
 )
 
-# Base image for OCR, expiry, dialogue, recipe
 base_image = (
     modal.Image.debian_slim()
     .pip_install(
-        "torch",
-        "torchvision",
-        "transformers>=5.7.0",
-        "accelerate",
-        "pillow",
-        "einops",
-        "sentencepiece",
-        "timm",
-        "open_clip_torch",
-        "av",
-        "numpy",
-        "huggingface_hub",
-        "hf_transfer"
+        "torch", "torchvision", "transformers>=5.7.0",
+        "accelerate", "pillow", "einops", "sentencepiece",
+        "timm", "open_clip_torch", "av", "numpy",
+        "huggingface_hub", "hf_transfer"
     )
     .env({"HF_HUB_ENABLE_HF_TRANSFER": "1"})
 )
@@ -51,7 +38,7 @@ tts_image = (
     )
 )
 
-# ── 0. DOWNLOAD MODELS ───────────────────────────────────────
+
 @app.function(
     image=download_image,
     volumes={str(MODEL_DIR): model_volume},
@@ -69,33 +56,22 @@ def download_models():
         local_dir=str(MODEL_DIR / "minicpm-v"),
         token=token
     )
-
     print("Downloading tiny-aya-fire...")
     snapshot_download(
         "CohereLabs/tiny-aya-fire",
         local_dir=str(MODEL_DIR / "tiny-aya-fire"),
         token=token
     )
-
     print("Downloading Nemotron 4B...")
     snapshot_download(
         "nvidia/Nemotron-Mini-4B-Instruct",
         local_dir=str(MODEL_DIR / "nemotron-4b"),
         token=token
     )
-
-    print("Downloading Magpie TTS...")
-    snapshot_download(
-        "nvidia/magpie_tts_multilingual_357m",
-        local_dir=str(MODEL_DIR / "magpie-tts"),
-        token=token
-    )
-
     model_volume.commit()
     print("All models downloaded.")
 
 
-# ── 1. RECEIPT OCR ──────────────────────────────────────────
 @app.function(
     gpu="T4",
     image=base_image,
@@ -175,7 +151,6 @@ Return as JSON only, no other text:
         return {"shop": "unknown", "items": [], "raw": str(e)}
 
 
-# ── 2. EXPIRY ESTIMATION ─────────────────────────────────────
 @app.function(
     gpu="T4",
     image=base_image,
@@ -231,7 +206,6 @@ Reply with a single integer only. No explanation.
     return expiry_map
 
 
-# ── 3. CHARACTER DIALOGUE ────────────────────────────────────
 @app.function(
     gpu="T4",
     image=base_image,
@@ -242,6 +216,7 @@ Reply with a single integer only. No explanation.
 )
 def generate_dialogue(character: str, expiring_items: list, cuisine: str) -> str:
     import torch
+    import re
     from transformers import AutoTokenizer, AutoModelForCausalLM
 
     tokenizer = AutoTokenizer.from_pretrained(str(MODEL_DIR / "tiny-aya-fire"))
@@ -252,11 +227,13 @@ def generate_dialogue(character: str, expiring_items: list, cuisine: str) -> str
     )
 
     character_prompts = {
-        "Grandma (Ammamma)": """You are Ammamma, a warm but dramatic grandma who loves cooking.
-    Speak in English only. Scold gently about food waste but show love. 2-3 sentences only. Be concise.""",
-        "Chef": "You are a sharp professional chef. Be direct and impatient but brilliant. 2-3 sentences only.",
-        "Fitness Coach": "You are an enthusiastic fitness coach obsessed with gains and clean eating. 2-3 sentences only.",
-        "Food Critic": "You are a pompous food critic who is secretly warm. Be dramatic. 2-3 sentences only.",
+        "Grandma (Ammamma)": """You are Ammamma, a warm loving grandma who hates food waste.
+Speak warmly in English. Mention the specific expiring items by name.
+Express urgency about using them before they go bad.
+Show love and excitement about cooking. 2-3 sentences only. Be specific and concise.""",
+        "Chef": "You are a sharp professional chef. Be direct and impatient but brilliant. Mention the expiring items. 2-3 sentences only.",
+        "Fitness Coach": "You are an enthusiastic fitness coach obsessed with gains and clean eating. Mention the expiring items. 2-3 sentences only.",
+        "Food Critic": "You are a pompous food critic who is secretly warm. Be dramatic about the expiring items. 2-3 sentences only.",
     }
 
     system = character_prompts.get(character, character_prompts["Grandma (Ammamma)"])
@@ -267,7 +244,7 @@ def generate_dialogue(character: str, expiring_items: list, cuisine: str) -> str
 <|user|>
 These grocery items are expiring soon: {items_str}
 We are cooking {cuisine} food today.
-Give your in-character reaction about the expiring food. 2-3 sentences maximum.
+Give your in-character reaction. 2-3 sentences maximum.
 <|assistant|>
 """
 
@@ -287,15 +264,25 @@ Give your in-character reaction about the expiring food. 2-3 sentences maximum.
         skip_special_tokens=True
     ).strip()
 
-    # Clean up leaked system prompt
+    # Clean up leaked content
     response = response.split("<|system|>")[0].strip()
     response = response.split("system|>")[0].strip()
+    response = response.split("<|user|>")[0].strip()
+    response = response.split("user|>")[0].strip()
     response = response.split("<|")[0].strip()
     response = response.split("##")[0].strip()
+    response = response.split("Can you")[0].strip()
+    response = response.split("Please")[0].strip()
+    response = response.split("Ammamma,")[0].strip()
+    response = re.sub(r'#\w*', '', response).strip()
+
+    # Limit to 3 sentences
+    sentences = [s.strip() for s in response.split(".") if s.strip()]
+    response = ". ".join(sentences[:3]) + "."
+
     return response
 
 
-# ── 4. RECIPE GENERATION ─────────────────────────────────────
 @app.function(
     gpu="T4",
     image=base_image,
@@ -343,7 +330,7 @@ STEPS:
 3. ...
 4. ...
 5. ...
-IMPORTANT: Maximum 5 steps. Stop here. Do not write more than 5 steps.
+Stop here. Do not write more than 5 steps.
 <|assistant|>
 """
 
@@ -363,7 +350,6 @@ IMPORTANT: Maximum 5 steps. Stop here. Do not write more than 5 steps.
         skip_special_tokens=True
     ).strip()
 
-    response = response.split("<|")[0].strip()
     response = response.split("##")[0].strip()
 
     # Force truncate to 5 steps
@@ -377,11 +363,10 @@ IMPORTANT: Maximum 5 steps. Stop here. Do not write more than 5 steps.
             if step_count >= 5:
                 break
 
-    response = "\n".join(truncated)    
+    response = "\n".join(truncated)
     return response
 
 
-# ── 5. TEXT TO SPEECH ────────────────────────────────────────
 @app.function(
     gpu="T4",
     image=tts_image,
@@ -393,13 +378,13 @@ IMPORTANT: Maximum 5 steps. Stop here. Do not write more than 5 steps.
 def text_to_speech(text: str, character: str) -> bytes:
     import torch
     import io
-    import soundfile as sf
     import os
+    import soundfile as sf
 
-    print("Downloading XTTS-v2...")
     os.environ["COQUI_TOS_AGREED"] = "1"
     from TTS.api import TTS
-    tts  = TTS("tts_models/multilingual/multi-dataset/xtts_v2")    
+
+    tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2")
 
     speaker_map = {
         "Grandma (Ammamma)": "Claribel Dervla",
@@ -409,14 +394,21 @@ def text_to_speech(text: str, character: str) -> bytes:
     }
     speaker = speaker_map.get(character, "Claribel Dervla")
 
-    # Keep short
+    # Deduplicate and limit sentences
     sentences = text.replace("!", ".").replace("?", ".").split(".")
-    short_text = ". ".join([s.strip() for s in sentences[:2] if s.strip()]) + "."
+    seen = []
+    for s in sentences:
+        s = s.strip()
+        if s and s not in seen:
+            seen.append(s)
+        if len(seen) >= 3:
+            break
+    short_text = ". ".join(seen) + "."
 
     wav = tts.tts(
-    text=short_text,
-    speaker=speaker,
-    language="en"
+        text=short_text,
+        speaker=speaker,
+        language="en"
     )
 
     buf = io.BytesIO()
@@ -426,16 +418,12 @@ def text_to_speech(text: str, character: str) -> bytes:
 
 @app.local_entrypoint()
 def main():
-    print("Downloading all models to volume...")
-    download_models.remote()
-    print("Done.")
-    # Test expiry estimation
     test_items = ["INDIA WALA PANEER", "CURRIS LEAVES PACKET", "BASMATI RICE"]
+
     print("Testing expiry estimation...")
     expiry_map = estimate_expiry_llm.remote(test_items)
     print(expiry_map)
 
-    # Test dialogue
     print("\nTesting Ammamma dialogue...")
     dialogue = generate_dialogue.remote(
         "Grandma (Ammamma)",
@@ -444,14 +432,12 @@ def main():
     )
     print(dialogue)
 
-    # Test recipe
     print("\nTesting recipe generation...")
     all_items = [{"name": i} for i in test_items]
     expiring = [{"name": i} for i in test_items[:2]]
     recipe = generate_recipe_llm.remote(all_items, expiring, "South Indian")
     print(recipe)
 
-    # Test TTS
     print("\nTesting TTS...")
     audio_bytes = text_to_speech.remote(dialogue, "Grandma (Ammamma)")
     with open("test_audio.wav", "wb") as f:

utils.py

@@ -51,9 +51,9 @@ def format_parsed_items(items):
         lines.append(f"{emoji} {item['name']} (qty: {item.get('quantity','1')}) — {label}")
     return "\n".join(lines)
 
-def init_db():
-    os.makedirs(os.path.dirname(DB_PATH), exist_ok=True) if os.path.dirname(DB_PATH) else None
-    conn = sqlite3.connect(DB_PATH)
+def init_db(db_path=DB_PATH):
+    os.makedirs(os.path.dirname(db_path) if os.path.dirname(db_path) else ".", exist_ok=True)
+    conn = sqlite3.connect(db_path)
     c = conn.cursor()
     c.execute('''CREATE TABLE IF NOT EXISTS pantry (
         id INTEGER PRIMARY KEY AUTOINCREMENT,
@@ -68,13 +68,11 @@ def init_db():
     conn.commit()
     conn.close()
 
-def save_to_db(items: list, shop: str):
-    init_db()
-    conn = sqlite3.connect(DB_PATH)
+def save_to_db(items: list, shop: str, db_path=DB_PATH):
+    init_db(db_path)
+    conn = sqlite3.connect(db_path)
     c = conn.cursor()
     for item in items:
-        # If item exists, update quantity and expiry
-        # If not, insert new row
         c.execute('''
             INSERT INTO pantry (item_name, quantity, price, purchase_date, estimated_expiry, shop)
             VALUES (?, ?, ?, ?, ?, ?)
@@ -82,8 +80,7 @@ def save_to_db(items: list, shop: str):
                 quantity = quantity || '+' || excluded.quantity,
                 estimated_expiry = excluded.estimated_expiry,
                 purchase_date = excluded.purchase_date
-        ''',
-        (
+        ''', (
             item["name"],
             item.get("quantity", "1"),
             item.get("price", ""),