Spaces:
Running
Running
Update main.py
Browse files
main.py
CHANGED
|
@@ -1,12 +1,12 @@
|
|
| 1 |
"""
|
| 2 |
-
main.py — Pricelyst Shopping Advisor (Jessica 2026
|
| 3 |
|
| 4 |
✅ Flask API
|
| 5 |
✅ Firebase Admin Persistence
|
| 6 |
-
✅ Gemini 2.5 Flash
|
| 7 |
✅ "Analyst Engine": Python Math for Baskets & ZESA
|
| 8 |
-
✅ "Creative Engine": 4-Step Hybrid Plans
|
| 9 |
-
✅
|
| 10 |
|
| 11 |
ENV VARS:
|
| 12 |
- GOOGLE_API_KEY=...
|
|
@@ -48,7 +48,6 @@ except Exception as e:
|
|
| 48 |
logger.error("google-genai not installed. pip install google-genai. Error=%s", e)
|
| 49 |
|
| 50 |
GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY", "")
|
| 51 |
-
# Updated to 2.5 Flash as requested for 2026 timeline
|
| 52 |
GEMINI_MODEL = os.environ.get("GEMINI_MODEL", "gemini-2.5-flash")
|
| 53 |
|
| 54 |
_gemini_client = None
|
|
@@ -168,18 +167,18 @@ def fetch_and_flatten_data() -> pd.DataFrame:
|
|
| 168 |
rows = []
|
| 169 |
for p in all_products:
|
| 170 |
try:
|
| 171 |
-
p_id = p.get("id")
|
| 172 |
-
p_name = p.get("name") or "Unknown"
|
| 173 |
clean_name = _norm(p_name)
|
| 174 |
|
| 175 |
cat_obj = p.get("category") or {}
|
| 176 |
-
cat_name = cat_obj.get("name") or "General"
|
| 177 |
|
| 178 |
brand_obj = p.get("brand") or {}
|
| 179 |
-
brand_name = brand_obj.get("brand_name") or ""
|
| 180 |
|
| 181 |
views = int(p.get("view_count") or 0)
|
| 182 |
-
image = p.get("thumbnail") or p.get("image")
|
| 183 |
|
| 184 |
prices = p.get("prices") or []
|
| 185 |
|
|
@@ -200,7 +199,7 @@ def fetch_and_flatten_data() -> pd.DataFrame:
|
|
| 200 |
|
| 201 |
for offer in prices:
|
| 202 |
retailer = offer.get("retailer") or {}
|
| 203 |
-
r_name = retailer.get("name") or "Unknown Store"
|
| 204 |
price_val = _coerce_price(offer.get("price"))
|
| 205 |
|
| 206 |
if price_val > 0:
|
|
@@ -268,6 +267,7 @@ def search_products_fuzzy(df: pd.DataFrame, query: str, limit: int = 10) -> pd.D
|
|
| 268 |
def calculate_basket_optimization(item_names: List[str]) -> Dict[str, Any]:
|
| 269 |
"""
|
| 270 |
Determines the best store for a list of items.
|
|
|
|
| 271 |
"""
|
| 272 |
df = get_market_index()
|
| 273 |
if df.empty:
|
|
@@ -289,12 +289,11 @@ def calculate_basket_optimization(item_names: List[str]) -> Dict[str, Any]:
|
|
| 289 |
# Best match based on popularity
|
| 290 |
best_prod = hits.iloc[0]
|
| 291 |
found_items.append({
|
| 292 |
-
"query": item,
|
| 293 |
-
"product_id": best_prod['product_id'],
|
| 294 |
-
"name": best_prod['product_name']
|
| 295 |
})
|
| 296 |
|
| 297 |
-
# Note: We return actionable=True even if products missing, so Gemini can estimate
|
| 298 |
if not found_items:
|
| 299 |
return {
|
| 300 |
"actionable": True,
|
|
@@ -319,10 +318,10 @@ def calculate_basket_optimization(item_names: List[str]) -> Dict[str, Any]:
|
|
| 319 |
found_names = [x['name'] for x in found_items if x['product_id'] in retailer_pids]
|
| 320 |
|
| 321 |
retailer_stats.append({
|
| 322 |
-
"retailer": retailer,
|
| 323 |
-
"total_price": float(total_price),
|
| 324 |
-
"item_count": found_count,
|
| 325 |
-
"coverage_percent": (found_count / len(found_items)) * 100,
|
| 326 |
"found_items": found_names
|
| 327 |
})
|
| 328 |
|
|
@@ -373,13 +372,13 @@ def calculate_zesa_units(amount_usd: float) -> Dict[str, Any]:
|
|
| 373 |
breakdown.append(f"All {bought:.1f}u @ ${t1['rate']}")
|
| 374 |
|
| 375 |
return {
|
| 376 |
-
"amount_usd": amount_usd,
|
| 377 |
-
"est_units_kwh": round(units, 1),
|
| 378 |
"breakdown": breakdown
|
| 379 |
}
|
| 380 |
|
| 381 |
# =========================
|
| 382 |
-
# 3. Gemini Helpers
|
| 383 |
# =========================
|
| 384 |
|
| 385 |
def gemini_detect_intent(transcript: str) -> Dict[str, Any]:
|
|
@@ -423,7 +422,7 @@ def gemini_generate_4step_plan(transcript: str, analyst_result: Dict) -> str:
|
|
| 423 |
USER CONTEXT: "{transcript}"
|
| 424 |
|
| 425 |
ANALYST DATA (REAL PRICES):
|
| 426 |
-
{json.dumps(analyst_result, indent=2)}
|
| 427 |
|
| 428 |
REQUIRED SECTIONS (Do not deviate):
|
| 429 |
|
|
@@ -432,7 +431,7 @@ def gemini_generate_4step_plan(transcript: str, analyst_result: Dict) -> str:
|
|
| 432 |
- Show product name, retailer, and exact price from data.
|
| 433 |
|
| 434 |
2. **Not in Catalogue (Estimates) 😔**
|
| 435 |
-
- List items from 'global_missing'
|
| 436 |
- Provide a realistic USD estimate for Zimbabwe.
|
| 437 |
|
| 438 |
3. **Totals 💰**
|
|
@@ -471,7 +470,7 @@ def health():
|
|
| 471 |
|
| 472 |
@app.post("/chat")
|
| 473 |
def chat():
|
| 474 |
-
"""Text Chat
|
| 475 |
body = request.get_json(silent=True) or {}
|
| 476 |
msg = body.get("message", "")
|
| 477 |
pid = body.get("profile_id")
|
|
@@ -489,13 +488,12 @@ def chat():
|
|
| 489 |
elif intent_data["intent"] == "PRODUCT_SEARCH" and intent_data.get("items"):
|
| 490 |
analyst_data = calculate_basket_optimization(intent_data["items"])
|
| 491 |
|
| 492 |
-
# Quick chat summary
|
| 493 |
reply = "I'm processing that for you."
|
| 494 |
if analyst_data:
|
| 495 |
try:
|
| 496 |
resp = _gemini_client.models.generate_content(
|
| 497 |
model=GEMINI_MODEL,
|
| 498 |
-
contents=f"Summarize this shopping data for chat:\n{json.dumps(analyst_data)}"
|
| 499 |
)
|
| 500 |
reply = resp.text
|
| 501 |
except: pass
|
|
@@ -541,7 +539,7 @@ def call_briefing():
|
|
| 541 |
top = df[df['is_offer']].sort_values('views', ascending=False).drop_duplicates('product_name').head(60)
|
| 542 |
lines = []
|
| 543 |
for _, r in top.iterrows():
|
| 544 |
-
lines.append(f"{r['product_name']} (
|
| 545 |
catalogue_str = ", ".join(lines)
|
| 546 |
|
| 547 |
kpi_snapshot = {
|
|
@@ -558,12 +556,7 @@ def call_briefing():
|
|
| 558 |
@app.post("/api/log-call-usage")
|
| 559 |
def log_call_usage():
|
| 560 |
"""
|
| 561 |
-
|
| 562 |
-
1. Update Memory.
|
| 563 |
-
2. Detect Intent.
|
| 564 |
-
3. Run Analyst.
|
| 565 |
-
4. Generate 4-Step Creative Plan.
|
| 566 |
-
5. Save.
|
| 567 |
"""
|
| 568 |
body = request.get_json(silent=True) or {}
|
| 569 |
pid = body.get("profile_id")
|
|
@@ -573,7 +566,6 @@ def log_call_usage():
|
|
| 573 |
|
| 574 |
logger.info(f"Log Call: Processing {pid}. Transcript Len: {len(transcript)}")
|
| 575 |
|
| 576 |
-
# 1. Update Memory
|
| 577 |
if len(transcript) > 20 and db:
|
| 578 |
try:
|
| 579 |
curr_mem = db.collection("pricelyst_profiles").document(pid).get().to_dict().get("memory_summary", "")
|
|
@@ -583,21 +575,16 @@ def log_call_usage():
|
|
| 583 |
except Exception as e:
|
| 584 |
logger.error(f"Memory Update Error: {e}")
|
| 585 |
|
| 586 |
-
# 2. Intent Detection
|
| 587 |
intent_data = gemini_detect_intent(transcript)
|
| 588 |
logger.info(f"Intent: {intent_data.get('intent')}")
|
| 589 |
|
| 590 |
plan_data = {}
|
| 591 |
|
| 592 |
-
# 3. Actionable Logic
|
| 593 |
if intent_data.get("actionable"):
|
| 594 |
|
| 595 |
-
# Run Analyst
|
| 596 |
if intent_data.get("items"):
|
| 597 |
analyst_result = calculate_basket_optimization(intent_data["items"])
|
| 598 |
|
| 599 |
-
# 4. Generate Creative Plan (4-Step)
|
| 600 |
-
# We generate if we have actionable basket results OR missing items
|
| 601 |
if analyst_result.get("actionable"):
|
| 602 |
md_content = gemini_generate_4step_plan(transcript, analyst_result)
|
| 603 |
|
|
@@ -615,7 +602,6 @@ def log_call_usage():
|
|
| 615 |
doc_ref.set(plan_data)
|
| 616 |
logger.info(f"Plan Saved: {doc_ref.id}")
|
| 617 |
|
| 618 |
-
# 5. Log Call
|
| 619 |
if db:
|
| 620 |
db.collection("pricelyst_profiles").document(pid).collection("call_logs").add({
|
| 621 |
"transcript": transcript,
|
|
|
|
| 1 |
"""
|
| 2 |
+
main.py — Pricelyst Shopping Advisor (Jessica Edition 2026)
|
| 3 |
|
| 4 |
✅ Flask API
|
| 5 |
✅ Firebase Admin Persistence
|
| 6 |
+
✅ Gemini 2.5 Flash
|
| 7 |
✅ "Analyst Engine": Python Math for Baskets & ZESA
|
| 8 |
+
✅ "Creative Engine": 4-Step Hybrid Plans
|
| 9 |
+
✅ Type Safety: Explicit Casting for JSON Serialization (Fixes int64 error)
|
| 10 |
|
| 11 |
ENV VARS:
|
| 12 |
- GOOGLE_API_KEY=...
|
|
|
|
| 48 |
logger.error("google-genai not installed. pip install google-genai. Error=%s", e)
|
| 49 |
|
| 50 |
GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY", "")
|
|
|
|
| 51 |
GEMINI_MODEL = os.environ.get("GEMINI_MODEL", "gemini-2.5-flash")
|
| 52 |
|
| 53 |
_gemini_client = None
|
|
|
|
| 167 |
rows = []
|
| 168 |
for p in all_products:
|
| 169 |
try:
|
| 170 |
+
p_id = int(p.get("id") or 0)
|
| 171 |
+
p_name = str(p.get("name") or "Unknown")
|
| 172 |
clean_name = _norm(p_name)
|
| 173 |
|
| 174 |
cat_obj = p.get("category") or {}
|
| 175 |
+
cat_name = str(cat_obj.get("name") or "General")
|
| 176 |
|
| 177 |
brand_obj = p.get("brand") or {}
|
| 178 |
+
brand_name = str(brand_obj.get("brand_name") or "")
|
| 179 |
|
| 180 |
views = int(p.get("view_count") or 0)
|
| 181 |
+
image = str(p.get("thumbnail") or p.get("image") or "")
|
| 182 |
|
| 183 |
prices = p.get("prices") or []
|
| 184 |
|
|
|
|
| 199 |
|
| 200 |
for offer in prices:
|
| 201 |
retailer = offer.get("retailer") or {}
|
| 202 |
+
r_name = str(retailer.get("name") or "Unknown Store")
|
| 203 |
price_val = _coerce_price(offer.get("price"))
|
| 204 |
|
| 205 |
if price_val > 0:
|
|
|
|
| 267 |
def calculate_basket_optimization(item_names: List[str]) -> Dict[str, Any]:
|
| 268 |
"""
|
| 269 |
Determines the best store for a list of items.
|
| 270 |
+
Explicitly casts int64/float64 to native Python types for JSON safety.
|
| 271 |
"""
|
| 272 |
df = get_market_index()
|
| 273 |
if df.empty:
|
|
|
|
| 289 |
# Best match based on popularity
|
| 290 |
best_prod = hits.iloc[0]
|
| 291 |
found_items.append({
|
| 292 |
+
"query": str(item),
|
| 293 |
+
"product_id": int(best_prod['product_id']), # Explicit int cast
|
| 294 |
+
"name": str(best_prod['product_name']) # Explicit str cast
|
| 295 |
})
|
| 296 |
|
|
|
|
| 297 |
if not found_items:
|
| 298 |
return {
|
| 299 |
"actionable": True,
|
|
|
|
| 318 |
found_names = [x['name'] for x in found_items if x['product_id'] in retailer_pids]
|
| 319 |
|
| 320 |
retailer_stats.append({
|
| 321 |
+
"retailer": str(retailer),
|
| 322 |
+
"total_price": float(total_price), # Explicit float cast
|
| 323 |
+
"item_count": int(found_count), # Explicit int cast
|
| 324 |
+
"coverage_percent": float((found_count / len(found_items)) * 100),
|
| 325 |
"found_items": found_names
|
| 326 |
})
|
| 327 |
|
|
|
|
| 372 |
breakdown.append(f"All {bought:.1f}u @ ${t1['rate']}")
|
| 373 |
|
| 374 |
return {
|
| 375 |
+
"amount_usd": float(amount_usd),
|
| 376 |
+
"est_units_kwh": float(round(units, 1)),
|
| 377 |
"breakdown": breakdown
|
| 378 |
}
|
| 379 |
|
| 380 |
# =========================
|
| 381 |
+
# 3. Gemini Helpers
|
| 382 |
# =========================
|
| 383 |
|
| 384 |
def gemini_detect_intent(transcript: str) -> Dict[str, Any]:
|
|
|
|
| 422 |
USER CONTEXT: "{transcript}"
|
| 423 |
|
| 424 |
ANALYST DATA (REAL PRICES):
|
| 425 |
+
{json.dumps(analyst_result, indent=2, default=str)}
|
| 426 |
|
| 427 |
REQUIRED SECTIONS (Do not deviate):
|
| 428 |
|
|
|
|
| 431 |
- Show product name, retailer, and exact price from data.
|
| 432 |
|
| 433 |
2. **Not in Catalogue (Estimates) 😔**
|
| 434 |
+
- List items from 'global_missing'.
|
| 435 |
- Provide a realistic USD estimate for Zimbabwe.
|
| 436 |
|
| 437 |
3. **Totals 💰**
|
|
|
|
| 470 |
|
| 471 |
@app.post("/chat")
|
| 472 |
def chat():
|
| 473 |
+
"""Text Chat."""
|
| 474 |
body = request.get_json(silent=True) or {}
|
| 475 |
msg = body.get("message", "")
|
| 476 |
pid = body.get("profile_id")
|
|
|
|
| 488 |
elif intent_data["intent"] == "PRODUCT_SEARCH" and intent_data.get("items"):
|
| 489 |
analyst_data = calculate_basket_optimization(intent_data["items"])
|
| 490 |
|
|
|
|
| 491 |
reply = "I'm processing that for you."
|
| 492 |
if analyst_data:
|
| 493 |
try:
|
| 494 |
resp = _gemini_client.models.generate_content(
|
| 495 |
model=GEMINI_MODEL,
|
| 496 |
+
contents=f"Summarize this shopping data for chat:\n{json.dumps(analyst_data, default=str)}"
|
| 497 |
)
|
| 498 |
reply = resp.text
|
| 499 |
except: pass
|
|
|
|
| 539 |
top = df[df['is_offer']].sort_values('views', ascending=False).drop_duplicates('product_name').head(60)
|
| 540 |
lines = []
|
| 541 |
for _, r in top.iterrows():
|
| 542 |
+
lines.append(f"{r['product_name']} (~${r['price']:.2f})")
|
| 543 |
catalogue_str = ", ".join(lines)
|
| 544 |
|
| 545 |
kpi_snapshot = {
|
|
|
|
| 556 |
@app.post("/api/log-call-usage")
|
| 557 |
def log_call_usage():
|
| 558 |
"""
|
| 559 |
+
Post-Call Orchestrator.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 560 |
"""
|
| 561 |
body = request.get_json(silent=True) or {}
|
| 562 |
pid = body.get("profile_id")
|
|
|
|
| 566 |
|
| 567 |
logger.info(f"Log Call: Processing {pid}. Transcript Len: {len(transcript)}")
|
| 568 |
|
|
|
|
| 569 |
if len(transcript) > 20 and db:
|
| 570 |
try:
|
| 571 |
curr_mem = db.collection("pricelyst_profiles").document(pid).get().to_dict().get("memory_summary", "")
|
|
|
|
| 575 |
except Exception as e:
|
| 576 |
logger.error(f"Memory Update Error: {e}")
|
| 577 |
|
|
|
|
| 578 |
intent_data = gemini_detect_intent(transcript)
|
| 579 |
logger.info(f"Intent: {intent_data.get('intent')}")
|
| 580 |
|
| 581 |
plan_data = {}
|
| 582 |
|
|
|
|
| 583 |
if intent_data.get("actionable"):
|
| 584 |
|
|
|
|
| 585 |
if intent_data.get("items"):
|
| 586 |
analyst_result = calculate_basket_optimization(intent_data["items"])
|
| 587 |
|
|
|
|
|
|
|
| 588 |
if analyst_result.get("actionable"):
|
| 589 |
md_content = gemini_generate_4step_plan(transcript, analyst_result)
|
| 590 |
|
|
|
|
| 602 |
doc_ref.set(plan_data)
|
| 603 |
logger.info(f"Plan Saved: {doc_ref.id}")
|
| 604 |
|
|
|
|
| 605 |
if db:
|
| 606 |
db.collection("pricelyst_profiles").document(pid).collection("call_logs").add({
|
| 607 |
"transcript": transcript,
|