Update app.py

app.py

@@ -29,8 +29,8 @@ town_to_row = {
 valid_towns = list(df_bfs_data["bfs_name"].sort_values().unique())
 
 
-# TODO 1:
-# Implement town matching from user text to canonical bfs_name.
+# Core Pipeline Functions
+
 def match_town(user_town: str):
     """Return the canonical town name from the dataset, or None."""
     if not user_town or not user_town.strip():
@@ -50,10 +50,6 @@ def match_town(user_town: str):
     return None
 
 
-# TODO 2 (LLM REQUIRED):
-# Implement one helper to call your chosen LLM and return JSON text.
-# Requirement: raise an error if API key/model is missing instead of using fallback logic.
-# Hint: the Week 1 OpenAI example uses `client.responses.create(...)`.
 def call_llm_json(system_prompt: str, user_prompt: str) -> str:
     """Call LLM with system and user prompts, return JSON response text."""
     if not LLM_API_KEY or not LLM_MODEL:
@@ -61,14 +57,16 @@ def call_llm_json(system_prompt: str, user_prompt: str) -> str:
     
     client = OpenAI(api_key=LLM_API_KEY)
 
-    response = client.responses.create(
+    response = client.chat.completions.create(
         model=LLM_MODEL,
-        input=user_prompt,
-        instructions=system_prompt,
-        max_output_tokens=500,
+        messages=[
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": user_prompt}
+        ],
+        max_tokens=500,
     )
 
-    return (response.output_text or "").strip()
+    return (response.choices[0].message.content or "").strip()
 
 
 # Validate the LLM response before the rest of the app depends on it.
@@ -98,8 +96,6 @@ def parse_json_response(raw: str, required_keys: tuple[str, ...]) -> dict:
     return parsed
 
 
-# TODO 3 (LLM REQUIRED):
-# Use your LLM to extract: rooms, area_m2, town from free text.
 def extract_preferences(user_text: str) -> dict:
     """Extract rooms, area_m2, and town from free text using LLM."""
     system_prompt = """Du bist ein Assistent, der Wohnungswünsche in strukturierte Daten umwandelt.
@@ -122,10 +118,6 @@ Antworte ausschließlich mit gültigem JSON in diesem Format:
     return parsed
 
 
-# TODO 4:
-# Implement numeric model prediction with exactly these features:
-# [rooms, area_m2, pop, pop_dens, frg_pct, emp, tax_income]
-# The provided model is a pickled scikit-learn regressor loaded above.
 def predict_apartment_price(rooms: float, area_m2: float, town: str) -> float:
     """Predict monthly rent using the loaded model."""
     # Get town data
@@ -151,8 +143,6 @@ def predict_apartment_price(rooms: float, area_m2: float, town: str) -> float:
     return round(prediction, 2)
 
 
-# TODO 5 (LLM REQUIRED):
-# Use your LLM to generate a concise explanation with one uncertainty note.
 def generate_explanation(preferences: dict, prediction: float) -> str:
     """Generate a user-friendly explanation using LLM."""
     system_prompt = "Du bist ein hilfsbereiter Assistent für Immobilienvorhersagen."
@@ -176,8 +166,6 @@ Antworte ausschließlich mit gültigem JSON in diesem Format:
     return parsed["answer"]
 
 
-# TODO 6:
-# Implement the end-to-end pipeline.
 def run_pipeline(user_text: str):
     """End-to-end pipeline: extract -> predict -> explain."""
     try: