Upload folder using huggingface_hub

ai_engine.py

@@ -1,36 +1,45 @@
-import google.generativeai as genai
 import os
 import json
+from groq import Groq
 
-# Configure the Gemini API
-genai.configure(api_key=os.getenv("GOOGLE_API_KEY"))
-
-# Initialize the model (using the standard gemini-pro which is universally supported)
-model = genai.GenerativeModel(
-    model_name='gemini-pro',
-    generation_config={"response_mime_type": "application/json"} # Forces perfect JSON!
-)
+# Initialize the Groq client
+client = Groq(api_key=os.getenv("GROQ_API_KEY"))
 
 def analyze_lead(name: str, company: str, company_summary: str) -> dict:
     prompt = f"""
     You are a B2B sales expert. Analyze this lead.
     Lead: {name} at {company}. Info: {company_summary}
     
-    Return ONLY this JSON structure:
+    1. Score the lead from 1 to 10 based on how likely they are to need AI software automation. 
+       (E.g., outdated businesses = high score, modern tech = lower score).
+    2. Write a 1-sentence reason for the score.
+    3. Write a short, highly personalized 100-word cold email offering our AI services to them. Start with Hi {name},. End by asking for a 10-min chat.
+
+    Output EXACTLY in this JSON format:
     {{
-      "score": <integer 1-10>,
-      "score_reason": "<one sentence why this score>",
-      "cold_email": "<a 100-word personalized cold email pitching AI automation services. Start with Hi {name},. End by asking for a 10-min chat.>"
+        "score": 8,
+        "score_reason": "They have a lot of manual processes...",
+        "cold_email": "Hi {name}..."
     }}
     """
     
     try:
-        response = model.generate_content(prompt)
-        # Because we forced JSON mime type, this is 100% safe to parse
-        return json.loads(response.text)
+        response = client.chat.completions.create(
+            messages=[
+                {"role": "system", "content": "You are an expert B2B sales qualifier. Output ONLY a valid JSON object."},
+                {"role": "user", "content": prompt}
+            ],
+            model="llama3-70b-8192",
+            response_format={"type": "json_object"}
+        )
+        
+        # Parse the JSON returned by Llama 3
+        result = json.loads(response.choices[0].message.content)
+        return result
         
     except Exception as e:
         # Fallback if something goes wrong
+        print(f"Error calling Groq: {e}")
         return {
             "score": 0, 
             "score_reason": f"AI Error: {str(e)}", 

requirements.txt

@@ -1,6 +1,6 @@
 fastapi
 uvicorn
-google-generativeai
+groq
 supabase
 requests
 python-dotenv