Update app.py

app.py

@@ -29,8 +29,7 @@ from pydantic import BaseModel, Field
 import litellm
 from langchain.tools import Tool
 
-LLM._get_litellm_model_name = lambda self, model_name: f"gemini/{model_name}" if not "/" in model_name else model_name
-os.environ["LITELLM_MODEL_DEFAULT_PROVIDER"] = "gemini"
+
 
 # Configure logging
 logging.basicConfig(level=logging.INFO)
@@ -62,7 +61,7 @@ if 'board_plan_generated' not in st.session_state:
 # Load environment variables
 #load_dotenv()
 
-os.environ["LITELLM_MODEL_DEFAULT_PROVIDER"] = "gemini"
+
 
 
 # Page title and description
@@ -73,6 +72,7 @@ st.write("Developed for BIA 568 (Business Intelligence and Analytics) -- Managem
 st.write("---")
 
 # Sidebar for API key configuration
+# In the sidebar section:
 with st.sidebar:
     st.title("⚙️ Configuration")
     
@@ -81,7 +81,17 @@ with st.sidebar:
     if api_key:
         os.environ["GEMINI_API_KEY"] = api_key
         os.environ["GOOGLE_API_KEY"] = api_key
-        #os.environ["OPENAI_API_KEY"] = api_key
+        os.environ["LITELLM_MODEL_DEFAULT_PROVIDER"] = "gemini"
+
+        # Configure litellm
+        litellm.set_verbose = True
+        litellm_config = {
+            "model": "gemini/gemini-2.0-flash",
+            "api_key": api_key,
+            "provider": "gemini"
+        }
+    
+
     
     st.divider()
     
@@ -393,10 +403,13 @@ class SectionContent(BaseModel):
 class CaseBreakdownCrew:
     def __init__(self, api_key):
         self.api_key = api_key
-        
+        self.llm = LLM(
+            model='gemini/gemini-2.0-flash',
+            api_key=self.api_key,
+            provider="gemini"
+        )
+    
     def create_metadata_agent(self):
-        self.api_key = api_key
-        self.llm = LLM(model='gemini/gemini-2.0-flash', api_key=self.api_key)  # Create a Gemini LLM instance
         return Agent(
             role="Metadata Analyzer",
             goal="Extract title and author information from document content",
@@ -404,11 +417,16 @@ class CaseBreakdownCrew:
             such as titles, authors, and other publication information. You have a keen eye for
             identifying the most important and relevant document metadata, even when it's not
             explicitly labeled.""",
+            llm=LLM(
+            model='gemini/gemini-2.0-flash',
+            api_key=self.api_key,
+            provider="gemini"
+        ),
             verbose=True
         )
     
     def create_content_generator_agent(self):
-        llm = LLM(model='gemini/gemini-2.0-flash', api_key=self.api_key)
+        
         return Agent(
             role="Case Study Content Generator",
             goal="Generate comprehensive case analysis content based on section requirements",
@@ -416,11 +434,12 @@ class CaseBreakdownCrew:
             You excel at breaking down complex business cases into structured, insightful content
             that highlights key learning points, strategies, and insights. You have extensive experience
             in business education and know how to create content that is valuable for teaching and learning.""",
+            llm=self.llm,
             verbose=True
         )
     
     def create_content_reviewer_agent(self):
-        llm = LLM(model='gemini/gemini-2.0-flash', api_key=self.api_key)
+        
         return Agent(
             role="Content Quality Reviewer",
             goal="Evaluate and score content for quality, relevance, and depth",
@@ -428,6 +447,7 @@ class CaseBreakdownCrew:
             business case studies and educational content. You have a strong understanding of what makes
             effective case study material and can provide constructive feedback to improve content quality.
             You carefully analyze content for relevance, clarity, depth, and educational value.""",
+            llm=self.llm,
             verbose=True
         )
     
@@ -506,7 +526,6 @@ class CaseBreakdownCrew:
             agents=[self.create_metadata_agent()],
             tasks=[metadata_task],
             process=Process.sequential,
-            llm=LLM(model='gemini/gemini-2.0-flash', api_key=self.api_key),
             verbose=False
         )
         result = crew.kickoff()
@@ -603,17 +622,15 @@ def extract_text(file_path: str) -> str:
     except Exception as e:
         return f"Error extracting text: {str(e)}"
 
-def create_teaching_plan_crew(file_paths, llm_provider="gemini"):
-    # Initialize the agent tracker
-    tracker = AgentTracker()
-    tracker.set_placeholder(st.empty())
-    
-    # Initialize LLM based on provider
-    my_llm = LLM(model='gemini/gemini-2.0-flash',
-        api_key=os.environ.get("GEMINI_API_KEY")
+def create_teaching_plan_crew(file_paths):
+    # Initialize LLM 
+    my_llm = LLM(
+        model='gemini/gemini-2.0-flash',
+        api_key=os.environ.get("GEMINI_API_KEY"),
+        provider="gemini"
     )
     
-    # Define agents with callbacks for UI updates
+    # Create agents with Gemini configuration
     pdf_analyzer = Agent(
         role='Case Study Analyzer',
         goal='Extract key concepts, objectives, and data from case study files',
@@ -707,9 +724,13 @@ class BoardPlanAnalyzer:
         api_key = os.environ.get('GEMINI_API_KEY')
         if not api_key:
             raise ValueError("Gemini API key not found")
-        # Create an LLM instance configured for Gemini
-        self.llm = LLM(model='gemini/gemini-2.0-flash', api_key=api_key)
             
+        self.llm = LLM(
+            model='gemini/gemini-2.0-flash',
+            api_key=api_key,
+            provider="gemini"
+        )
+        
         litellm.set_verbose = True
 
         # Create agents
@@ -811,11 +832,13 @@ class BoardPlanAnalyzer:
         
         try:
             response = litellm.completion(
-                model=self.llm,
+                model="gemini/gemini-2.0-flash",
                 messages=messages,
                 api_key=os.environ.get("GEMINI_API_KEY"),
+                provider="gemini",
                 response_format={"type": "json_object"}
             )
+
             
             # Extract and parse the JSON response
             content = response.choices[0].message.content