Initial commit with full functionality extend app req tools

app.py

@@ -1,5 +1,5 @@
 # app.py
-from smolagents import CodeAgent, HfApiModel
+from smolagents import CodeAgent, HfApiModel, Tool
 import logging
 from Gradio_UI import GradioUI
 
@@ -10,9 +10,22 @@ logging.basicConfig(
 )
 logger = logging.getLogger(__name__)
 
+class AnalysisTool(Tool):
+    """Tool for analyzing web content."""
+    
+    name = "web_analyzer"
+    description = "Analyzes web content for summaries, sentiment, and topics"
+    
+    def forward(self, text: str, analysis_type: str) -> str:
+        """Process the text based on analysis type."""
+        return text
+
 def create_agent():
     """Create and configure the agent."""
     try:
+        # Initialize tools
+        analyzer = AnalysisTool()
+        
         # Initialize model
         model = HfApiModel(
             model_id='Qwen/Qwen2.5-Coder-32B-Instruct',
@@ -20,9 +33,10 @@ def create_agent():
             temperature=0.5
         )
         
-        # Create agent with simple configuration
+        # Create agent with tools
         return CodeAgent(
             model=model,
+            tools=[analyzer],  # Add the tools here
             max_steps=3,
             verbosity_level=1
         )

prompts.yaml

@@ -1,50 +1,28 @@
 # prompts.yaml
 system_prompt: |
-  You are a sophisticated AI assistant specialized in web content analysis. You have access to these capabilities:
-  - Text extraction: Clean and extract meaningful content from URLs
-  - Sentiment analysis: Determine emotional tone and context
-  - Summarization: Create concise, informative summaries
-  - Topic detection: Identify main themes and subjects
-  - Web search: Gather additional context when needed
-  - Temporal analysis: Consider time-based context
+  You are an AI assistant specialized in analyzing web content. You can:
+  - Extract and clean text from web pages
+  - Create concise summaries
+  - Analyze sentiment
+  - Identify main topics
   
-  Always structure your responses in JSON format with these keys:
-  - clean_text: The extracted and cleaned content
-  - summary: A concise summary if requested
-  - sentiment: Sentiment analysis if requested
-  - topics: Main topics if requested
-  
-  Think step by step and use the most appropriate tools for each task.
-
-user: |
-  User query: {input}
-  
-  Analysis process:
-  1. Validate and process the URL
-  2. Determine required analysis types
-  3. Execute analysis in order: extraction → summary → sentiment → topics
-  4. Format results in JSON structure
-  
-  Available tools: {tools}
+  Always provide clear, structured responses.
 
-A: |
-  I'll analyze the content systematically:
+user_prompt: |
+  Please analyze the following content: {input}
+  Analysis types requested: {analysis_types}
 
-  {thoughts}
+  Return your analysis in a clear, organized format.
 
-  Executing analysis with appropriate tools...
+assistant_prompt: |
+  I'll analyze that content step by step:
+  
+  {analysis}
 
 observation: |
-  Tool response: {output}
-
-final: |
-  Analysis complete. Results formatted in JSON:
+  Analysis results: {results}
 
-  {response}
-
-error: |
-  An error occurred: {error}
-  
-  Technical details: {error_details}
+final_prompt: |
+  Here's the complete analysis:
   
-  I'll attempt an alternative approach or provide partial results if possible.
+  {final_results}

requirements.txt

@@ -3,4 +3,5 @@ gradio>=4.0.0
 requests>=2.31.0
 beautifulsoup4>=4.12.2
 smolagents>=0.2.0
-python-dotenv>=1.0.0
+python-dotenv>=1.0.0
+pyyaml>=6.0.1