Delete confluence_integration_example_2.py

confluence_integration_example_2.py

@@ -1,414 +0,0 @@
-"""
-Phase 2: Enhanced Confluence Integration with Space Filtering and Caching
-
-This example builds on confluence_integration_example.py with:
-- Specialized search tools for different Confluence spaces
-- Smart tool selection based on query type
-- Caching for improved performance
-
-Prerequisites:
-    - Complete Phase 1 (confluence_integration_example.py)
-    - Install: pip install -r requirements-with-confluence.txt
-    - Configure .env with Confluence credentials
-"""
-
-import os
-import warnings
-from functools import lru_cache
-from typing import Optional
-
-# Suppress ResourceWarning for cleaner output
-warnings.filterwarnings("ignore", category=ResourceWarning)
-os.environ["PYTHONWARNINGS"] = "ignore::ResourceWarning"
-
-# Load environment variables from .env file
-try:
-    from dotenv import load_dotenv
-    load_dotenv()
-except ImportError:
-    print("⚠ Warning: python-dotenv not installed. Install with: pip install python-dotenv")
-    print("   Environment variables must be set manually.")
-
-import gradio as gr
-from strands import Agent
-from strands.models.openai import OpenAIModel
-
-# Import confluence-ingestor
-from confluence_ingestor import ConfluenceRAG
-from confluence_ingestor.adapters.strands import (
-    create_confluence_search_tool,
-    create_confluence_loader_tool,
-)
-
-# Import your existing fraud tools
-from app import (
-    get_application_summary,
-    explain_fraud_score,
-    compare_to_population,
-    check_fair_lending_flags,
-    get_identity_network,
-    get_model_performance,
-    SYSTEM_PROMPT as ORIGINAL_PROMPT,
-)
-
-
-# =============================================================================
-# STEP 1: Initialize Confluence RAG (Same as Phase 1)
-# =============================================================================
-
-_confluence_rag: Optional[ConfluenceRAG] = None
-
-
-def init_confluence():
-    """Initialize Confluence RAG (runs once at startup)."""
-    global _confluence_rag
-
-    if _confluence_rag is None:
-        print("🔧 Initializing Confluence integration...")
-
-        # Check if required environment variables are set
-        required_vars = ["CONFLUENCE_URL", "CONFLUENCE_EMAIL", "CONFLUENCE_API_TOKEN"]
-        missing_vars = [var for var in required_vars if not os.getenv(var)]
-
-        if missing_vars:
-            print(f"\n❌ ERROR: Missing required environment variables: {', '.join(missing_vars)}")
-            print("\n📝 Setup Instructions:")
-            print("   1. Copy .env.example to .env:")
-            print("      cp .env.example .env")
-            print("   2. Edit .env with your Confluence credentials")
-            print("   3. See CONFLUENCE_SETUP_GUIDE.md for detailed setup instructions")
-            print("\n⚠  App will run WITHOUT Confluence integration.\n")
-            raise ValueError(f"Missing Confluence credentials: {', '.join(missing_vars)}")
-
-        # Create RAG pipeline with free local embeddings
-        _confluence_rag = ConfluenceRAG.from_env(
-            embedding_provider="huggingface",  # Free, runs locally
-            vector_store_type="chroma",
-        )
-
-        # Ingest your Confluence spaces (runs once, then cached)
-        spaces = {
-            "fraud-model-governance": 100,  # Model docs, validation reports
-            "compliance-policies": 50,  # Fair lending, ECOA policies
-            "fraud-investigation-playbooks": 75,  # Investigation procedures
-        }
-
-        for space_key, max_pages in spaces.items():
-            try:
-                stats = _confluence_rag.ingest_space(
-                    space_key, max_pages=max_pages, force=False
-                )
-
-                if stats.get("skipped"):
-                    print(f"   ✓ {space_key}: {stats['reason']}")
-                else:
-                    print(f"   ✓ {space_key}: {stats['pages']} pages indexed")
-
-            except Exception as e:
-                print(f"   ⚠ {space_key}: Failed - {e}")
-
-        print("✅ Confluence integration ready!")
-
-    return _confluence_rag
-
-
-# =============================================================================
-# PHASE 2 ENHANCEMENT 1: Space-Filtered Search Tools
-# =============================================================================
-
-def create_specialized_search_tools(rag: ConfluenceRAG) -> dict:
-    """Create specialized search tools for different Confluence spaces.
-
-    This allows the agent to search specific spaces for targeted results,
-    improving relevance and response time.
-
-    Returns:
-        Dictionary of specialized search tools
-    """
-    return {
-        "search_compliance": create_confluence_search_tool(
-            rag=rag,
-            k=3,  # Fewer results for focused searches
-            space_filter="compliance-policies"
-        ),
-        "search_model_governance": create_confluence_search_tool(
-            rag=rag,
-            k=3,
-            space_filter="fraud-model-governance"
-        ),
-        "search_investigation": create_confluence_search_tool(
-            rag=rag,
-            k=3,
-            space_filter="fraud-investigation-playbooks"
-        ),
-        "search_all": create_confluence_search_tool(
-            rag=rag,
-            k=5,  # More results for general searches
-            space_filter=None
-        ),
-    }
-
-
-# =============================================================================
-# PHASE 2 ENHANCEMENT 2: Caching Layer
-# =============================================================================
-
-@lru_cache(maxsize=100)
-def cached_confluence_search(query: str, space_filter: Optional[str] = None) -> str:
-    """Cached Confluence search for frequently asked questions.
-
-    Args:
-        query: Search query
-        space_filter: Optional space key filter
-
-    Returns:
-        Formatted search results
-    """
-    rag = init_confluence()
-    results = rag.search(query, k=5, space_filter=space_filter)
-
-    if not results:
-        return "No relevant documents found in Confluence."
-
-    # Format results
-    output = [f"Found {len(results)} relevant documents from Confluence:\n"]
-
-    for i, result in enumerate(results, 1):
-        output.append(f"\n{i}. {result.title}")
-        output.append(f"   Space: {result.space_key}")
-        output.append(f"   Relevance Score: {result.score:.4f}")
-        output.append(f"   Source: {result.source}")
-
-        # Truncate content for readability
-        content = (
-            result.content[:300] + "..."
-            if len(result.content) > 300
-            else result.content
-        )
-        output.append(f"   Content: {content}")
-
-    return "\n".join(output)
-
-
-# =============================================================================
-# STEP 2: Enhanced System Prompt with Smart Tool Selection
-# =============================================================================
-
-ENHANCED_PROMPT_PHASE2 = """
-You are a Fraud Model Explainability Assistant for a major financial services company.
-Your role is to help fraud analysts, data scientists, and executives understand
-fraud model decisions and their implications.
-
-You have access to tools that can:
-1. Retrieve application summaries and fraud scores
-2. Explain why applications received specific fraud scores (SHAP-style explanations)
-3. Compare applications to approved/denied populations statistically
-4. Check for fair lending compliance concerns
-5. Analyze identity networks and linkages
-6. Show model performance metrics
-7. **Search company Confluence documentation** with specialized tools:
-   - search_compliance: Search compliance policies and fair lending documentation
-   - search_model_governance: Search model validation reports and SR 11-7 docs
-   - search_investigation: Search fraud investigation playbooks and procedures
-   - search_all: Search all Confluence spaces
-
-SMART TOOL SELECTION GUIDELINES:
-- For fair lending questions → Use search_compliance FIRST
-- For model documentation questions → Use search_model_governance FIRST
-- For investigation procedures → Use search_investigation FIRST
-- For general or multi-topic questions → Use search_all
-
-When answering questions:
-- Be precise and data-driven
-- Highlight the most important risk factors first
-- Explain technical concepts in business terms when speaking to executives
-- **Use the most specific Confluence search tool for the question type**
-- **Cite specific Confluence pages when referencing procedures**
-- **For compliance questions, ALWAYS search for relevant policy documentation**
-- **For model governance questions, reference actual validation reports when available**
-- Always mention fair lending implications when relevant
-- Provide actionable insights, not just data
-
-For flagged applications, structure your response as:
-1. Quick summary (score, decision, risk level)
-2. Top contributing factors
-3. How unusual this is compared to the population
-4. Any compliance considerations (with policy references from Confluence)
-5. Recommended next steps
-
-Remember: Your explanations may be used in regulatory examinations and audits,
-so be accurate and thorough. When citing company policies, always reference the
-source Confluence page.
-""".strip()
-
-
-# =============================================================================
-# STEP 3: Create Agent with Phase 2 Enhancements
-# =============================================================================
-
-def create_enhanced_agent_phase2():
-    """Create fraud agent with Phase 2 Confluence enhancements.
-
-    Phase 2 features:
-    - Specialized search tools for different Confluence spaces
-    - Caching for improved performance
-    - Smart tool selection guidance in system prompt
-    """
-    openai_api_key = os.environ.get("OPENAI_API_KEY")
-
-    # Initialize Confluence
-    try:
-        rag = init_confluence()
-
-        # Create specialized search tools
-        specialized_tools = create_specialized_search_tools(rag)
-
-        # Create loader tool
-        load_confluence_page = create_confluence_loader_tool(max_pages=3)
-
-        # Add all tools (fraud + specialized Confluence)
-        tools = [
-            get_application_summary,
-            explain_fraud_score,
-            compare_to_population,
-            check_fair_lending_flags,
-            get_identity_network,
-            get_model_performance,
-            specialized_tools["search_compliance"],  # NEW: Compliance-specific search
-            specialized_tools["search_model_governance"],  # NEW: Model governance search
-            specialized_tools["search_investigation"],  # NEW: Investigation playbook search
-            specialized_tools["search_all"],  # NEW: General Confluence search
-            load_confluence_page,  # Confluence page loader
-        ]
-
-        system_prompt = ENHANCED_PROMPT_PHASE2
-
-    except Exception as e:
-        print(f"⚠ Confluence disabled: {e}")
-
-        # Fall back to original tools if Confluence fails
-        tools = [
-            get_application_summary,
-            explain_fraud_score,
-            compare_to_population,
-            check_fair_lending_flags,
-            get_identity_network,
-            get_model_performance,
-        ]
-
-        system_prompt = ORIGINAL_PROMPT
-
-    # Create agent
-    if openai_api_key:
-        model = OpenAIModel(
-            client_args={"api_key": openai_api_key},
-            model_id="gpt-4o",
-            params={"temperature": 0.1, "max_tokens": 2048},
-        )
-        return Agent(model=model, system_prompt=system_prompt, tools=tools)
-    else:
-        # Default to Bedrock
-        return Agent(system_prompt=system_prompt, tools=tools)
-
-
-def query_enhanced(question: str) -> str:
-    """Process question with Phase 2 enhanced agent."""
-    try:
-        agent = create_enhanced_agent_phase2()
-        result = agent(question)
-        return str(result)
-    except Exception as e:
-        return f"Error: {str(e)}"
-
-
-# =============================================================================
-# STEP 4: Gradio Interface
-# =============================================================================
-
-def process_question(question: str) -> str:
-    """Wrapper for Gradio."""
-    return query_enhanced(question)
-
-
-# Create interface
-with gr.Blocks(title="Fraud Model Explainability Assistant - Phase 2") as iface:
-    gr.Markdown(
-        """
-    # 🔍 Fraud Model Explainability Assistant (Phase 2)
-
-    Enhanced with **specialized Confluence search tools** and **performance caching**.
-
-    **Phase 2 Enhancements:**
-    - **Space-filtered search**: Targeted searches in compliance, model governance, and investigation spaces
-    - **Smart tool selection**: Agent automatically chooses the best search tool for each question
-    - **Performance caching**: Frequently asked questions return instantly
-
-    **Example Questions:**
-    - "What does our fair lending policy say about synthetic ID detection?" (→ search_compliance)
-    - "Find the model validation report for XGBoost v3.2" (→ search_model_governance)
-    - "What are the procedures for escalating high-risk applications?" (→ search_investigation)
-    """
-    )
-
-    with gr.Row():
-        with gr.Column(scale=2):
-            question_input = gr.Textbox(
-                label="Ask a Question",
-                placeholder="e.g., What does our fair lending policy say about phone type features?",
-                lines=3,
-            )
-            submit_btn = gr.Button("🔍 Analyze", variant="primary")
-
-    with gr.Row():
-        output = gr.Textbox(label="Analysis Results", lines=25)
-
-    gr.Markdown("### 💡 Example Questions")
-
-    examples = gr.Examples(
-        examples=[
-            ["Why was application APP-78432 flagged as high risk?"],
-            ["Explain the fraud score for APP-12345 and compare it to approved applications"],
-            ["Check fair lending compliance for APP-55555 and cite relevant policies"],
-            ["Show me the identity network analysis for APP-78432"],
-            ["What's the current model performance for the Retail Card portfolio?"],
-            ["What does our fair lending policy say about synthetic ID detection?"],
-            ["Find the model validation report for XGBoost v3.2"],
-            ["What are the procedures for escalating high-risk applications?"],
-            ["I need to present APP-99999 to the CCO. Give me a complete risk summary with compliance review and policy references."],
-        ],
-        inputs=question_input,
-    )
-
-    submit_btn.click(fn=process_question, inputs=question_input, outputs=output)
-    question_input.submit(fn=process_question, inputs=question_input, outputs=output)
-
-    gr.Markdown(
-        """
-    ---
-    *Powered by Strands Agents + Confluence Integration (Phase 2)*
-
-    **Phase 2 Features:**
-    - ✅ Specialized search tools for compliance, model governance, and investigations
-    - ✅ LRU caching for frequently asked questions
-    - ✅ Smart tool selection guidance
-
-    **Next Steps:**
-    - See `confluence_integration_phase3.py` for monitoring, logging, and scheduled updates
-    """
-    )
-
-
-# =============================================================================
-# MAIN
-# =============================================================================
-
-if __name__ == "__main__":
-    # Pre-initialize Confluence (optional, speeds up first query)
-    try:
-        init_confluence()
-    except Exception as e:
-        print(f"Warning: Confluence initialization failed: {e}")
-        print("App will run without Confluence integration.")
-
-    # Launch Gradio UI
-    iface.launch(ssr_mode=False)