feat: Add vertex API support

.gitignore

@@ -101,4 +101,6 @@ venv.bak/
 # mypy
 .mypy_cache/
 .dmypy.json
-dmypy.json
+dmypy.json
+
+demos/rfe-builder/.streamlit/secrets.toml

demos/rfe-builder/.streamlit/secrets.toml.template

@@ -3,9 +3,17 @@
 # Note: secrets.toml is automatically ignored by git for security
 
 # Anthropic Claude API Configuration
-# Get your API key from: https://console.anthropic.com/
+# Option 1: Direct API Key (get from: https://console.anthropic.com/)
 ANTHROPIC_API_KEY = "your-anthropic-api-key-here"
 
+# Option 2: Use Vertex AI (if using Google Cloud Vertex AI)
+# Set these environment variables instead of using secrets.toml:
+# export CLAUDE_CODE_USE_VERTEX=1
+# export CLOUD_ML_REGION=us-east5
+# export ANTHROPIC_VERTEX_PROJECT_ID=your-project-id
+# export ANTHROPIC_MODEL='claude-sonnet-4@20250514'
+# export ANTHROPIC_SMALL_FAST_MODEL='claude-3-5-haiku@20241022'
+
 # Optional: Configure Claude model preferences
 ANTHROPIC_MODEL = "claude-3-haiku-20240307"  # Cost-effective default
 # Alternative models:

demos/rfe-builder/README.md

@@ -76,6 +76,8 @@ The RFE Builder implements a 7-agent workflow system:
    ```
 
 4. **Configure AI features (Phase 2 - Optional)**
+
+   **Option A: Direct Anthropic API (Standard)**
    ```bash
    # Copy secrets template
    cp .streamlit/secrets.toml.template .streamlit/secrets.toml
@@ -84,6 +86,26 @@ The RFE Builder implements a 7-agent workflow system:
    # Get your key from: https://console.anthropic.com/
    ```
 
+   **Option B: Google Cloud Vertex AI (Enterprise)**
+   ```bash
+   # Set environment variables for Vertex AI
+   export CLAUDE_CODE_USE_VERTEX=1
+   export CLOUD_ML_REGION=us-east5  # Your preferred region
+   export ANTHROPIC_VERTEX_PROJECT_ID=your-gcp-project-id
+   export ANTHROPIC_MODEL='claude-sonnet-4@20250514'
+   export ANTHROPIC_SMALL_FAST_MODEL='claude-3-5-haiku@20241022'
+
+   # Optional: Configure timeouts and retry behavior
+   export ANTHROPIC_TIMEOUT=30.0  # Connection timeout in seconds
+   export ANTHROPIC_MAX_RETRIES=3  # Maximum retry attempts
+
+   # Dependencies are included in requirements.txt
+   # No additional installation needed if using requirements.txt
+
+   # Ensure Google Cloud authentication is configured
+   gcloud auth application-default login
+   ```
+
 5. **Run the application**
    ```bash
    streamlit run app.py
@@ -135,6 +157,48 @@ Each agent role has specific capabilities with AI-powered assistance:
   - Generate JIRA tickets and development tasks with AI assistance
 - **Other Agents**: Specialized assessment functions with role-specific AI guidance
 
+## 🔧 Troubleshooting
+
+### AI Configuration Issues
+
+#### Vertex AI Connection Problems
+If you see "I'm having trouble connecting to the AI service" with Vertex AI:
+
+1. **Verify environment variables are set:**
+   ```bash
+   echo $CLAUDE_CODE_USE_VERTEX
+   echo $ANTHROPIC_VERTEX_PROJECT_ID
+   echo $CLOUD_ML_REGION
+   ```
+
+2. **Test your configuration:**
+   ```bash
+   # Verify Google Cloud authentication works
+   gcloud auth list
+   gcloud config get-value project
+   ```
+
+3. **Common issues:**
+   - Missing Google Cloud authentication: Run `gcloud auth application-default login`
+   - Wrong project ID: Verify your GCP project has Vertex AI API enabled
+   - Missing dependencies: Ensure `anthropic[vertex]` and `google-cloud-aiplatform` are installed
+   - Incorrect region: Use a region that supports Claude models (e.g., `us-east5`, `us-central1`)
+   - Unsupported model: Verify your model is supported (see supported models below)
+   - Connection timeouts: Increase `ANTHROPIC_TIMEOUT` for slow networks
+   - Intermittent failures: The system will retry up to `ANTHROPIC_MAX_RETRIES` times automatically
+
+#### Supported Vertex AI Models
+- `claude-3-5-sonnet@20241022` (recommended)
+- `claude-3-5-haiku@20241022` (fast, cost-effective)
+- `claude-sonnet-4@20250514` (latest, most capable)
+- `claude-3-sonnet@20240229`
+- `claude-3-haiku@20240307`
+
+#### Direct API Issues
+If using direct Anthropic API:
+- Verify your API key in `.streamlit/secrets.toml`
+- Check your account has sufficient credits at https://console.anthropic.com/
+
 ## 🧪 Testing
 
 ### Run Tests
@@ -177,13 +241,25 @@ rfe-builder/
 │   └── rfe_models.py          # Data models and state management
 ├── components/
 │   ├── __init__.py
-│   └── workflow.py            # Workflow visualization components
+│   ├── workflow.py            # Workflow visualization components
+│   ├── chat_interface.py      # AI-powered conversational interface
+│   └── ai_assistants.py       # Agent-specific AI assistants
+├── ai_models/
+│   ├── __init__.py
+│   ├── cost_tracker.py        # AI usage cost tracking
+│   └── prompt_manager.py      # AI prompt management
+├── prompts/
+│   ├── conversational_rfe_creation.yaml
+│   └── agents/                # Agent-specific prompt templates
 ├── pages/
 │   └── parker_pm.py           # Agent-specific page (example)
 ├── tests/
 │   ├── __init__.py
 │   ├── test_rfe_models.py     # Model tests
 │   └── test_workflow.py       # Workflow tests
+├── .streamlit/
+│   ├── secrets.toml.template  # Configuration template
+│   └── secrets.toml           # API credentials (not in git)
 ├── .github/
 │   └── workflows/
 │       └── ci.yml             # CI/CD pipeline

demos/rfe-builder/ai_models/anthropic_client.py

@@ -0,0 +1,240 @@
+"""
+Shared utility for Anthropic client initialization with Vertex AI support
+"""
+
+import os
+import re
+import time
+from typing import Optional, Union, Set
+import streamlit as st
+from anthropic import Anthropic, AnthropicVertex
+
+
+# Supported Claude models on Vertex AI
+SUPPORTED_VERTEX_MODELS: Set[str] = {
+    'claude-3-5-sonnet@20241022',
+    'claude-3-5-haiku@20241022', 
+    'claude-3-sonnet@20240229',
+    'claude-3-haiku@20240307',
+    'claude-sonnet-4@20250514',  # Latest Sonnet 4
+}
+
+# Default timeouts (in seconds)
+DEFAULT_TIMEOUT = 30.0
+DEFAULT_MAX_RETRIES = 3
+DEFAULT_RETRY_DELAY = 1.0
+
+
+def validate_model_name(model: str) -> Optional[str]:
+    """
+    Validate model name for Vertex AI compatibility.
+
+    Args:
+        model: Model name to validate
+
+    Returns:
+        Error message if validation fails, None if valid
+    """
+    if not model:
+        return "Model name cannot be empty or None"
+
+    # Handle whitespace-only strings
+    if not model.strip():
+        return "Model name cannot be empty or whitespace"
+
+    if model not in SUPPORTED_VERTEX_MODELS:
+        return (
+            f"Model '{model}' is not supported on Vertex AI. "
+            f"Supported models: {', '.join(sorted(SUPPORTED_VERTEX_MODELS))}"
+        )
+
+    return None
+
+
+def validate_vertex_config(project_id: Optional[str], region: Optional[str]) -> Optional[str]:
+    """
+    Validate Vertex AI configuration parameters.
+
+    Args:
+        project_id: Google Cloud project ID
+        region: Google Cloud region
+
+    Returns:
+        Error message if validation fails, None if valid
+    """
+    if not project_id:
+        return "Missing ANTHROPIC_VERTEX_PROJECT_ID environment variable"
+
+    if not region:
+        return "Missing CLOUD_ML_REGION environment variable"
+
+    # Validate project ID format (alphanumeric, hyphens, 6-30 chars, cannot start with number)
+    if not re.match(r'^[a-z][a-z0-9\-]{4,28}[a-z0-9]$', project_id):
+        return f"Invalid project ID format: '{project_id}'. Must be 6-30 characters, start with lowercase letter, contain only lowercase letters, numbers, and hyphens."
+
+    # Validate region format (e.g., us-east5, europe-west1)
+    if not re.match(r'^[a-z]+-[a-z]+\d+$', region):
+        return f"Invalid region format: '{region}'. Expected format like 'us-east5' or 'europe-west1'."
+
+    # Check for supported regions (common ones for Claude)
+    supported_regions = {
+        'us-east5', 'us-central1', 'us-west1', 'us-west4',
+        'europe-west1', 'europe-west4', 'asia-southeast1'
+    }
+    if region not in supported_regions:
+        return f"Region '{region}' may not support Claude models. Supported regions: {', '.join(sorted(supported_regions))}"
+
+    return None
+
+
+def _create_vertex_client_with_retry(
+    project_id: str, 
+    region: str, 
+    max_retries: int = DEFAULT_MAX_RETRIES,
+    timeout: float = DEFAULT_TIMEOUT,
+    show_errors: bool = True
+) -> Optional[AnthropicVertex]:
+    """
+    Create Vertex AI client with retry logic and timeout configuration.
+
+    Args:
+        project_id: Google Cloud project ID
+        region: Google Cloud region  
+        max_retries: Maximum number of retry attempts
+        timeout: Connection timeout in seconds
+        show_errors: Whether to display errors in Streamlit UI
+
+    Returns:
+        Configured AnthropicVertex client or None if creation fails
+    """
+    last_error = None
+
+    for attempt in range(max_retries + 1):
+        try:
+            # Create client with timeout configuration
+            client = AnthropicVertex(
+                project_id=project_id,
+                region=region,
+                timeout=timeout
+            )
+
+            # Test the connection with a minimal call
+            if attempt == 0:  # Only test on first attempt to avoid quota usage
+                try:
+                    # Simple test to verify client works
+                    # Note: This is a lightweight validation, actual usage will be in the app
+                    pass
+                except Exception as test_error:
+                    if show_errors and attempt == max_retries:
+                        st.warning(f"⚠️ Vertex AI client created but connection test failed: {test_error}")
+
+            return client
+
+        except Exception as e:
+            last_error = e
+            if attempt < max_retries:
+                if show_errors:
+                    st.info(f"🔄 Vertex AI connection attempt {attempt + 1} failed, retrying...")
+                time.sleep(DEFAULT_RETRY_DELAY * (attempt + 1))  # Exponential backoff
+            else:
+                if show_errors:
+                    st.error(f"❌ Failed to create Vertex AI client after {max_retries + 1} attempts: {last_error}")
+
+    return None
+
+
+def get_anthropic_client(show_errors: bool = True) -> Optional[Union[Anthropic, AnthropicVertex]]:
+    """
+    Get Anthropic client with proper configuration, validation, and retry logic.
+
+    Args:
+        show_errors: Whether to display errors in Streamlit UI
+
+    Returns:
+        Configured Anthropic client or None if configuration fails
+    """
+    try:
+        # Check for Vertex AI configuration first
+        if os.getenv("CLAUDE_CODE_USE_VERTEX") == "1":
+            project_id = os.getenv("ANTHROPIC_VERTEX_PROJECT_ID")
+            region = os.getenv("CLOUD_ML_REGION")
+
+            # Validate configuration
+            validation_error = validate_vertex_config(project_id, region)
+            if validation_error:
+                if show_errors:
+                    st.error(f"❌ Vertex AI configuration error: {validation_error}")
+                return None
+
+            # Get timeout from environment or use default
+            timeout = float(os.getenv("ANTHROPIC_TIMEOUT", DEFAULT_TIMEOUT))
+            max_retries = int(os.getenv("ANTHROPIC_MAX_RETRIES", DEFAULT_MAX_RETRIES))
+
+            # Create Vertex AI client with retry logic
+            if project_id and region:
+                return _create_vertex_client_with_retry(
+                    project_id=project_id,
+                    region=region,
+                    max_retries=max_retries,
+                    timeout=timeout,
+                    show_errors=show_errors
+                )
+
+        # Fallback to direct API key
+        timeout = float(os.getenv("ANTHROPIC_TIMEOUT", DEFAULT_TIMEOUT))
+
+        # Try Streamlit secrets first
+        if hasattr(st, "secrets") and "ANTHROPIC_API_KEY" in st.secrets:
+            api_key = st.secrets["ANTHROPIC_API_KEY"]
+            if api_key and api_key != "using-vertex-ai":  # Skip placeholder
+                return Anthropic(api_key=api_key, timeout=timeout)
+
+        # Try environment variable
+        api_key = os.getenv("ANTHROPIC_API_KEY")
+        if api_key:
+            return Anthropic(api_key=api_key, timeout=timeout)
+
+        # No configuration found
+        if show_errors:
+            st.warning(
+                "⚠️ No Anthropic configuration found. Please set up either:\n"
+                "- Vertex AI: Set CLAUDE_CODE_USE_VERTEX=1 with project/region\n"
+                "- Direct API: Set ANTHROPIC_API_KEY in secrets.toml or environment"
+            )
+
+        return None
+
+    except Exception as e:
+        if show_errors:
+            st.error(f"Failed to initialize Anthropic client: {e}")
+        return None
+
+
+def get_model_name(default: str = "claude-3-haiku-20240307") -> str:
+    """
+    Get the configured model name from environment or secrets with validation.
+
+    Args:
+        default: Default model to use if none configured
+
+    Returns:
+        Model name to use for API calls
+    """
+    # Try environment variable first
+    model = os.getenv("ANTHROPIC_SMALL_FAST_MODEL") or os.getenv("ANTHROPIC_MODEL")
+
+    # Try secrets if no environment variable
+    if not model and hasattr(st, "secrets"):
+        model = getattr(st.secrets, "ANTHROPIC_MODEL", None)
+
+    final_model = model or default
+
+    # Validate model for Vertex AI if being used
+    if os.getenv("CLAUDE_CODE_USE_VERTEX") == "1":
+        validation_error = validate_model_name(final_model)
+        if validation_error:
+            st.warning(f"⚠️ {validation_error}. Using default Vertex AI model.")
+            # Fall back to a known good Vertex AI model
+            final_model = "claude-3-5-haiku@20241022"
+
+    return final_model