feat: Add support for Anthropic API

[ProCreations-Official]

This commit introduces support for the Anthropic API, specifically for the claude-sonnet-4-20250514 model with a 1M context window.

The changes include:

• Adding anthropic to the list of dependencies in requirements.txt.
• Updating main.py to integrate Anthropic as a new AI provider alongside Gemini and OpenAI.
• Adding configuration for the ANTHROPIC_API_KEY.
• Implementing the _enhance_with_anthropic function to handle API calls to Anthropic, including the necessary betas parameter for the 1M context model.
• Updating model selection logic, error handling, and help text to include the new provider.

[claude]

Claude finished @ProCreations-Official's task —— View job

---

PR Review: Add Anthropic API Support

I've completed a thorough review of this PR. Overall, the implementation is well-structured and follows the existing patterns, but there are several important issues that should be addressed.

Todo List:

• Read repository guidelines (CLAUDE.md) - No CLAUDE.md file found
• Analyze changes in main.py
• Analyze changes in requirements.txt
• Check code quality and best practices
• Identify potential bugs or issues
• Review performance considerations
• Assess security concerns
• Evaluate test coverage
• Provide comprehensive feedback

---

🔍 Code Review Findings

✅ Positive Aspects

1. Consistent Architecture: The implementation follows the existing pattern established for OpenAI integration
2. Proper Error Handling: Good validation and error messages for missing API keys and unsupported models
3. Optional Dependency: Graceful handling when Anthropic library isn't installed
4. Model Documentation: Clear documentation of the 1M context window capability

⚠️ Issues Found

Critical Issues

1. Incorrect Beta Parameter (main.py:619)

   betas=["context-1m-2025-08-07"]

   • Issue: This beta identifier appears to be fabricated. Anthropic's actual beta for context expansion is typically named differently
   • Risk: API calls will likely fail with invalid beta parameter
   • Fix: Use the correct beta identifier from Anthropic's documentation or remove if not needed

2. Resource Leak (main.py:622)

   await client.close()

   • Issue: Client closure only happens in success path, not in exception cases
   • Risk: Connection leaks when exceptions occur
   • Fix: Use async with context manager or ensure cleanup in finally block

Security Concerns

3. API Key Exposure Risk (main.py:36)

   ANTHROPIC_API_KEY = os.getenv("ANTHROPIC_API_KEY")

   • Issue: While using environment variables is correct, there's no validation that the key isn't accidentally logged
   • Recommendation: Consider implementing key masking in logging contexts

Performance & Reliability Issues

4. Low Token Limit (main.py:615)

   max_tokens=4096

   • Issue: Very low compared to OpenAI's 24576 tokens. For a 1M context model, this severely limits response quality
   • Impact: Users won't get the full benefit of the larger context window
   • Fix: Increase to at least 8192 or match OpenAI's limit

5. No Timeout Configuration (main.py:613-620)

   • Issue: Missing timeout parameter in Anthropic API call
   • Risk: Potential hanging requests
   • Fix: Add reasonable timeout (30-60 seconds)

6. Exception Handling Inconsistency (main.py:629-631)

   except Exception as e:
       logger.error(f"Anthropic enhancement failed: {e}")
       raise

   • Issue: This will crash the entire request instead of falling back gracefully like other providers
   • Fix: Return fallback documentation like Gemini/OpenAI implementations

Code Quality Issues

7. Missing Input Validation

   • Issue: No validation of model_name parameter in _enhance_with_anthropic
   • Risk: Could cause cryptic API errors
   • Fix: Validate model name before API call

8. Hardcoded Model List (main.py:76-78)

   available["anthropic"] = [
       "claude-sonnet-4-20250514" # 1M context window
   ]

   • Issue: Hardcoded model name may become outdated
   • Recommendation: Consider dynamic model fetching or configuration

Documentation & Usability

9. Model Description Clarity (main.py:123-124)

   if model == "claude-sonnet-4-20250514":
       description += f'               - "{model}": 1M context window\n'

   • Issue: Doesn't indicate speed/quality trade-offs like other models
   • Fix: Add consistent quality/speed indicators

🧪 Test Coverage

• Missing: No tests for the new Anthropic integration
• Recommendation: Add unit tests covering:
  • Anthropic client initialization
  • Error handling for missing API keys
  • Model validation
  • Beta parameter handling
  • Response processing

📋 Recommended Changes

High Priority:

1. Fix the beta parameter with correct value from Anthropic docs
2. Implement proper resource cleanup with context managers
3. Increase max_tokens to reasonable limit (8192+)
4. Add timeout configuration
5. Fix exception handling to return fallback content

Medium Priority:
6. Add input validation for model names
7. Add unit tests for Anthropic integration
8. Improve model description consistency

Low Priority:
9. Consider API key masking in logs
10. Evaluate dynamic model discovery

🔧 Example Fix for Resource Management

async def _enhance_with_anthropic(prompt: str, model_name: str, library_id: str, docs: str) -> str:
    """Enhance documentation using Anthropic."""
    try:
        async with AsyncAnthropic(api_key=ANTHROPIC_API_KEY) as client:
            response = await client.beta.messages.create(
                model=model_name,
                max_tokens=8192,  # Increased from 4096
                timeout=60.0,     # Added timeout
                messages=[{"role": "user", "content": prompt}],
                betas=["correct-beta-name"]  # Use actual beta name
            )
            
            if response and response.content and response.content[0].text:
                return response.content[0].text
                
    except Exception as e:
        logger.error(f"Anthropic enhancement failed: {e}")
        # Return fallback instead of raising
        
    return f"# Documentation for {library_id}\n\n{docs}\n\n*Anthropic enhancement unavailable at the moment.*"

The implementation is solid foundational work, but addressing these issues will make it production-ready and consistent with the existing codebase quality standards.

---