fix: add DoS protection to OpenAI API calls

[deosha]

Summary

This PR adds DoS protection to the get_chat_completion function to prevent API abuse and cost explosion.

The Problem

The get_chat_completion() function in services/openai.py lacks:

• Request timeout configuration
• Token/response limits
• Input length validation

This allows attackers to send extremely large inputs or trigger expensive API calls without limits.

The Fix

Added configurable protections:

• OPENAI_TIMEOUT - Request timeout (default: 30 seconds)
• MAX_TOKENS - Maximum response tokens (default: 500)
• MAX_INPUT_LENGTH - Maximum input characters (default: 10,000)

All settings are configurable via environment variables for flexibility.

Changes

# Before
response = openai.ChatCompletion.create(model=model, messages=messages)

# After  
if total_input_length > MAX_INPUT_LENGTH:
    raise ValueError(f"Input length exceeds maximum")
    
response = openai.ChatCompletion.create(
    model=model,
    messages=messages,
    max_tokens=max_tokens,
    timeout=timeout,
)

---

Identified using aisentry, an AI/LLM security scanner for OWASP LLM Top 10 vulnerabilities.