CRM AI Assistant Optimization

We have an existing internal CRM used by our company to manage customers, leads, notes, and sales communication. The CRM includes an AI assistant that summarizes customer activity and generates suggested responses for our sales team.

The AI functionality is already implemented, but it has recently become unreliable. Some requests fail, generated responses are inconsistent, and API/model errors are not handled correctly.

We need an experienced Python/AI developer to review the existing integration and fix the current issues.

The work includes reviewing the existing FastAPI endpoints and AI service, debugging the OpenAI/LangChain integration, correcting model response parsing and prompt handling, improving validation and exception handling, and ensuring API credentials/configuration are managed securely through environment variables.

The existing CRM frontend, database structure, customer records, and normal CRM functionality should remain unchanged.

The implementation should also keep the AI provider layer reasonably modular so Anthropic or a local Ollama model could be added later without redesigning the CRM.

Expected deliverables:

Working CRM AI summary feature
Working AI suggested-reply feature
Correct OpenAI API integration
Improved prompt and response handling
Timeout and API error handling
Environment-based API configuration
Basic testing of existing CRM workflows
Changes pushed to our GitHub repository
Short setup/configuration documentation