reliableGPT

ReliableGPT is a powerful tool designed to ensure zero dropped requests for your LLM (Language Model) app in production. It handles errors effectively by employing various strategies such as retrying with alternate models, larger context window models, semantic similarity-based cached responses, and fallback API keys.

Key Features:

Alternate Model Retry: Retry failed requests with alternate models such as GPT-4, GPT3.5, GPT3.5 16k, or text-davinci-003.

Larger Context Window Models: Retry requests with larger context window models to address Context Window Errors.

Semantic Similarity-based Cached Response: Provide cached responses based on semantic similarity to handle errors efficiently.

Fallback API Key Retry: Retry requests with a fallback API key in case of Invalid API Key errors.

Switch between Azure OpenAI and raw OpenAI: Seamlessly switch between Azure OpenAI and raw OpenAI to meet your specific requirements.

Caching for Overloaded Servers: Handle overloaded servers with caching mechanisms to ensure smooth operation.

Rotated Key Handling: Effortlessly handle rotated keys to avoid disruptions in service.