Generative AI

Generative AI refers to a class of artificial intelligence techniques and models that are designed to generate new, original content that resembles human-created data.

TotalAgility supports integration with OpenAI and Azure OpenAI generative AI providers, provides the capability for integration with any custom Large Language Model (LLM) and AI Agent. This integration helps create workflows along with associated forms and use them within your solutions. Additionally, TotalAgility provides a ready-to-use Tungsten AI provider, a ready-to-use AI provider for Copilot for development and Copilot for extraction. The Tungsten AI provider also supports the use of images.

OpenAI allows customers to set “custom instructions” in their account. For example, an account may have instructions such as "I am a florist and please restrict your responses to flower-related topics".

TotalAgility does not support accounts that have custom instructions set.

Retry codes

In a Generative AI provider configuration, "retry codes" usually refer to HTTP status codes and specific error types from the provider that trigger automatic retry logic in your client or orchestration layer. The goal is to attempt retries only for transient failures, which include issues like rate limits, timeouts, and temporary outages. It's important to avoid retries for permanent errors, such as bad requests or authentication failures.

HTTP Status Description
408 Request timed out / canceled.
409 Resource state conflict occurred.
429 Rate limit is exceeded.
502 Upstream service error occurred.
503 Service is overloaded/unavailable.
504 Upstream service timed out.

How to:

See also: Generative AI chat control