Cognify documentation#
Useful links: Installation | Source Repository | Blog Post
Cognify is a fully automated optimizer for generative AI (gen-AI) workflows. It transforms user-provided workflows into a set of optimized versions with diverse quality-cost-latency trade-offs. Cognify performs optimizations on individual steps within a workflow and also the workflow structure, with a multi-layer hierarchical optimization framework. Cognify offers improvements of up to 2.8x in generation quality, up to 10x cost reduction, and up to 2.7x reduction in end-to-end latency.
Getting started
New to Cognify? Check out our hello world example to get started. It shows how to quickly set up Cognify and optimize a simple workflow.
Tutorial
The tutorial explains each step involved in using Cognify. It covers workflow connector, evaluator, dataloader, optimizer configuration, the CLI, and how to interpret the results.
Fundamentals
The fundamentals section provides an in-depth explanation of the key concepts and features of Cognify.
API Reference
Want to dive into the code? The reference provides detailed information on the functions, classes and data structures used in Cognify.