Cohere launched a new toolkit to empower developers building AI applications

The Cohere Toolkit is an open-source repository of production-ready applications that integrates with most cloud providers. The applications access Cohere's Command, Embed, and Rerank models on the developer's choice between AWS, Azure, and the Cohere platform, with additional platforms upcoming. Moreover, the applications can also be deployed in private environments to comply with organizations' security standards, and they can also be connected to private data sources. The Cohere Toolkit also includes building blocks for unique, scalable AI application development.

Initially, the Toolkit contains the blueprint to construct a knowledge assistant, similar to the one hosted on the Cohere platform. Knowledge assistants connected to enterprise data and tailored to specific teams are ideal AI applications for enterprises because they boost productivity by simplifying information access, automating routine tasks, and facilitating team collaboration. The Cohere Toolkit contains the tools to ensure that personalized knowledge assistants are capable of understanding conversation intent, remembering conversation history, fulfilling RAG-based tasks, adding fine-grained, relevant citations from private data sources to support their answers, and are endlessly customizable using Cohere’s 100+ pre-built connectors. The availability of every component in a unified packaging spares developers from putting together components from various sources, ensuring they are compatible, and the environment they are deployed in is secure.

The Cohere Toolkit includes plug-and-play components and source code for Cohere's open-sourced interface for its knowledge assistant. The chat UI supports multiturn conversation, fine-grained citations, document upload, and conversation history. It also includes a module enabling interaction with Command R and R+ models in any AI model-hosting platform. Finally, a set of components dedicated to building retrieval systems at the core of retrieval-augmented generation pipelines includes over a hundred connectors with OAuth authentication for enterprise data sources, integrations from popular libraries like LangChain and LlamaIndex, and the ability to use Cohere's Embed and Rerank to improve vector search and retrieval in existing systems.

Platform-specific instructions for AWS, Microsoft Azure, GCP, and more simplify the deployment of the Toolkit. Developers must only direct the code to the models' location to deliver a scalable deployable application, and they are encouraged to customize the application, create and contribute new ones to the repository, or reskin them for easy integration with individual branding. Those interested in trying the Toolkit out should follow the instructions in the GitHub repository. Cohere is also inviting contributions by submitting pull requests to GitHub and has opened registrations for the Cohere Build Days, where the toolkit will be put into action and those interested can join a workshop to build a powerful RAG application at one of Cohere's offices.