Cohere Launches Public Beta for Chat API with Retrieval-Augmented Generation

Cohere Launches Public Beta for Chat API with Retrieval-Augmented Generation
Image Credit: Cohere

Cohere, an AI startup building large language models for businesses, has announced the launch of a public beta for its Chat API integrated with retrieval-augmented generation (RAG). This new capability allows developers to build conversational AI products that produce grounded and verifiable responses.

The Chat API makes it easy for developers to integrate Cohere's large language model, Command, into applications to enable smooth and natural conversations. Whether building a virtual assistant, chatbot, or customer service agent, the API simplifies creating reliable AI-powered interactions.

Developers have control over customizing outputs by selecting models, adjusting temperature settings, and leveraging chat history. The Chat API expands Cohere's offerings for creating AI products and features beyond its existing Generate and Summarize APIs.

A major challenge with generative AI is establishing user trust, as incorrect or biased outputs can occur. Cohere aims to address this through RAG, which improves accuracy by connecting the model to external data sources.

Example of how the Chat API with RAG can use a data source to generate verifiable responses with citations

The Chat API is RAG-enabled, allowing developers to anchor model generations to external data points. For the current public beta, the developers can connect Command to web search and plain text documents. This ensures that the generated responses remain relevant, verifiable, and up-to-date. An example use-case highlighted by Cohere is a market research chatbot, capable of searching the web for the freshest insights on market trends and competition.

Coral Chat with the web search grounding connector enabled. In this test, Coral was able to respond accurately about recent events and even referenced a Maginative article as a source.

Adding another layer of trust, Cohere enables the inclusion of citations within generated responses. Such citations are pivotal, not just for transparency, but also to boost the content's credibility. Users can delve deeper into the sources, fostering a richer understanding of the subject.

RAG's strength lies in its ability to tie an AI model like Command to varied information sources. Cohere has focused on optimizing for RAG-oriented tasks like search, results ranking, and citation generation. This ensures that the content remains contextual, relevant, and timely. This eliminates the need for frequent retraining, an often cumbersome task for traditional generative models.

The Chat API public beta is now accessible to all Cohere users. Developers can leverage RAG as an end-to-end solution or use individual modular components (Document mode, Query-generation mode, and Connector mode) to align with their specific needs. The company has also launched a Coral Showcase to preview chat capabilities.

With the launch of its public beta Chat API, Cohere empowers developers to build the next generation of trusted and intelligent conversational AI. RAG integration represents a major step toward mitigating inaccuracies in generative models.

Chris McKay is the founder and chief editor of Maginative. His thought leadership in AI literacy and strategic AI adoption has been recognized by top academic institutions, media, and global brands.

Let’s stay in touch. Get the latest AI news from Maginative in your inbox.