Toronto-based startup Cohere, founded by former Google employees, has entered the AI chatbot race with the release of a new API. The API, called Chat, allows third-party developers to build chat applications using Cohere’s large language model (LLM), Command. Cohere’s chatbot demo, called Coral Showcase, is also available for users to test. The Coral chatbot, powered by Command, provides accurate and up-to-date responses, although it is slower than some competing chatbots. Cohere’s Chat API features Retrieval-Augmented Generation (RAG), a method that controls a chatbot’s information sources to improve relevance and accuracy. Developers can add web search or plain text documents from their enterprise as additional information sources.
In addition to the Chat API, Cohere also offers three modular components for developers: document mode, query-generation mode, and connector mode. These components allow developers to specify which documents the chatbot references, generate search queries based on user prompts, and connect the chatbot to the web or other information sources. Cohere plans to expand this modular ecosystem in the future. The release of Cohere’s Chat API comes shortly after rival OpenAI reintroduced web browsing capabilities to its ChatGPT, signaling the competitiveness and demand in the AI chatbot market.
Overall, Cohere’s entry into the AI chatbot market with its Chat API and the use of RAG technology provides developers with an opportunity to build reliable and accurate conversational AI products. While the Coral chatbot may be slower than some competitors, it delivers accurate responses without visible hallucinations. Cohere’s modular components offer flexibility and customization options for developers, allowing them to enhance their chatbot’s performance. The market for AI chatbots is growing, and Cohere’s release positions the company as a leading player in this crowded space.