Highlights:

  • Ragie enables RAG-enabled LLMs to access data from widely used cloud applications like Salesforce, Google Drive, and Notion.
  • Ragie offers a chunking tool that breaks documents used by RAG-enabled LLMs into smaller, more manageable pieces.

Recently, a new startup supporting developers in building retrieval-augmented generation applications, Ragie Corp., launched with USD 5.5 million in seed funding.

The funding was provided by Craft Ventures, Saga VC, Chapter One, and Valor.

Retrieval-augmented generation (RAG) is a machine learning technique that simplifies the process of updating large language models with new information. Traditionally, expanding an LLM’s knowledge base required retraining, a process that can be very costly. RAG allows additional data to be incorporated into the model without the need for retraining.

While the technology simplifies certain aspects of AI projects, implementing it still requires considerable effort. Consequently, developing RAG-enabled applications can sometimes take months. San Francisco-based Ragie has created a cloud platform designed to streamline this process.

Ragie enables RAG-powered LLMs to access data from popular cloud applications like Salesforce, Google Drive, and Notion. According to the company, adding new data sources through its platform requires just a few clicks. Ragie also tracks changes in the datasets it processes and automatically updates the LLMs that utilize this information.

Ragie’s platform doesn’t directly stream raw information to AI models. Instead, it converts the data into embeddings, which are mathematical representations that LLMs use to store knowledge. These embeddings are then stored in a cloud-based vector database specifically designed for such files.

Ragie also offers a tool called chunking within its feature set. Chunking involves breaking down documents used by RAG-enabled LLMs into smaller, more manageable segments. This reorganization of datasets can enhance the quality of the AI models’ outputs.

A reranking algorithm is another essential component for developing an RAG application. These algorithms assess the documents LLM references to answer a question, prioritizing them by relevance to ensure that only the most pertinent information is included in the model’s response. Ragie’s platform consists of a built-in reranking feature, saving developers valuable time.

The platform also includes several advanced features. One allows for the creation of AI chatbots that can retrieve information from multiple sources when responding to user queries. Another enhances RAG applications’ ability to carry out data extraction tasks, like pinpointing all the earnings reports within a large set of business documents.

The company announced its funding round alongside the official launch of its platform into general availability. Ragie offers a free tier that lets developers process up to 10 RAG requests per minute at no cost. Revenue comes from two paid versions of the platform, Pro and Enterprise, which offer higher rate limits.