Frequently Asked Questions (FAQs)

What is Medullar?

Medullar is an advanced AI-powered universal search platform designed to enhance productivity and decision-making. The platform connects to over 60 applications, offering secure and private data interactions through innovative search optimization and collaborative features.

How does Medullar's universal search tool work?

Medullar's universal search tool connects to your personal data sources using APIs and federated search technology to search across various connected platforms simultaneously, providing relevant results from emails, documents, chats, and more in a unified interface.

What is federated search technology?

Federated search technology allows users to search multiple applications simultaneously without moving or storing the underlying data. This ensures that information remains in its original location while retrieving results from various connected platforms in a unified interface.

How does Medullar differentiate itself from other enterprise search platforms?

Medullar differentiates itself from other enterprise search platforms through its federated search approach and AI-powered features to support personalized/collaborative interactions between each user’s personal knowledge graph. Unlike other platforms, Medullar performs real-time searches across 60+ apps without moving / migrating / centralizing data, ensuring enhanced privacy and security. Medullar’s AI-powered features (ex., semantic reranking and contextual app selection) provide highly relevant and timely results tailored to user needs.

Medullar also offers best-in-class privacy features, such as AES-256 end-to-end encryption and granular access controls to differentiate from other, less secure competitor platforms.

What apps does Medullar support?

Medullar supports a wide range of over 60+ apps including Asana, Box, Crunchbase, DocuSign, Google (Drive, Calendar, Mail), HubSpot, Microsoft 365 (Outlook, OneDrive, SharePoint, Teams), New York Times, Notion, Salesforce, Slack, and more. View the entire app catalog here.

Can Medullar search through attachments?

Yes - Medullar can search through the content of attachments in supported applications and file formats.

What types of file formats can Medullar search through?

Medullar can search through a wide variety of file formats, including documents, emails, chat logs, and files stored in cloud services.

Can I filter search results in Medullar?

Yes - Medullar offers various filtering options to refine search results by date, file type, source, and more.

How does Medullar perform relevancy ranking?

Medullar enhances its relevancy ranking by leveraging Large Language Models (LLMs), primarily Cohere ReRank. Cohere ReRank semantically re-ranks search results based on their contextual relevance to the query. By understanding the meaning and nuances of the query and results, the LLM prioritizes the most meaningful information rather than just relying on keyword matches. This semantic reranking ensures that users get the most contextually appropriate results first.

In addition to semantic reranking, Medullar employs basic Natural Language Processing (NLP) techniques to deduplicate results. These NLP techniques identify and remove redundant information from different sources, ensuring users receive distinct, high-quality results without repeated entries. This combination of LLM-powered reranking and NLP-based deduplication creates a more streamlined and efficient search experience.

Does Medullar offer AI summarization of search results?

Yes - Medullar provides AI summarization for all major connectors to enhance user productivity. The summary is embedded directly in the search result to save users a step in the review process. Auto-summarization can be turned on / off under user settings.

How are AI summaries generated?

Medullar enables users with the ability to select their preferred LLM (ChatGPT, Claude, Llama, etc) to generate summaries. To change LLMs access Workspace > Settings (NOTE: ChatGPT LLMs are used by default).

How accurate are the summaries generated by Medullar?

The accuracy of the summaries depends on the complexity of the document and the effectiveness of the underlying AI algorithms, which are designed to produce relevant and coherent summaries.

Does Medullar support Natural Language Processing (NLP) queries?

Yes - Medullar employs large language models (LLMs) to convert user inputs into natural language queries for improved search accuracy and efficiency.

How does Medullar's search tool deal with misspellings?

Medullar handles misspellings through a pre-processing step as part of its search workflow. When a user enters a query, Medullar cleans and spell-checks the input. This step automatically detects and corrects spelling errors before further processing the query. Once cleaned, the query is passed to an LLM, which interprets it and does a second spelling check.

If a misspelled word is found in the pre-processing step, Medullar gives the user the option to search for the misspelled word, like Google. (ex. “Showing results for twitter. Search instead for twiter”)

For users with large numbers of data sources, what is the impact on query speed / response time? i.e. is Medullar only as fast as its slowest content source?

Medullar depends on the slowest source to show results, but each source is assigned a timeout to avoid excess waiting. To further improve latency, Medullar developed a “Smart App Chooser” process that leverages LLMs to understand search context and determine which app(s) are best suited to answer the user’s question.

For example - if searching for a document file type with both Dropbox and NYT connectors enabled, Medullar’s “Smart App Chooser” will search only through BOX and disregard to improve search efficiency. For more information on this feature please visit: https://www.medullar.com/us/en/blog/medullar-smart-app-chooser.