Locusive Is Currently Free

Locusive utilizes enterprise-grade security protections for your business data including access controls, encryption, data isolation, and compliance with regulations like CASA Tier 2. Only specific authorized datasets are synced with Locusive. Client data is separated logically. All data is transferred over SSL, and we implement a policy of least privilege.

‍

We’ve also taken the following steps to work on ensuring data privacy and security:

‍

Data storage:

• We don’t store the entire contents of a document anywhere. Rather, we break it up into chunks and store those chunks in a vector database. That vector database does not provide an index of what documents are available, and that information is stored in a separate database. This helps ensure that even if one database was compromised, it would be difficult to get access to any data.
  ‍
• We limit access to all data stores to the top personnel at Locusive who need to be able to work with data systems
  ‍
• All data is partitioned by customer so that there is no intermingling of documents between customers

Data transmission

• Data is transmitted securely with SSL

Data access

• Locusive users make their documents available to the users in their Slack channel, and only those users have access to query documents that have been explicitly added by their org’s admins
• For uploaded documents, we generate a one-time, signed URL that expires after a few minutes. This URL is only generated at the time that an authorized user explicitly wants to view a document they have uploaded.

Yes! Unlike with other AI systems, LLMs have already been pre-trained on lots of data, so you don't need to have millions of pages of documents to use them. Our system will work with whatever information you have, even if it's just a one-page PDF.

You select the apps and data sources that you want our system to know about, securely connect them to our system, and then start asking questions. Our system will find content from your connected data sources that's likely to have the answer to your question and then use a large language model like ChatGPT to provide the final answer.

To use ChatGPT with your own data sources, you have three options:

1. Training an LLM from scratch on your dataset
2. Fine-tuning an existing model with your data
3. Use an existing LLM by providing it context from your data (known as retrieval augmented generation or RAG)

Most companies don't have the resources to train their own LLM from scratch, and fine-tuning can be a lot of work for potentially no benefits. We generally recommend using retrieval augmented generation.

Yes! We refresh public and private PDFs every 30 days, and we'll check all other sources (webpages, Google Sheets, etc.) every day. When we detect there has been a change, we'll update our indexes to make sure we have the latest information.

Locusive works best with long-form text data or structured data like datasets or Google Sheets. You might have trouble using it with poorly formed data (for example, insurance policy documents that have unusual layouts, or PDFs that have been converted from slides that have strange layouts or only images).

Locusive's Concierge will be an assistant that customer-facing teams in SaaS companies can use to handle user requests automatically.

The system takes your query and then looks through all your content to find the most relevant paragraphs and sentences. It then iteratively sends those paragraphs and sentences to ChatGPT to find the answer. So it's possible that just having a single document that contains the answer to your question will be sufficient. However, if your question is extremely complicated or if it's phrased in a way that makes it hard for the system to find the right answer, it may require a few more documents. And finally, if the answer isn't available anywhere in the documents, or if the system is unable to find areas in your documents where the answer might reside, then no amount of documents will help.