How are connectors different from other integrations?

Connectors are a type of integration. Starmind Connectors imports data related to the user's knowledge from other applications and enhances their profiles on Starmind. In other words, connectors bring a user's know-how from different applications to Starmind. In contrast, Starmind integrations surface Starmind outputs such as Questions/Answers and Experts into other applications to improve that workflow.

When should connectors be activated?

Connectors should be enabled at the beginning of the Starmind lifecycle. They eliminate the "cold start" problem, which occurs when the Starmind platform lacks information about a new user to build an expertise profile. When connectors are enabled, Starmind's algorithm can automatically create expertise profiles for every employee, regardless of whether they have interacted with the Starmind platform.

How does Starmind ensure data confidentiality and data privacy?

We only include publicly available data and relevant information, filtering out non-relevant information. We do not learn from private data, such as a private MS Teams conversation between two people.

We are requesting only the essential API permissions that we need.

We do not store the whole text document. We only extract key topics associated with a user's knowledge. Starmind only stores these extracted topics in the user profile.

Can we use the same connector to learn about a user's expertise and improve StarGPT answers?

Yes, this is solved with the same Connector. The Connector is the part that integrates with the platform's API. We can then utilise the same information (e.g., SharePoint documents) to enhance the Knowledge Engine and/or index it with StarGPT to improve the quality of answers.

Can we implement a custom connector that is not part of Starmind's out-of-the-box options? And what makes a good data source to derive expertise from and improve the Knowledge Engine?

Yes. We have an API for this.

Generally, all data sources where data can be clearly linked to a user on Starmind are considered reliable sources. For example, Confluence Pages, with their wiki-style format, serve as a good example. They clearly associate content with an author, allowing us to confidently identify the user as a subject matter expert for the page's topics.
PDFs are not an ideal source of information because we cannot verify that the uploader is also the author. If the uploader is not the author, the inferred connections may not be accurate.

The four essential requirements for data to be useful are:

• Each entry or item contains consistent and relevant text from which we can extract topics or the topics directly.
• Each entry or item has at least one person associated with it who can be unambiguously identified (for example, by email address).
• Each entry or item has a date and time when it happened.

Examples of data sources that are not good sources for Starmind:

• Videos or images, as we cannot extract topics from them.
• PDFs, as it may be difficult to associate the content with specific people correctly.

Technical

From which IP address does the Starmind Connector access my data source?

The infrastructure for indexing data from external data sources (e.g., SharePoint) is hosted in the same datacenter as your Starmind network. This ensures that your data is kept secure in the same location and doesn't leave the premises.