Power-Chat is an artificial intelligence (AI) based chat feature similar to ChatGPT or Bing. However, Power-Chat works on your company's knowledge, it is able to answer queries on the know-how you have aggregated in your Unusual Suite wiki articles and documents.
To start a Power-Chat click the 'Search' icon in the top right next to the 'My account' icon to open a new search tab and click the 'Power-Chat' icon on the top right to enter the Power-Chat window. Submit your queries in the bottom section of the Power-Chat window and start your AI conversation. Power-Chat browses your wiki articles and your documents to find the matching information and calls the external AI to compose a reply. The AI response is displayed with references to wiki articles and documents which have been used to craft the reply. Continue your Power-Chat by refining your inquiry.
To start a new Power-Chat on a different subject click the 'Start a new Power-Chat on a different topic' icon on the top right.
To switch back to Power-Search click the 'Use Power-Search to deep search your data' icon on the top right.
Running Power-Chat on Selected Data Records
Power-Chat also can be run on any explicitly selected set of projects, contacts, organizations, activities, wiki articles, emails, issues, tickets, and documents. Click the Power-Chat icon on the respective UI (projects, contacts, organizations, activities, wiki articles, emails, issues, tickets, or documents) to open a new Power-Chat session.
If a Power-Chat tab is already open, you can add the current data record to the existing session or start a new one. This allows you to configure Power-Chat with any combination of projects, contacts, organizations, activities, wiki articles, emails, issues, tickets, and documents.
Up to 10 of the most recently uploaded linked documents of any data record added to Power-Chat will be added automatically. You can manually add more documents if needed.
Unusual Suite will filter and add up to 10 of the most relevant linked emails of any data record added to Power-Chat. More emails can be added manually if required.
Technical Summary
Power-Chat uses Retrieval-Augmented Generation (RAG) to answer your queries. Your wiki articles and documents are converted to plain text as you save them and the text is converted to a vector of floating point values. In the context of an AI those are called 'embeddings'. These vectors are stored in a database which support searching for vectors. Your query also is converted to such a vector of floating point values and the database tries to find the best matches - this is called 'nearest neighbor' search. The best matches then are fed together with your query to the AI, which composes a response to your query.
Note that this process does not have any memory.
On your follow up questions Unusual Suite needs to feed your prior queries on the same subject together with your current query to the AI.