AI Assistant
Answers clients and staff strictly from your own documents: it searches the loaded knowledge base by meaning and does not invent what is not there.
Knowledge base
The answer is assembled from your own documents. The technique is called retrieval-augmented generation: first the relevant fragments are found, and the answer is built only from them.
-
Loading documents
PDF, DOCX, XLSX, TXT and Markdown, dragged into the window
-
Automatic processing
text extraction, splitting into fragments, vectorisation; the state of every file is visible
-
Answers from the documents
search by meaning that takes in the history of the dialogue, delivered as it is generated or all at once
Channels
-
A Telegram bot for the organisation
private and group chats, voice messages, access by key or open
-
Web widget
added to a site with a single fragment of code; text and voice
-
A card page
a separate brandable page for the assistant with its own access token
-
REST and MCP
connecting outside systems and AI clients to the knowledge base
Management
-
Administrator panel
the knowledge base, the settings of the bot and the model, access, statistics
-
Operator console
a live operator takes the dialogue over from the assistant and answers themself
-
Several organisations
the knowledge base, the bot and the settings of each are isolated from the rest
Platform
-
Russian jurisdiction
the language model, the vectorisation and the speech recognition are domestic, and the data is processed in Russia
-
Domestic operating systems
confirmed to run on Astra Linux, ALT Linux and RED OS
-
No client to install
a web application in any modern browser
Projects
-
Rosneft Pavilion — “Russia” Expo 2023
/2023 — Events
-
VK Store (Sochi)
/2023 — Exhibitions
-
History of Fortification
/2023 — Museums
-
UAC Irkut — NAIS
/2023 — Events
-
Central Election Commission — “Russia” Expo 2023
/2023 — Exhibitions
-
“Battlefield Mariupol” Museum
/2023 — Museums
-
Gazprombank
/2023 — Events
-
SSJ100VR — UAC Irkut
/2023 — Exhibitions
