Working with the Document Base
The agent answers not only from the tracker and knowledge base — it also answers from your documents: specifications, contracts, protocols, procedures. This guide covers how to bring documents into the system and use them.
How to upload documents
Documents are uploaded straight into the conversation: attach a file or drag it into the chat window. The system parses the document (including tables) and adds it to the knowledge base.
Text and office formats are supported: PDF, DOCX, Markdown, TXT, and others. After upload, a document becomes searchable after a few minutes while the index is built.
Three access levels
| Level | What it is | When to use it |
|---|---|---|
| Session documents | Materials of the current conversation: drafts, attached files | Work inside a single conversation |
| Project documents | The durable project base, available to all participants and across all conversations of the project | Documents that must persist between sessions |
| Global corpus | Standards and materials shared across projects (PMO level) | Corporate standards, methodologies |
From session to project: promote
By default, a document uploaded into a conversation lives in the session. When a document is ready to become part of the project, the agent proposes promoting it to the project base. The promotion is confirmed by you: the agent does not publish session drafts to the project without an explicit decision.
How to search the documents
Search works by meaning, not by exact word match. Ask questions in plain language:
- "What does the uploaded specification say about acceptable API response latency?"
- "Find the force-majeure notice-period requirement in the contract."
- "What access restrictions to the test environment are described in the procedure?"
In its answer, the agent quotes the document fragment and gives a source reference — you can open the original and verify. If several documents contain the answer, the agent compares them and shows the discrepancies.
Document registry: /documents
Besides the search base, the system keeps a registry of project documents — a structured list of which project documents exist. Type /documents (or upload a file and ask it to be classified):
- the agent determines the document type (charter, scope, BRD, specification, contract, risk register, plan, and others);
- it adds a short note on which project question the document covers ("covers");
- it shows which document types are still missing.
A type that does not apply to the project (for example, a contract for an internal project) is marked "not applicable" with a reason — the agent will stop reminding you about it.
Good to know
- The agent reads documents in full, but in its answers quotes only the relevant fragments — full texts are not disclosed outside your environment.
- A revised document: upload the new version — the index updates and search uses the fresh version.
- Project documents are distinguished by level: a session draft and a project document are different things, and moving between them is a deliberate action.
Where to continue
- Project documents: creation and upkeep — documents the agent creates itself.
- Knowledge library — how the knowledge base works as a whole.
- Navigator — the working mode where document search is the primary tool.