In one sentence
A POT (Knowledge Pot) is a specialized container where all knowledge for one account or domain lives: documents, meeting transcripts, extracted facts, and the relationships between them. 1 account = 1 POT.Three letters, three meanings
“POT” has a deliberate three-layer etymology:
When you read “the POT” or “your POT” — that’s the container. When you read “POT Score 0.92” — that’s the confidence metric.
What a POT contains
- Constitution: a small set (typically 3–7) of constitutional axioms — the foundational facts of that account, with POT Score = 1.0. Only a Human Curator can modify them.
- Documents: the PDFs, DOCX, Markdown, and images you uploaded.
- Transcripts: from meetings recorded with kb2b Desktop.
- Extracted facts: automatically from documents and transcripts, each one with its POT Score.
- Edges (relationships): typed connections between facts — supports, contradicts, refines, derives from, etc.
- Conversations: the chat history about the account.
POT Score — confidence per fact
Every fact inside the POT carries a POT Score between 0.0 and 1.0. The SciPot whitepaper defines five levels:
When kb2b answers, the score tells you how confident it is. Transparency, not marketing.
More on POT Scores
The full POT Score mechanic — propagation through the knowledge graph, temporal decay, and human curation — lives in the official SciPot documentation.
When a POT is created
- When you create your kb2b workspace, a main POT is created automatically.
- When you create a new account (a customer, a project), a new POT is created.
- 1 account = 1 POT, always. No multi-POT per account.
When the POT model doesn’t quite apply
- If your workspace is for open research (you don’t have “accounts” in the B2B sense), use the main POT as inbox.
- If your team works by topics rather than accounts (legal works by cases, not by clients), name your POTs as cases: “Case Customer-X 2026-Q2”.