Getting Started
Memory vs Knowledge
Decide whether your data belongs in a knowledge engine, a memory space, or both.
Knowledge is what your organization knows. Memory is what your app learns about the people and work it handles. Most data clearly belongs to one of them, and most assistants use both.
Side by side
| Knowledge Engines | Docklee Universal Memory | |
|---|---|---|
| Holds | Documents, pages, records, and app data | Facts, preferences, events, and conversation history |
| Example | A refund policy, a product manual, a help article | "Maya prefers email", "The customer renewed in March" |
| Written by | Sources in the dashboard, and Dock with a container | Dock with a memory_space_id |
| Stored in | Vector and OKF containers | Memory spaces |
| Read through | An engine, with lynk.query or lynk.retrieve | A memory space, with lynk.recall |
| Returns | An answer with citations, or ranked chunks | Memory records with a type and a confidence score |
| Per end user | No. An engine answers the same for everyone who can reach it. | Yes, with end_user. |
| Changes when | The source content changes | Your app captures something new |
Choose Knowledge if
- The content already exists as documents, pages, or app records.
- Everyone asking the same question should get the same answer.
- Answers must cite the source they came from.
Choose Memory if
- The data is about one user, customer, or conversation.
- Facts change, and a newer fact should replace an older one.
- Each of your end users must only ever see their own data.
Use both
A support assistant answers policy questions from an engine and remembers each customer in a memory space. One key with both runtimes reaches each. This example reads DOCKLEE_ENGINE_ID and DOCKLEE_MEMORY_SPACE_ID from the environment:
from docklee import Lynk
lynk = Lynk()
policy = lynk.query_sync("What is the refund window?")
history = lynk.recall_sync("past refund requests", end_user="user_8f2c")Put the answer and the customer's history in the same prompt, and the reply is both correct and personal.
Next steps
Once you know where your data goes, see how each path works end to end.