Use OKF
OKF in memory
How Docklee also turns a memory space into a concept graph, so recall can reason over concepts as well as single memories.
Individual memories answer "what do we know about Maya?". A question like "how is Maya's team connected to the Northwind renewal?" needs the concepts behind those memories and how they relate. Docklee builds a concept graph from each memory space, using the same OKF approach as knowledge.
How it works
- Compile. Docklee turns a memory space's memories into typed concepts and the relationships between them, with the same generation steps used for knowledge. See Generate OKF.
- Store. The concepts are encrypted with your workspace's key and stored apart from your knowledge bundle. Knowledge and memory never share storage.
- Index. Each concept is indexed for search, tagged with its memory space, so a search never crosses spaces.
- Recall. When you call
lynk.recall, Docklee searches the space's memories and its concept graph, and combines the two.
What changes for you
Nothing in your code. lynk.recall returns memory records as it always does. The concept graph adds related context, and if it can't be searched, recall still returns the memories themselves.
records = lynk.recall_sync("How is Maya's team connected to the Northwind renewal?")Knowledge and memory stay separate
| Knowledge OKF | Memory concepts | |
|---|---|---|
| Built from | Sources and Dock writes to container="okf" | A memory space's memories |
| Stored | Your workspace's OKF bundle | Separately, per memory space |
| Read by | Engines, through query and retrieve | recall on that memory space |
| Exported | Yes | No |
Next steps
You've seen OKF across knowledge and memory. Explore how memory reasons over time next.