mirror of
https://github.com/vegu-ai/talemate.git
synced 2025-12-18 04:37:47 +01:00
23 lines
1.1 KiB
Markdown
23 lines
1.1 KiB
Markdown
|
|
# Testing Embeddings
|
||
|
|
|
||
|
|
You can test the performance of the selected embedding, by using talemate normally and then inspecting the memory request in the debug tools view.
|
||
|
|
|
||
|
|

|
||
|
|
|
||
|
|
Once the debug tools are open, select the :material-processor: Memory tab.
|
||
|
|
|
||
|
|
Then wait for the next talemate generation (for example conversation) and all the memory requests will be shown in the list.
|
||
|
|
|
||
|
|

|
||
|
|
|
||
|
|
In this particular example we are asking Kaira when we first met, and the expectation is for the memory agent to find and return the memory of the first meeting.
|
||
|
|
|
||
|
|
Click the memory request to see the details.
|
||
|
|
|
||
|
|

|
||
|
|
|
||
|
|
Up to 10 results are shown, however only those that fall within the acceptable distance are included in the context.
|
||
|
|
|
||
|
|
Selected entries will have their distance function colored green, while the others will be going from yellow to red.
|
||
|
|
|
||
|
|
If you find that accuracy is lacking you may need to tweak the [Embedding settings](/talemate/user-guide/agents/memory/embeddings).
|