What it does
The included benchmark uses a small fictional memory set designed around semantic-search misses, so its comparison should not be generalized without broader data.
Integrations
Retrieve memories for their consequence to the current request rather than their embedding similarity.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
The included benchmark uses a small fictional memory set designed around semantic-search misses, so its comparison should not be generalized without broader data.
You can use this tool to pick out important memories before making a decision. If someone orders dessert, the app checks past notes for a nut allergy. Your developer can use this tool to send a situation and several memories to TypeSafe.
The outside AI service scores each memory. It keeps only the notes that matter. To connect the software, your developer must provide an access key from TypeSafe. The tool limits how much text you can send in one request.