Add auto container tags feature#10
Conversation
![vorflux[bot]](https://avatars.githubusercontent.com/in/2483280?s=60&v=4){kind=small}
| additionalContextLength: additionalContext.length, | ||
| }); | ||
| exitWithContext(additionalContext); | ||
| } else { |
There was a problem hiding this comment.
Container catalog is only injected when newFacts.length > 0. If recall returns zero new facts (all previously seen), the else branch emits empty context and the AI never learns about custom containers for that turn. Fix: inject the catalog independently of new facts.
|
|
||
| const client = new SupermemoryClient(); | ||
| const projectTag = getProjectTag(process.cwd()); | ||
| const effectiveTag = containerTag || getProjectTag(process.cwd()); |
There was a problem hiding this comment.
No validation that containerTag matches a configured customContainers[].tag. A typo like --container wrk silently creates an orphaned space in Supermemory that wont appear in searches.
|
|
||
| const content = rawContent | ||
| .replace(/\[SUPERMEMORY CONTAINERS\][\s\S]*?\[END SUPERMEMORY CONTAINERS\]\s*/g, "") | ||
| .replace(/<supermemory-containers>[\s\S]*?<\/supermemory-containers>\s*/g, "") |
There was a problem hiding this comment.
This regex strips supermemory-containers XML tags but getContainerCatalog() only emits the [SUPERMEMORY CONTAINERS] bracket format. This XML variant doesnt exist anywhere in the codebase.
1 participant
Summary
Custom Memory Containers: Users can now organize memories into specific tags (e.g., code_preferences, personal, architecture) via a new --container flag in the save, search, and forget skills.
Dynamic Container Catalog: Available custom containers are now injected into the model's context, allowing the agent to proactively choose the right storage location.
The below image shows the save working with container tags the search working and in the userpomptsubmit hook container tags are visible