curate-memory

Automatically deduplicate chat transcripts by scanning conversations against existing memory entries, routing content to memory journals, notes, or learnings, and merging fragments into existing topics to maintain a coherent knowledge base.

Classify long-running conversation logs using context-tree guided routing to determine whether entries belong in MEMORY.md, memory journals, notes, references, or learnings, ensuring safe deterministic writes to memory structures.

Conversation curation handles overlapping or conflicting technical decisions by scanning for duplicates, resolving conflicts, and merging fragmented content into existing topics to prevent knowledge base fragmentation.

Yes, you can use context-tree guidelines to automatically route entries to MEMORY.md and learnings without manual curation, as the system applies deterministic classification rules to organize conversation data safely.

Yes, a context tree is required to merge fragmented conversation notes into existing topics, as it provides the necessary guidelines to classify, deduplicate, and route entries to the correct memory structures.

Knowledge base fragmentation occurs when overlapping technical decisions are not deduplicated and merged into existing topics; automated curation scans for these overlaps and consolidates fragments to maintain searchable coherence.