memory-ingest

Paste the transcript and ask for it to be processed into Basic Memory. The workflow extracts people, organizations, and topics, searches for existing matches, proposes new entities for your approval, then creates a source note with verbatim content plus categorized observations and relations.

Each extracted entity is searched with multiple query variations, such as full name, last name, abbreviations, and acronyms. Entities found in Basic Memory are linked with wiki-links instead of being recreated, and only unmatched entities are proposed as new.

Yes. Before anything is written, the workflow presents a proposal listing existing entities and proposed new ones with context for each. You can approve all, select individually, or skip entity creation entirely.

Meeting transcripts, conversation logs, email threads, pasted documents, articles, and freeform notes are supported. The workflow detects the format and applies an appropriate note template, such as meeting notes or reference notes.

No. Source content is preserved verbatim in the note as the ground truth. Structured observations, categories, and relations are layered on top as interpretation without rewriting or summarizing the original material.