episodic-archiver

You embed message turns and store full transcripts in ArangoDB to index conversation histories. This applies timestamps and category labels, enabling rapid search, retrieval, and auditability for long-running sessions.

The best way to index chat transcripts for memory-assisted retrieval is by generating per-turn embeddings and storing them with dedupe keys in ArangoDB. This preserves user stories and enables searchable memory across sessions.

Yes, archiving conversations in ArangoDB requires an embedding model to process message turns for search. It also relies on optional external categorization services to apply category labels and track metadata for analysis.

You can apply quality checks to archived episodic conversations using optional edge verification for critical episodes. This supports auditing and recall within memory-assisted workflows while maintaining data accuracy.

To start archiving a transcript JSON for auditability, provide the transcript data to the archive script and run it to store the conversations in ArangoDB. This preserves full context and metadata for later retrieval.

Using per-turn embeddings for conversation archiving enables rapid search and retrieval of specific dialogue points within long-running sessions. It solves context loss and memory decay by storing searchable metadata in ArangoDB.