Retrieval & memory
Making a large library searchable, one evidence layer at a time
A substantial keyword index is useful before a vector phase is finished. The Wisdom Library records make that distinction explicit.
- Gutenberg books in the recorded FTS index
- 13,468
- Indexed chunks
- 1,883,249
- Embedded chunks at that checkpoint
- 0
The completed layer
The saved preflight records a completed keyword/full-text-search index with 13,468 Gutenberg books and 1,883,249 chunks. A separate unified quality check passed database integrity, FTS parity, and retrieval probes. These figures describe that dated source-side index.
A whole-archive audit and a later local-interface check are separate records. The later successful interface probe did not retroactively change the earlier full-audit failure state.
Preparing the R640 handoff
The Drive manifest records 53 expected staging files, including 48 split Gutenberg archive parts, a separate candidate index archive, support files, and checksums. Local archive-integrity tests passed before upload; the Drive listing matched expected file names and sizes.
Remote content hashes were not supplied by the connector. Download-side integrity and the external source-drive copy still required destination verification.
What was not complete
Full-corpus BGE-M3 embeddings were pending at that checkpoint: zero of the 1,883,249 chunks were embedded, and the local Qdrant directory for this corpus was empty. This is a keyword-retrieval milestone, not a completed vector search system.
This corpus is separate from the small Riku pilot’s functioning vector index. Success in one retrieval project does not complete the other.
What we are trying to achieve
We want a library that makes accumulated reading and research useful in everyday work. The path is layered: preserve sources, make them searchable, validate retrieval, then add further retrieval methods with their own checks. An answer should remain connected to what the source actually says.
Sources & reading notes
- R640 transfer preflightLatest verified state and transfer limitations. Drive record · access may be restricted Source record retained privately.
- Wisdom Library Drive staging manifest53-file staging inventory and destination-verification boundary. Drive record · access may be restricted Source record retained privately.
This is a dated, source-based research account, not a live machine report. Historical results, engineering preparation, research designs, open review gates, and measured pilots remain distinct.