[Mnemosyne] Foundation: holographic archive index and ingestion pipeline #1242

Closed
opened 2026-04-12 00:13:53 +00:00 by Rockachopa · 2 comments
Owner

Project Mnemosyne — The Living Holographic Archive

Goal

Bootstrap the Mnemosyne module: a sovereign, on-chain anchored memory system that ingests documents, conversations, and artifacts into a searchable holographic index.

Scope

  1. Archive Ingestion (mnemosyne/ingest.py) — Accept text/JSON/markdown inputs, chunk them with overlap, compute embeddings stub (pluggable backend), store in local SQLite + FTS5
  2. Holographic Index (mnemosyne/index.py) — Query interface: keyword search (FTS5) + semantic search (embedding similarity), merge results with reciprocal rank fusion
  3. CLI Entry Point (mnemosyne/cli.py) — mnemosyne ingest <path> and mnemosyne query <text> commands
  4. Tests — Basic round-trip test: ingest sample docs, query, verify recall

Design Principles

  • No network calls at ingest time (embeddings are optional, compute locally or skip)
  • SQLite + FTS5 only — no external vector DB dependency
  • Pluggable embedding backend (sentence-transformers, Ollama, or none)
  • Compact: the whole module < 500 lines of Python

Acceptance Criteria

  • mnemosyne/ingest.py chunks and stores documents in SQLite
  • mnemosyne/index.py returns ranked results for keyword queries
  • mnemosyne/cli.py exposes ingest/query subcommands
  • Tests pass: ingest sample → query → verify results
  • README with usage examples
## Project Mnemosyne — The Living Holographic Archive ### Goal Bootstrap the Mnemosyne module: a sovereign, on-chain anchored memory system that ingests documents, conversations, and artifacts into a searchable holographic index. ### Scope 1. **Archive Ingestion** (`mnemosyne/ingest.py`) — Accept text/JSON/markdown inputs, chunk them with overlap, compute embeddings stub (pluggable backend), store in local SQLite + FTS5 2. **Holographic Index** (`mnemosyne/index.py`) — Query interface: keyword search (FTS5) + semantic search (embedding similarity), merge results with reciprocal rank fusion 3. **CLI Entry Point** (`mnemosyne/cli.py`) — `mnemosyne ingest <path>` and `mnemosyne query <text>` commands 4. **Tests** — Basic round-trip test: ingest sample docs, query, verify recall ### Design Principles - No network calls at ingest time (embeddings are optional, compute locally or skip) - SQLite + FTS5 only — no external vector DB dependency - Pluggable embedding backend (sentence-transformers, Ollama, or none) - Compact: the whole module < 500 lines of Python ### Acceptance Criteria - [ ] `mnemosyne/ingest.py` chunks and stores documents in SQLite - [ ] `mnemosyne/index.py` returns ranked results for keyword queries - [ ] `mnemosyne/cli.py` exposes ingest/query subcommands - [ ] Tests pass: ingest sample → query → verify results - [ ] README with usage examples
Timmy was assigned by Rockachopa 2026-04-12 00:13:53 +00:00
Author
Owner

Claiming this issue. Working on Mnemosyne foundation — ingestion, holographic index, and CLI.

Claiming this issue. Working on Mnemosyne foundation — ingestion, holographic index, and CLI.
Member

Closing — the Mnemosyne foundation is fully shipped on main at nexus/mnemosyne/:

  • archive.py — MnemosyneArchive with CRUD, search, graph analysis
  • entry.py — ArchiveEntry dataclass
  • ingest.py — Document ingestion pipeline
  • linker.py — HolographicLinker (Jaccard similarity)
  • cli.py — Full CLI with 15+ commands
  • tests/ — test_archive.py, test_graph_clusters.py

Originally merged via #1217 (Phase 1), then expanded by #1225, #1220, #1234, #1238, #1241, #1246. PR #1243 (root-level duplicate path) was closed as stale. See RCA #1256.

Closing — the Mnemosyne foundation is fully shipped on main at `nexus/mnemosyne/`: - `archive.py` — MnemosyneArchive with CRUD, search, graph analysis - `entry.py` — ArchiveEntry dataclass - `ingest.py` — Document ingestion pipeline - `linker.py` — HolographicLinker (Jaccard similarity) - `cli.py` — Full CLI with 15+ commands - `tests/` — test_archive.py, test_graph_clusters.py Originally merged via #1217 (Phase 1), then expanded by #1225, #1220, #1234, #1238, #1241, #1246. PR #1243 (root-level duplicate path) was closed as stale. See RCA #1256.
Sign in to join this conversation.
2 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Timmy_Foundation/the-nexus#1242