Design adr 0001 pluggable storage#40
Conversation
Pluggable storage backend, network access, and emergent domain classification. Introduces MemoryStore + Embedder traits, PgMemoryStore alongside SqliteMemoryStore, HTTP MCP + API key auth, and HDBSCAN-based domain clustering. Phase 5 federation deferred to a follow-up ADR. - docs/adr/0001-pluggable-storage-and-network-access.md -- Accepted - docs/plans/0001-phase-1-storage-trait-extraction.md - docs/plans/0002-phase-2-postgres-backend.md - docs/plans/0003-phase-3-network-access.md - docs/plans/0004-phase-4-emergent-domain-classification.md - docs/prd/001-getting-centralized-vestige.md -- source RFC
added amends to the postgres backend/phase2
|
Thank you for your contribution! Many people have requested this feature, so I will get working on db portablility ASAP. |
|
Jan, yes — please go ahead and open the Phase 1 implementation PR for the storage/embedder trait extraction when you are ready.\n\nI reviewed this ADR against the v2.1.1 portable-sync work. They line up cleanly: v2.1.1 solves the immediate cross-machine file-sync path, while this ADR remains the right roadmap for full pluggable primary storage. The shape I want next is exactly the first phase you mentioned: keep SQLite behavior stable, introduce the trait boundary, keep the concrete SQLite alias/source compatibility where possible, and let the full suite prove no behavior regression.\n\nFor PostgreSQL: I agree with the Phase 2 direction here. Postgres/pgvector should come after the trait boundary lands, behind an explicit feature/config path, not as a forced dependency for local users. Open the Phase 1 PR and I’ll review it as the base for the Postgres follow-up. |
2 participants
Quite a lot of text, but I hate rabbit-holing with AI...
Most of the time when it all fits, I can do hours long runs and have working deliverables.
Vestige is really really swell, together with some skills to keep it within predefined lines, it's a godsend.
But I have home, work desktops, a portable ... sometimes exporting and importing , but then I lose too much for comfort.
Not included in the thought process (yet) would be a user-level hierarchy, but I haven't gathered my thoughts on that
Anywayz, it's just for your review if you'd be so inclined to read it, otherwise you can just drop it, no grudge will be held ;-)
I already have phase 1 (traiting the db/embedder) that passes all tests, if you want, I'll already create a pr for it too
Jan