ffsdb

release map

FFS roadmap from core engine to public proof ladder.

These are sequence labels, not calendar promises. The point is to say what each release should make better in plain English and what evidence has to exist before the claim is credible.

Release Map

sequence labels, not calendar promises

Label Theme What to expect in plain English Proof bar
Now / v0.5 Core exists FFS can open a database file, write data, reopen it, verify storage/qlog posture, run Cypher-shaped queries, search HNSW vectors, and expose ffsd commands. 920 passing tests and current local comparator evals.
Due v0.6 Trust + index hardening Fewer sharp edges on restart, import/export, schema backfills, secondary indexes, and vector-index maintenance. Normal users should feel this as "less scary to operate." Crash/reopen tests, index invariant checks, and safer planner fallbacks.
Due v0.7 Execution speed pass Bigger scans, joins, traversals, sorts, and write-back paths should do less row-at-a-time work. Any "10x" style claim waits for measured workloads; the release goal is benchmark-backed speedups, not a slogan. Morsel/threaded operators and before/after benchmark deltas.
Due v0.8 Movement-aware planner The engine should choose where work happens: scan columns, use a B+-tree, expand graph edges, search vectors first, filter first, stream, materialize, or write back. EXPLAIN output that names movement cost and tests that reject obvious bad plans.
Due v0.9 Write-back loop The signature workflow becomes first-class: retrieve, traverse, derive, write back, update indexes, and commit in one engine instead of stitching sidecars together. A realistic loop demo plus commit/abort/reopen tests.
Due v0.10 Proof ladder Public, reproducible numbers against the well-known systems people already understand, with raw commands, machine specs, and losses included. 10GB, 100GB, and 1TB-class ladders where feasible, plus comparator raw outputs.

what the map serves

Use Cases The app and agent workloads that drive the release order. Engine The current storage, query, vector, and daemon surface. Comparisons The known products used as bars to clear. Evals Current measured paths and caveated numbers. Home The short positioning page.