2.2.0 resolves from Maven Central, eleven artifacts and thirteen modules are published, every planned tier through Tier 9 is shipped, and over the last fortnight this repository had sixteen unique visitors and three referrals from outside GitHub, which is the number that decides whether any of the rest of it matters.

Done: Tiers 0 through 9. 2.1.0 put the REST engine on every target the core compiles for, added scoped tokens, built the GraalVM image the README had been claiming and made collection migrations resumable. 2.2.0 closed Tiers 8 and 9 on 6 August with payload index parameters, sparse and hybrid search proven against a real server rather than against a body assertion, server-side inference, an ingest that owns batching and resume, TLS trust a native consumer can configure, a vocabulary for a degraded cluster, and kdrant-cli. It took three release attempts, all three failing in the same job, which is its own item now.

Scheduled: four tiers and four releases, none started. Tier 10 answers Qdrant 1.19 and the agent-era query surface and closes as 2.3.0 on 26 October. Tier 11 is operability at the size people run and closes as 2.4.0 on 4 December. Tier 12 is the distance between what this repository claims and what its build checks, plus Android, and closes as 2.5.0 on 19 February 2027. Tier 13 is what a caller cannot check today, payload keys the compiler checks, an in-memory transport the shared contract keeps honest, recall measured against exact search, and it closes as 2.6.0 on 16 April 2027.

Not scheduled: anything after 16 April, and anything on the other boards after December. Both are the roadmap week of 14 December, which also decides whether 2.6.0 is really 3.0.

The status is At risk and the risk is not the code. Every tier on this board has been argued from the artifact rather than from a user, because there have been no users to argue from, and the one item that tries to change that is a newsletter submission and a repository description. What would change the plan: M73 producing referrers, which reorders everything behind whatever an actual user hit first; a Metis pilot signing, which takes the main track for about four weeks and moves the 2027 dates right; and Qdrant's 2027 roadmap, since a client's plan is downstream of what the server ships.

Sunbelt Computer Software

PL/B Language Development and Support