{"row":{"id":"ADR-002","title":"Forever v0.x -- no v1.0 target","file":"adr-002-forever-v0x.md","status":"accepted","date":"2026-04-13","cites":6,"supersedes":[],"superseded_by":[],"layers":[],"tags":["versioning","release-model","semver"],"room":true},"pin":{"release_tag":"v0.107.0","commit":"5d61b7e13f1f6a10c17414f61146648f76bcf2bd"}}

[{"k":"h","depth":1,"text":"ADR-002: Forever v0.x — no v1.0 target","id":"adr-002-forever-v0x--no-v10-target"},{"k":"quote","inline":[{"k":"strong","text":"⚠️ AMENDED 2026-06-20 (D-2026-06-20-N1) — this doctrine is RETIRED."},{"k":"text","text":"\nThe \"forever-v0.x · no v1.0 target\" model described below is "},{"k":"strong","text":"superseded"},{"k":"text","text":".\nNika now follows "},{"k":"strong","text":"real semver toward a 1.0.0 public launch"},{"k":"text","text":" (the engine\nis at 0.90.0 RC-grade; 1.0.0 ships when the 7 shadow zones are green; the\nremaining crates land additively across the 1.x minors). The "},{"k":"code","text":"nika: v1"},{"k":"text","text":"\nLANGUAGE envelope is frozen forever and is unaffected (a different axis).\nThe original text is preserved below per "},{"k":"code","text":"cross-source-validation.md"},{"k":"text","text":" §2.7."}]},{"k":"quote","inline":[{"k":"strong","text":"⚠️ AMENDED 2026-06-20 (D-2026-06-20-N1) · \"forever-v0.x · no v1.0 ever\" is RETIRED → real semver toward a 1.0 launch."},{"k":"text","text":"\nThe decision below is SUPERSEDED at the version-policy layer. Canon courant ·"},{"k":"text","text":"The "},{"k":"code","text":"v0.8X.Y"},{"k":"text","text":" layer-phase tag scheme and the \"42 crates @ v0.90 milestone /\nno v1.0\" framing are SUPERSEDED · 42 crates is now a post-1.0 target reached\nacross 1.x minors. The original body is preserved below per\ncross-source-validation §2.7 (zero silent rewrite · this ADR's filename + title\nstay frozen as historical anchors)."}]},{"k":"h","depth":2,"text":"Context","id":"context"},{"k":"p","inline":[{"k":"text","text":"Nika is developed primarily by a solo author augmented by AI. Committing to a v1.0 target introduces two failure modes: (a) rushing breaking changes into a 1.0 window to \"lock the API,\" producing a worse API than iteration would; (b) refusing to change load-bearing APIs post-1.0 because SemVer punishes breaking changes on major releases. Neither outcome serves correctness."}]},{"k":"p","inline":[{"k":"text","text":"Context: legacy Nika had 17 releases in v0.x space (v0.1 → v0.79.3) across ~6 months of solo+AI work. Phase D (Diamond) is projected at ~10–11 months to v0.90 milestone. Applying strict SemVer 1.0 semantics over that horizon forces compat decisions that conflict with Diamond's \"zero dead code\" philosophy."}]},{"k":"h","depth":2,"text":"Decision","id":"decision"},{"k":"p","inline":[{"k":"strong","text":"Nika remains at v0.x indefinitely."},{"k":"text","text":" No v1.0 target. Ever."}]},{"k":"list","ordered":false,"items":[[{"k":"text","text":"Phase milestones (v0.90, v0.95, v0.100) are "},{"k":"strong","text":"markers"},{"k":"text","text":", not contracts. They signal accumulated crate admissions (40–42 @ v0.90, ~63 @ v0.95, ~75 @ v0.100, cap 100)."}],[{"k":"text","text":"Breaking changes ship on MINOR bumps (v0.90 → v0.91) without ceremony. "},{"k":"code","text":"#[non_exhaustive]"},{"k":"text","text":" + "},{"k":"code","text":"pub fn new()"},{"k":"text","text":" (ADR-007) preserve forward-compat "},{"k":"em","text":"by construction"},{"k":"text","text":" within minor ranges for downstream struct-literal use."}],[{"k":"text","text":"The CHANGELOG distinguishes "},{"k":"code","text":"Breaking"},{"k":"text","text":", "},{"k":"code","text":"Added"},{"k":"text","text":", "},{"k":"code","text":"Changed"},{"k":"text","text":", "},{"k":"code","text":"Fixed"},{"k":"text","text":", "},{"k":"code","text":"Deprecated"},{"k":"text","text":" sections. Breaking changes are explicit but not gated by a SemVer major."}],[{"k":"text","text":"Public documentation ("},{"k":"code","text":"README.md"},{"k":"text","text":", "},{"k":"code","text":"MANIFESTO.md"},{"k":"text","text":") states this explicitly so users adopt with correct expectations."}]]},{"k":"h","depth":2,"text":"Consequences","id":"consequences"},{"k":"h","depth":3,"text":"Positive","id":"positive"},{"k":"list","ordered":false,"items":[[{"k":"text","text":"Freedom to refactor load-bearing APIs when evidence demands (post-audit findings, shadow-zone resolutions)."}],[{"k":"text","text":"No artificial \"pre-1.0\" vs \"post-1.0\" discipline gap. Every release held to the same quality floor."}],[{"k":"text","text":"Quality-first signaling: a v0.95 release with 95% mutation coverage is tangibly better than a v1.0 with compat shims."}]]},{"k":"h","depth":3,"text":"Negative","id":"negative"},{"k":"list","ordered":false,"items":[[{"k":"text","text":"Some ecosystem tools (package registries, dependabot rules) treat pre-1.0 crates as \"unstable\" and apply stricter update policies. Acceptable — Nika distributes as a CLI + library, not a churned dependency."}],[{"k":"text","text":"Requires explicit user education: "},{"k":"code","text":"v0.x"},{"k":"text","text":" here ≠ \"pre-alpha quality.\" README must state this."}]]},{"k":"h","depth":3,"text":"Neutral","id":"neutral"},{"k":"list","ordered":false,"items":[[{"k":"text","text":"AGPL license already signals long-term project commitment independent of version number."}]]},{"k":"h","depth":2,"text":"Evidence","id":"evidence"},{"k":"list","ordered":false,"items":[[{"k":"code","text":"ROADMAP.md"},{"k":"text","text":" — \"forever v0.x\" in header, phased crate counts"}],[{"k":"code","text":"Cargo.toml"},{"k":"text","text":" line 19 — "},{"k":"code","text":"version = \"0.80.0\""},{"k":"text","text":" at workspace root, inherited"}],[{"k":"code","text":"CHANGELOG.md"},{"k":"text","text":" header — explicit \"Nika follows forever-v0.x\" statement"}],[{"k":"text","text":"memory: "},{"k":"code","text":"feedback_no_deadlines.md"},{"k":"text","text":" — \"never pressure with timelines\""}],[{"k":"text","text":"memory: "},{"k":"code","text":"POST_AUDIT_REVISIONS.md"},{"k":"text","text":" — \"Version model: Forever v0.x\""}]]},{"k":"h","depth":2,"text":"Alternatives considered","id":"alternatives-considered"},{"k":"h","depth":3,"text":"Alt A — Target v1.0 at milestone v0.90","id":"alt-a--target-v10-at-milestone-v090"},{"k":"p","inline":[{"k":"text","text":"Ship v1.0.0 once the 40–42 admitted crates hit the quality bar. Rejected: v0.90 is an arbitrary crate count, not an API stabilization moment. Memory features (v0.95) and agent-v2 (v0.100) are forward-known to require kernel API evolution; committing v1.0 before those land is dishonest."}]},{"k":"h","depth":3,"text":"Alt B — Calendar versioning (YYYY.MM)","id":"alt-b--calendar-versioning-yyyymm"},{"k":"p","inline":[{"k":"code","text":"2026.04"},{"k":"text","text":", "},{"k":"code","text":"2026.10"},{"k":"text","text":". Rejected: loses the \"quality ratchets up with number\" signal. Users infer from v0.80 → v0.90 that substantial progress happened."}]},{"k":"h","depth":2,"text":"Related","id":"related"},{"k":"list","ordered":false,"items":[[{"k":"text","text":"ADR-001 — orphan rewrite (why we're starting the v0.8x line)"}],[{"k":"text","text":"ADR-007 — forward-compat invariants (how we preserve struct-literal compat within minor ranges)"}]]},{"k":"h","depth":2,"text":"Notes","id":"notes"},{"k":"p","inline":[{"k":"text","text":"If the ecosystem changes such that \"forever v0.x\" actively harms adoption (e.g., major packaging tool blocks < 1.0 by policy), revisit. Not expected within 24 months."}]}]

ADR-002

Forever v0.x -- no v1.0 target

accepted · 2026-04-13 · cites 6

ADR-002: Forever v0.x — no v1.0 target #

⚠️ AMENDED 2026-06-20 (D-2026-06-20-N1) — this doctrine is RETIRED. The "forever-v0.x · no v1.0 target" model described below is superseded. Nika now follows real semver toward a 1.0.0 public launch (the engine is at 0.90.0 RC-grade; 1.0.0 ships when the 7 shadow zones are green; the remaining crates land additively across the 1.x minors). The nika: v1 LANGUAGE envelope is frozen forever and is unaffected (a different axis). The original text is preserved below per cross-source-validation.md §2.7.

⚠️ AMENDED 2026-06-20 (D-2026-06-20-N1) · "forever-v0.x · no v1.0 ever" is RETIRED → real semver toward a 1.0 launch. The decision below is SUPERSEDED at the version-policy layer. Canon courant ·The v0.8X.Y layer-phase tag scheme and the "42 crates @ v0.90 milestone / no v1.0" framing are SUPERSEDED · 42 crates is now a post-1.0 target reached across 1.x minors. The original body is preserved below per cross-source-validation §2.7 (zero silent rewrite · this ADR's filename + title stay frozen as historical anchors).

Context #

Nika is developed primarily by a solo author augmented by AI. Committing to a v1.0 target introduces two failure modes: (a) rushing breaking changes into a 1.0 window to "lock the API," producing a worse API than iteration would; (b) refusing to change load-bearing APIs post-1.0 because SemVer punishes breaking changes on major releases. Neither outcome serves correctness.

Context: legacy Nika had 17 releases in v0.x space (v0.1 → v0.79.3) across ~6 months of solo+AI work. Phase D (Diamond) is projected at ~10–11 months to v0.90 milestone. Applying strict SemVer 1.0 semantics over that horizon forces compat decisions that conflict with Diamond's "zero dead code" philosophy.

Decision #

Nika remains at v0.x indefinitely. No v1.0 target. Ever.

Phase milestones (v0.90, v0.95, v0.100) are markers, not contracts. They signal accumulated crate admissions (40–42 @ v0.90, ~63 @ v0.95, ~75 @ v0.100, cap 100).
Breaking changes ship on MINOR bumps (v0.90 → v0.91) without ceremony. #[non_exhaustive] + pub fn new() (ADR-007) preserve forward-compat by construction within minor ranges for downstream struct-literal use.
The CHANGELOG distinguishes Breaking, Added, Changed, Fixed, Deprecated sections. Breaking changes are explicit but not gated by a SemVer major.
Public documentation (README.md, MANIFESTO.md) states this explicitly so users adopt with correct expectations.

Consequences #

Positive #

Freedom to refactor load-bearing APIs when evidence demands (post-audit findings, shadow-zone resolutions).
No artificial "pre-1.0" vs "post-1.0" discipline gap. Every release held to the same quality floor.
Quality-first signaling: a v0.95 release with 95% mutation coverage is tangibly better than a v1.0 with compat shims.

Negative #

Some ecosystem tools (package registries, dependabot rules) treat pre-1.0 crates as "unstable" and apply stricter update policies. Acceptable — Nika distributes as a CLI + library, not a churned dependency.
Requires explicit user education: v0.x here ≠ "pre-alpha quality." README must state this.

Neutral #

AGPL license already signals long-term project commitment independent of version number.

Evidence #

ROADMAP.md — "forever v0.x" in header, phased crate counts
Cargo.toml line 19 — version = "0.80.0" at workspace root, inherited
CHANGELOG.md header — explicit "Nika follows forever-v0.x" statement
memory: feedback_no_deadlines.md — "never pressure with timelines"
memory: POST_AUDIT_REVISIONS.md — "Version model: Forever v0.x"

Alternatives considered #

Alt A — Target v1.0 at milestone v0.90 #

Ship v1.0.0 once the 40–42 admitted crates hit the quality bar. Rejected: v0.90 is an arbitrary crate count, not an API stabilization moment. Memory features (v0.95) and agent-v2 (v0.100) are forward-known to require kernel API evolution; committing v1.0 before those land is dishonest.

Alt B — Calendar versioning (YYYY.MM) #

2026.04, 2026.10. Rejected: loses the "quality ratchets up with number" signal. Users infer from v0.80 → v0.90 that substantial progress happened.

ADR-001 — orphan rewrite (why we're starting the v0.8x line)
ADR-007 — forward-compat invariants (how we preserve struct-literal compat within minor ranges)

Notes #

If the ecosystem changes such that "forever v0.x" actively harms adoption (e.g., major packaging tool blocks < 1.0 by policy), revisit. Not expected within 24 months.

read at v0.107.0 · the decision record ships with the engine

Every decision, with its status This one in the repo