v0.7.0

Changelog: https://github.com/yisyang/open-wish/blob/v0.7.0/CHANGELOG.md
Full Changelog: v0.6.1...v0.7.0

v0.6.1

Changelog: https://github.com/yisyang/open-wish/blob/v0.6.1/CHANGELOG.md
Full Changelog: v0.6.0...v0.6.1

v0.6.0

Changelog: https://github.com/yisyang/open-wish/blob/v0.6.0/CHANGELOG.md
Full Changelog: v0.5.0...v0.6.0

OpenWish v0.6.0

Better architectural guidance, more focused verification, and clearer management of durable intent.

OpenWish v0.6.0 builds on the Needs → Changes → Tasks workflow introduced in v0.5.0. This release improves how agents establish and evolve architecture, reuse needs, choose meaningful tests, and dispose of completed work.

Humans ask. AI identifies needs, aligns design and engineering, implements, and reviews. Humans review the results.

Architecture grounded in the product

• Shared architecture rubric: assess discoverability, rule ownership, dependency and data ownership, isolated effects, independent outcome checks, and justified complexity.
• Greenfield guidance: begin with a minimal, representative end-to-end slice and evolve affected boundaries as new asks arrive.
• Brownfield onboarding: inspect real behavior through code, tests, and operational context. Recommend retaining satisfactory architecture or propose targeted refactoring with incremental steps, baseline checks, risks, and verification or rollback.
• Explicit refactoring scope: seek human approval for additional restructuring before product edits.

Better need sizing and relationships

• Reuse existing needs when a new format, command, or mechanism serves the same outcome.
• Split independently valuable outcomes and consolidate duplicates while preserving acceptance, constraints, and references.
• Add optional, reasoned relationships directly to needs:
  • requires: the source outcome depends on the target outcome.
  • constrains: the source limits acceptable solutions for the target.
• Inspect recorded relationships with the new derived graph command:

ow graph --json
ow graph N-001 --json

Relationships remain sparse semantic facts. They survive change disposal, permit meaningful cycles, and require no separate graph registry. All current needs still constrain the work, including those without recorded links.

Strategic verification

Updated skills favor the smallest set of fast, independent checks that protects affected needs and consequential failure modes.

• Reuse useful existing checks before adding tests.
• Test stable outcomes, important decisions, and consequential boundary contracts.
• Avoid brittle assertions about incidental wording, layout, private call order, and speculative edge-case matrices.
• Preserve narrow exact assertions for actual contracts.
• Honor explicitly requested coverage targets without treating percentages or test counts as proof of correctness.
• Retain useful regression tests while allowing temporary debugging probes to disappear.

The greenfield evaluation now assesses independently rerunnable current tests, meaningful protection, and useful retained artifacts after disposal. It no longer requires historical first-slice execution evidence.

Clearer retention and context management

• ow status exposes attention groups for:
  • cleanup_candidates
  • awaiting_acceptance
  • blocked_changes
  • retired_needs
• Context warnings appear at 80% of the configured current-needs token budget or 400 context lines.
• Expanded guidance explains disposal, deferred cleanup, retired needs, and preservation of useful manual procedures and rationale.
• Historical evidence belongs in product data when a genuine need requires it; OW requires no permanent execution archive.

Performance and correctness

• Status, summary, and batch closure reuse a source fingerprint within each command, reducing repeated repository scans.
• Status builds reverse need-to-change links once.
• Each subsequent command reads a fresh snapshot.
• Misplaced or nested need artifacts now produce validation errors instead of silently disappearing from current intent.
• Executable-permission changes now invalidate pending result reviews.

Documentation and release tooling

• Added docs/workflows.md, explaining greenfield onboarding, brownfield onboarding, and the change lifecycle with vertical Mermaid diagrams.
• Expanded repository-grounded workflow evaluations for architecture, intent consolidation, and strategic verification.
• Version-tag pushes now trigger TestPyPI publication; stable GitHub releases remain the production PyPI trigger.

Compatibility and upgrading

The package and bundled integrations advance to 0.6.0. Artifact schema remains 3, and need relationships are optional. Valid v0.5.0 projects do not require a schema conversion.

uv tool upgrade openwish
cd /path/to/your/product
ow init

Reinitialization updates older bundled integrations while preserving project-owned needs, work, configuration, and local policy.

Refresh pending verification and AI reviews when reported stale after upgrading: repository fingerprints now include executable permissions. New relationship metadata requires v0.6.0 tooling.

OpenWish continues to require Python 3.11+ and has no third-party runtime dependencies.

v0.5.0

Changelog: https://github.com/yisyang/open-wish/blob/v0.5.0/CHANGELOG.md
Full Changelog: v0.4.0...v0.5.0

OpenWish v0.5.0 — Durable intent, disposable execution

v0.5.0 is NOT backward compatible with earlier versions. This release changes the artifact format and workflow commands. Existing workspaces and scripts cannot be used as-is. No migrations or compatibility shims are provided.

A simpler architecture

OpenWish now follows:

Needs → Changes → Tasks

Humans ask. AI identifies needs, aligns design and engineering, implements, and reviews. Humans review the results.

• Needs preserve durable intent: actors, outcomes, observable acceptance, and genuine constraints.
• Changes hold temporary working context: purpose, design approach, verification, and execution.
• Tasks organize work within a change: lightweight lists with optional local dependencies.
• After completion and review, changes, tasks, and review records can be disposed. Lasting behavior and verification live in code and tests; useful rationale lives near the implementation.

One change can serve several needs, and one need can be served by several changes. Current needs remain binding after delivery.

New workflow commands

• ow new change … --need N-001 — create bounded work linked directly to needs.
• ow record-review C-001 … — record outcomes and verification evidence.
• ow close C-001 --by human — close work after human acceptance.
• ow clean C-001 — dispose closed execution records while preserving needs, product files, and unknown content.

Explicit fast-forward acceptance is scoped to an individual change. Pending reviews become stale when relevant intent, plans, evidence, or repository content changes.

Removed

• Persistent specs and epics.
• Need-owned execution state and review sidecars.
• Stage gates and the authorize, approve, and reopen commands.
• Global task group/order requirements.
• Legacy template recognition and compatibility handling.

All eight bundled skills and supporting documentation now follow the new model.

Compatibility

Only artifact schema 3 is supported. ow init does not convert older workspaces.

Before v1.0.0, OpenWish prioritizes a simple current design over backward compatibility.

v0.4.0

Changelog: https://github.com/yisyang/open-wish/blob/v0.4.0/CHANGELOG.md
Full Changelog: v0.3.1...v0.4.0

v0.3.1

Initial release for OpenWish NDD framework!

Full Changelog: https://github.com/yisyang/open-wish/commits/v0.3.1