DocForge2/ACTIVE_SLICE.md

Milestone state

Last completed milestone: 4 — adapter SDK and product documentation.
Baseline: 6d06195950d33bcd2d712f8819bbfb3d6652ad03 on main, dev, origin/main, and origin/dev.
Active milestone: 5 — stabilization and first DocForge2 release.
Outcome: Compatibility, determinism, recovery, security, performance, and representative task advantage are proven for the first successor release.
Status: Active.
Release candidate: 1.4.0, subject to every gate below.

Authority

This contract activates Milestone 5 from the clean, merged, and pushed Milestone 4 closeout. The roadmap in /home/andraxion/.openclaw/workspace/DocForgeOutline.md supplies direction; this file freezes the executable scope and acceptance criteria.

main remains the Milestone 4 baseline until the complete Milestone 5 release candidate passes. Implementation occurs on dev.

Required release evidence

The release candidate must prove all of the following from maintained, reproducible gates:

1. Legacy one-method adapter compatibility and the frozen package, CLI, MCP, schema, rendering, descriptor, result-envelope, and no-AST surfaces.
2. Exact complete/incremental primary-graph and Logic equivalence for maintained incremental adapters.
3. Deterministic adapter, worker, renderer, command-reference, configuration, and publication output.
4. Detection of concurrent source mutation without publishing or serving a mixed or stale generation.
5. Atomic, crash-safe derived publication and exact recovery from interrupted publication.
6. Corrupt extraction-cache, index, attestation, receipt, fragment, and projection recovery using canonical sources as the authority.
7. No stale reads after source or policy change, including generation-pinned retrieval and viewer behavior.
8. Exact-hash canonical application with project-owned serialization, stale-proposal rejection, and post-application resynchronization.
9. Closed policy precedence across process capability, descriptor policy, --no-ast, projection policy, worker enforcement, and viewer enforcement.
10. Manual and portable-graph isolation, immutable package verification, bounded detached workers, rendering-policy enforcement, viewer-policy enforcement, and accessibility.
11. Comparative representative task evidence with fixed questions, answer keys, provenance, latency, and response-size accounting for both graph-assisted and source-only workflows.
12. A fresh-wheel and fresh-clone release rehearsal, full quality and browser gates, maintained benchmarks, secret scans, exact version identity, signed-off release notes, and reproducible release artifacts.

Deliverables

• A maintained Milestone 5 compatibility matrix and aggregate release gate.
• Reproducible migration, recovery, concurrency, and comparative-task evidence.
• One authoritative package version shared by package metadata, Python, CLI, MCP, viewer manager, reference MCP, generated configuration, and release documentation.
• A standard project license file, changelog or release notes, release baseline, closeout record, and machine-readable evidence.
• A clean main merge, synchronized main and dev, annotated v1.4.0 tag, and Forgejo release only after the final documentation-bearing commit passes a fresh-clone gate.

Fixed boundaries

• Preserve the docforge package, docforge CLI, docforge-mcp, MCP tool names, schema version 1 surfaces, effective policy version 1, projection policy version 2, and legacy adapter entry point.
• Preserve canonical project sources. Migration may rebuild disposable state but may not rewrite canonical content merely to satisfy the release.
• Full rebuild remains the recovery and equivalence oracle.
• Derived artifacts must fail closed on malformed, foreign, stale, oversized, or incompatible state.
• Version 1.4.0 is additive relative to 1.0.0; breaking a frozen contract requires a separately justified major-version decision.

Exclusions

• No WorldForge change or benchmark.
• No ScrapeStation change, production binding change, or production migration.
• No legacy-repository mutation.
• No production MCP repointing.
• No remote render service, render farm, third-party renderer ecosystem, graph federation, or dedicated graph database.
• No arbitrary adapter command execution, compiler execution, remote execution, or expanded launcher authority.
• No PyPI publication unless it is separately verified as an intended existing release channel.

Release sequence

1. Freeze and implement the compatibility, migration, recovery, concurrency, and task-evidence gates.
2. Stabilize version identity, packaging, license, security, and release automation.
3. Freeze one clean executable candidate and run the full repository, browser, benchmark, fresh-wheel, fresh-clone, and secret-scan gates.
4. Close documentation atomically against that candidate and rerun documentation-only validation.
5. Merge and push the final candidate.
6. Rehearse from a fresh anonymous clone at the exact commit.
7. Create and push the annotated v1.4.0 tag and publish the Forgejo release from the verified artifacts.

The completed Milestone 4 contract and exclusions remain preserved in SLICE_HISTORY.md.

Release-candidate evidence — 2026-07-29

The executable implementation is frozen at d2bb95fe6190e659cf66ba57c78be53b63b53240. The proof-bearing candidate base is 2b98059b44f4d46b4d4cce776f163e893c647c76; it includes exact legacy-tag verification for the fresh-clone gate and four maintained aggregate derived-recovery tests.

The clean executable release gate passed formatting, Python and web lint, strict Pyright, compilation, lock and dependency checks, builds, generated documentation, three accessibility flows, fresh-wheel adoption, artifact reproducibility, secret scans, and the full Milestone 0 through Milestone 4 benchmark sequence. Its exact test evidence was:

• 142 contract tests plus 272 subtests.
• 371 complete tests plus 419 subtests.
• 116 compatibility tests plus 263 subtests.
• 29 concurrency tests plus 2 subtests.
• 68 recovery tests plus 62 subtests.

The real migration gate preserves exact canonical and proposal bytes from the annotated v1.0.0 lineage while rebuilding the disposable index from schema 1 to schema 3. The real-package task gate uses the lock-pinned markdown-it-py 4.2.0 tree and proves exact graph-assisted and source-only answers for all reviewed tasks.

The later proof-only recovery commit adds four maintained tests without changing executable product code. The current recovery aggregate passes 72 tests plus 62 subtests.

Milestone 5 remains active. Documentation closeout, final branch synchronization, a fresh anonymous clone at the exact documentation-bearing remote commit, final artifact checksums, the annotated v1.4.0 tag, and the Forgejo release are still pending. No PyPI publication is planned.