feat: merge module map, bounded agent context, durable evidence, and review UX into main

[orenlab]

Summary

This PR merges the current feat/module-map development line into main.

The branch consolidates several completed product tracks:

• canonical Module Map projection and HTML visualization;
• Review Hub and Overview launchpad;
• shared report UI primitives and visual-system cleanup;
• analysis-phase and context-pressure observability;
• bounded MCP response governance;
• durable receipt, Patch Trail, and blast-artifact retrieval;
• continuation paging for Engineering Memory and implementation context;
• Engineering Memory sourcing and hydration performance improvements;
• runtime reachability extensions;
• updated plugin skills, help topics, README, and dependency pins.

The goal of this merge is to establish the current CodeClone 2.1 alpha foundation in main before the final 2.1.0a1 stabilization cycle.

Product changes

Module Map

• Added a canonical derived module_map report projection.
• Exposed Module Map through get_report_section.
• Added the Module Map HTML tab.
• Extracted shared deterministic dependency-graph sampling and layout.
• Unified Module Map and dependency graph rendering on shared graph primitives.

Review experience

• Added a derived review queue sourced from findings across all families.
• Added the Review Hub with a prioritized finding queue.
• Added the Overview launchpad.
• Introduced shared finding cards and reused them across structural and suggestion views.
• Moved overloaded-module guidance into the Module Map domain.
• Refined filters, badges, meters, focus states, contrast, and shared design tokens.

Bounded agent context

• Added passive context-governance envelopes.
• Added response governance to controlled-change start and finish flows.
• Added implementation-context facet paging.
• Added Engineering Memory continuation pages.
• Enforced response budgets for start, finish, implementation context, and compact memory retrieval.
• Added drill-down reachability metadata and continuation capabilities.
• Added context-governance contract tests and synchronized help guidance.

Durable evidence

• Added durable post-clear review receipt retrieval.
• Added durable Patch Trail retrieval.
• Added durable blast-artifact retrieval.
• Exposed receipt identity fields.
• Deduplicated finish receipts.
• Added idempotent start replay and stabilized replay identity against advisory drift.

Engineering Memory

• Added incremental semantic sourcing for unchanged rows.
• Prevented semantic-lane deletion when a source read fails.
• Removed duplicate semantic sourcing during rebuilds.
• Added delta experience replacement and batched store hydration.
• Added JetBrains as an IDE governance client.

Observability

• Added analysis-phase observability.
• Split the platform cockpit into focused tabs.
• Added deterministic phase and span visualizations.
• Aligned context estimates and renamed the display unit to context units.
• Improved SQL statement counting and reduced visual noise.

Core and packaging

• Extended runtime reachability for FastAPI, Pydantic, and Strawberry.
• Consolidated duplicate coercion helpers.
• Refreshed project dependencies and lockfile.
• Pinned analytics dependencies for Python 3.10+ compatibility.
• Updated Zensical.

Documentation

• Refreshed README positioning.
• Updated MCP context-governance guidance and help topics.
• Resynchronized plugin skill copies with their canonical sources.
• Updated architecture-triage, Engineering Memory, blast-radius, receipt, and start-replay documentation.
• Refreshed the changelog.

Validation

• Canonical analysis remains stable.
• Focused structural checks and gates remain deterministic.
• Implementation-context paging is bounded and reproducible.
• Durable evidence routes are available for receipts, Patch Trails, and blast artifacts.
• Context-governance truth-table tests are included.
• The latest contract test slice passes.

Known stabilization work

This PR intentionally merges the completed feature branch before the final alpha stabilization fixes.

The following issues will be addressed in a separate branch:

• section-specific bounding for get_report_section;
• compact continuation cursor size;
• zero-lane omitted-evidence overhead;
• claim-validation negation handling;
• trajectory_get compact/full behavior;
• Patch Trail digest lookup consistency;
• receipt outcomes when intent or structural evidence is unavailable;
• smaller MCP schema and documentation inconsistencies;
• final documentation deduplication and release synchronization.

These issues do not change the canonical analysis truth, but they must be resolved before tagging 2.1.0a1.

Follow-up

After merge:

1. create a dedicated 2.1.0a1 stabilization branch;
2. close the known boundedness and evidence-integrity defects;
3. complete the documentation audit;
4. run full suite, wheel, fresh-install, migration, and real-agent smoke tests;
5. tag and publish 2.1.0a1.

[github-actions]

CodeClone Review

✅ Passed · Health 91/100 (A) · Baseline ok · Cache miss · CodeClone 2.1.0a1

Review snapshot

Area	Signal	Review note
Clones	0 total, 0 new, 0 known	no new clone debt reported
Quality	CC max 20, CBO max 9, LCOM4 max 3, overloaded 42	structural metric snapshot
Dependencies	avg 6.9, p95 21, max 23, cycles 0	acyclic
Coverage Join	not joined	no coverage.xml facts in this report
Security Surfaces	250 surfaces, 6 categories, 125 production	report-only boundary inventory
API Surface	7110 symbols, 631 modules	0 breaking, 0 added
Dead code	0 high-confidence, 2 suppressed	clean

Review focus

• Treat 125 production security surface(s) as review-first boundary code when touched.
• Review 42 overloaded module candidate(s) when they intersect this PR.

_{Security Surfaces are report-only capability inventory, not vulnerability claims. Generated by CodeClone}