CTX Layer to PersistMind bridge 0.2.0a19

Compatibility bridge for existing CTX Layer installations. Adds PersistMind commands, transactional state migration, dual-generation uninstall, and signed legacy update assets. New installations should use the PersistMind release repository.

CTX Layer 0.2.0a18

CTX Layer 0.2.0a18

Released July 12, 2026.

This preview fixes Windows source indexing and recovery issues found during a real update and index run against the MARKETING repository.

What was fixed

• Resolves sqlite-vec's extensionless vec0 load path to the packaged vec0.dll on Windows.
• Accepts sqlite-vec's native v0.1.9 version string as matching Python package version 0.1.9.
• Removes partial source DB/WAL/SHM files when source-segment construction fails, so indexing can be retried safely.
• Extends storage repair to detect and remove catalog-unpublished source segments with storage repair --yes.
• Reports missing published source segments as verified-backup restore actions instead of incorrectly reporting storage as clean.

Field replay

The fix was exercised against C:\codes\MARKETING after generated deployment and dependency trees were removed from Git tracking:

• 1,964 UTF-8 text files indexed;
• 2,889 binary or over-budget files skipped intentionally;
• 10,376 source records built;
• 25,960,024 logical text bytes indexed;
• full split-storage verification passed;
• source search passed;
• repeated ctxlayer --repo . index completed successfully and reused the published generation.

Repository file limit

The 10,000-file discovery guard remains intentional. Repositories that track generated deployments, dependency trees, or build outputs should remove them from Git tracking and add them to .gitignore. This release fixes the Windows runtime and recovery failures that occurred after the repository was brought under that limit.

Install or update

ctxlayer update --channel preview

Direct wheel installation:

python -m pip install --upgrade https://github.com/abhilashsblai/ctxlayer-release/releases/download/v0.2.0a18/ctxlayer-0.2.0a18-py3-none-any.whl

CTX Layer 0.2.0a17

CTX Layer 0.2.0a17

Released July 12, 2026.

This release separates durable project migration from rebuildable source-index
budgets for large repositories.

Correction

When a repository exceeds source indexing limits—such as more than 10,000
eligible files—the updater now:

1. preserves the verified monolith backup;
2. completes authoritative knowledge, learning, task, activity, audit, and
   policy migration;
3. activates and verifies split storage;
4. records source indexing as deferred with remediation;
5. commits the update instead of rolling durable data back.

Index data is derived from repository files and can be rebuilt later. It is not
allowed to invalidate successfully verified learning or knowledge migration.

Real Database Rehearsal

An isolated 589 MB legacy database plus 10,001 source files completed:

• 142 authoritative tables copied;
• 530 rows copied with parity;
• 99 canonical records projected;
• final migration verification;
• index.status=deferred with the 10,000-file reason.

Install Or Update

python -m pip install --upgrade "https://github.com/abhilashsblai/ctxlayer-release/releases/download/v0.2.0a17/ctxlayer-0.2.0a17-py3-none-any.whl"

Installations on 0.2.0a16 use the normal updater:

ctxlayer update --check
ctxlayer update --dry-run
ctxlayer update

After migration, review .gitignore or source-index budgets and run
ctxlayer --repo . index to build the deferred source generation.

ctxlayer 0.2.0a12

CTX Layer 0.2.0a12 publishes the completed cognitive flywheel operationalization as an installable wheel. Source commit: 1ac212e978ae98a75b91098a3a7de66d39d45f0e. SHA256: 035bac0ae91289aa4ca10df9dc0622e9b23fc26a3611ef7351a5d1b74c68bdd8. Wheel size: 728279 bytes.

CTX Layer 0.2.0a11

CTX Layer 0.2.0a11

Published: 2026-07-05

Source commit: df6aabcef885010323bd16bac15456ad5edde5cb

Wheel: releases/ctxlayer-0.2.0a11-py3-none-any.whl

GitHub release:
https://github.com/abhilashsblai/ctxlayer-release/releases/tag/v0.2.0a11

SHA256:
d7ea956517b0e29f1debb005c175609402a5ab8b138427c46e9f09559b472c5a

Wheel size: 695269 bytes

Highlights

• Bumps the package metadata to 0.2.0a11.
• Publishes the memory storage-retention redesign as an installable wheel.
• Adds public documentation for storage-report, retention show, and
  retention set --preset balanced.
• Packages bounded dirty-pack retention, per-repo snapshot fairness,
  session-event/audit summarization, and debounced degraded-state maintenance.
• Preserves the 0.2.0a10 database-evolution, metacognition, governance,
  advanced memory, judgment, workspace-loop, workflow, anticipation, CIE, MCP,
  and write-time guardrail surfaces.

Upgrade Notes

Existing repositories do not need to re-index solely to adopt this version.
After upgrading the CTX Layer package, run:

ctxlayer --repo . doctor
ctxlayer --repo . storage-report
ctxlayer --repo . retention show

For large existing workspaces, back up .ctxlayer/workspace.db first, then use:

ctxlayer --repo . retention set --preset balanced
ctxlayer --repo . gc --deep --dry-run
ctxlayer --repo . gc --deep
ctxlayer --repo . memory bench --samples 3

CTX Layer stores indexed text chunks, vectors, graph edges, context packs,
sessions, and memory metadata in workspace.db. It does not store raw images or
videos as media blobs.

Verification

• python -m pytest tests/test_memory_optimization.py tests/test_service.py tests/test_reliability.py -q
• python -m ruff check src\ctxlayer\__init__.py
• python -m compileall -q src\ctxlayer
• python -m build --wheel
• Wheel smoke install in a temporary venv: ctxlayer --version returned
  ctxlayer 0.2.0a11
• CTX check-diff diff and path gates passed for the source README/version
  changes.

CTX Layer 0.2.0a10

CTX Layer 0.2.0a10

Published: 2026-07-04

Source commit: fde3d05

Wheel: releases/ctxlayer-0.2.0a10-py3-none-any.whl

GitHub release:
https://github.com/abhilashsblai/ctxlayer-release/releases/tag/v0.2.0a10

SHA256:
aec83a11b06c6d1b9259eca702909568d242aa5f5770c5b4a13153221a9dcf06

Wheel size: 686060 bytes

Highlights

• Bumps the package metadata to 0.2.0a10.
• Adds real sqlite-vec vector retrieval and explicit missing-vector failure
  behavior.
• Adds optional DuckDB analytics acceleration with SQLite parity for rollup,
  CIE score, workflow adaptation, judgment distillation, and eval scorecards.
• Adds the Wave 0 database backend benchmark harness and records the Wave 3
  go/no-go decision to defer libSQL and RocksDB/Pebble until evidence requires
  them.
• Includes metacognition, governance activation, judgment overlay, narrative,
  advanced memory architecture, and workspace-loop surfaces.

Verification

• python -m compileall -q src\ctxlayer
• ctxlayer --repo . workflow lint
• python -m ctxlayer.workflow.export_features --check
• pytest tests/test_database_evolution.py tests/test_cie_engine.py tests/test_capability_activation_governance.py tests/test_metacognition_and_agency.py tests/test_phase0_surfaces.py tests/test_phase0_efficacy.py tests/test_nextgen_runtime.py -q
• python -m build --wheel
• Wheel smoke install in a temporary venv: ctxlayer --version returned
  ctxlayer 0.2.0a10

Caveat

The source repo's broad all-path CTX check-diff pack timed out on the full
80-file staged release payload. Impact analysis, focused test/build
verification, smoke install, and representative exact-file CTX groups passed.

CTX Layer 0.2.0a9

CTX Layer 0.2.0a9 Release Notes

Latest wheel: ctxlayer-0.2.0a9-py3-none-any.whl

Release asset:
https://github.com/abhilashsblai/ctxlayer-release/releases/download/v0.2.0a9/ctxlayer-0.2.0a9-py3-none-any.whl

SHA256:
3d231ec7d9f2816acc3928a79764bd2774b2de0f980c43f7a61411b4488f7bbb

Wheel size: 630811 bytes

The 0.2.0a9 build is a preview release for development and non-critical
repositories. Existing users should back up .ctxlayer/workspace.db before the
first run after upgrading.

This release cuts a new wheel artifact for the memory and runtime wiring upgrade
instead of replacing the older 0.2.0a8 asset. It was built on 2026-07-03 from
the Advanced CTX Layer workspace at source base commit
e152f63b9c0a5f7fef2da2895acc13d51dccc921, with package metadata bumped to
0.2.0a9.

Why Upgrade From 0.2.0a8

0.2.0a9 preserves the workflow orchestrator, multi-agent surfaces,
anticipation preview, reliable-enforcement work, large-DB performance work,
memory optimization, deep-GC maintenance, skill evolution, write-time semantic
guardrails, and NextGen runtime surfaces from the latest CTX Layer source state.

The important release distinction is packaging: this version gives those memory
and wiring updates their own installable wheel, release tag, digest, and release
notes instead of publishing them as another replacement under v0.2.0a8.

Install

python -m pip install --upgrade "https://github.com/abhilashsblai/ctxlayer-release/releases/download/v0.2.0a9/ctxlayer-0.2.0a9-py3-none-any.whl"

Then verify:

ctxlayer --version
ctxlayer --repo . doctor
ctxlayer --repo . workflow lint
ctxlayer --repo . pack --task "release update smoke test"

CTX Layer 0.2.0a8

CTX Layer 0.2.0a8 preview release.

Includes the workflow orchestrator, workflow lint CLI, deterministic feature workbook drift check, snapshot-aware workflow resume, CIE surface expansion, multi-agent setup surfaces, anticipation preview, large-DB performance work, memory maintenance, skill evolution, and write-time semantic guardrails.

Wheel SHA256: a348d2147aca306ebdde40ad0734b2921dd219289ebdabc984e0baf2fc39cec6
Built from Advanced-CTX-Layer source commit: 9c6d227

ctxlayer 0.2.0a7

Preview wheel for the post-0.2.0a6 source line.

Includes:

• Multi-agent adapter layer for Codex, Claude Code, and Cursor.
• Full sixth-sense anticipation hardening path.
• Reliable-enforcement and large-DB performance updates.
• Fluid-memory optimization, deep-GC, CIE preview, skill-evolution, and write-time semantic guardrails.

Build details:

• Advanced-CTX-Layer source commit: 9f7d041
• Wheel SHA256: af17b6051e70683c1625725cf59d2b9a96081542957254c9eac0b33d06ef0cc2
• Wheel size: 550982 bytes

ctxlayer 0.2.0a6

Preview release built from Advanced-CTX-Layer source commit 01c82a1. Adds the anticipation layer preview with live Codex surprise scoring, calibrated world-model ECE, expected-free-energy lookahead, interoceptive gut state, online learning triggers, and enforce-mode calibration gating.