Skip to content

External project save/load: canonical files, durable execution and independent forks - #119

Merged
MrScripty merged 9 commits into
mainfrom
feat/external-project-save-load-20261022
Oct 23, 2026
Merged

MrScripty merged 9 commits into
mainfrom
feat/external-project-save-load-20261022

Conversation

@MrScripty

@MrScripty MrScripty commented Oct 22, 2026 •

Copy link
Copy Markdown
Owner

Users can own a Tuldok project in an external Git-compatible directory and reopen it with a fresh PostgreSQL projection. Canonical text retains annotations, decisions, observations, source history and lineage; explicit manifests describe untracked unique artifacts, weights, runtime binaries and generated data. Open is inert and imports no credentials, live grants, admissions or resumable execution.

Published complete implementation candidate: 5704965, based on main 133d53c. Coding-Standards reviewed at 5bce3110fc64d61ff8b4f79468184bd6295855d5. M1–M5 source is implemented; this remains draft / Verifying, with acceptance conditional on successful required checks for this exact candidate. Do not mark ready or merge while a gate is missing or failed.

  • New/Open/Save/Close use atomic canonical checkpoints and durable acknowledged edits; SQL is rebuildable. Current custody remains installation-local, with explicit original-host recovery and conservative fencing of uncertain effects.
  • Finite domain codecs preserve immutable scientific identities and history. Explicit generic script/native execution uses canonical and SQL acknowledgment before GO; historical imported facts cannot activate a process.
  • Save As creates a new project and editable identities while preserving exact historical hashes and copying required unique artifacts without hardlinks. Relocation, Git checkout/stale projections, explicit working-file adoption, and retained note-deletion/restoration history have fixed controls.
  • Mixed-catalog extraction is report-first and requires an actual stopped source, private backup/inventory and independent destination authority. The user's local surrogate lab was never inspected or migrated. Tuldok gained no Rheon-specific integration knowledge.

Validation and review:

  • Independent complete source review and targeted repair rereviews are clear. Final contract, enabled composition, bounded test deadlines and CI selection/portability review are clear. All 829 files in the final qualified product manifest match this candidate except three reviewed test-only portability/deadline deltas.
  • Disposable real PostgreSQL/process/HTTP/SDK/CLI/MCP/Chromium controls passed on recorded immutable subjects, including source-database CONNECT denial to a separate cold role. Actual CPU native Runs trained and separately verified checkpoints; retained two-Run Study proof and all 79 artifact hashes cold-reconstructed with zero imported execution authority and fresh disclosure required.
  • Earlier complete native cases remain ERROR/FAIL where fixture expectations or a 180-second test wait stopped them. Corrected no-fit tails reused the same measured evidence; original-host final Study SQL assertions were not reached. This is not a claim that those original whole cases passed. Final CI selects all three corrected native domain cases, real public/browser gates and the approved-agent cold-denial tail, with exact/nonzero counts and no skips.
  • M2 lifecycle run 39842528122 failed at the observation browser fixture; the other three workflows passed. Its response-arrival/DOM-application race repair has a scoped real Chromium pass. Failed evidence is retained; no duplicate full-suite or training campaign was run locally.
  • Exact candidate CI: current head f266a70 has three successful workflows: M1 interface 39903396897, extension catalog 39903396885 and captured CPU 39903396900. Lifecycle 39903396876 reached full Python discovery (step 44) at 05:23 UTC, with every preceding step successful, including native execution, SDK/CLI/MCP, native public controls and note deletion. Full discovery, later browser gates and terminal exact counts/no-skip evidence for this candidate remain pending. The PR remains draft / acceptance pending. Prior head 3ac3351 lifecycle 39893561852 FAILED at classification fixture readiness: its old make_handler wrapper rejected the production transfer_controller keyword before any proposal flow ran. That candidate passed full discovery (2,332 cases, zero failures/errors, 107 optional/intentional skips), native domain 3/3, native public 1/1 and note deletion 1/1 without skips, all six project browsers and the earlier observation browser. Later steps 153–180 were skipped. The same wrapper incompatibility was found in the later Pumas fixture before CI reached it. Both wrappers now explicitly forward the keyword, preserving all instrumentation; each full real-PG/Chromium browser procedure passed once in synthetic-HTTP mode, with no application exceptions and all 912 source/test hashes stable. Independent source/evidence review is clear. This six-file commit changes two fixtures and four evidence documents; product and workflows remain unchanged from 5704965. Prior 591eb24 Study wait failure, 5704965 full-suite failure and intermediate HTTP failures remain recorded, including unobserved timed-out Run/Open outcomes. No same-head rerun, broad local suite or native fit was used. Current test merge 63a7e24191e3c0c8e98b482f8dc1e5604a9c6434 and feature share tree 3aa6cd7a36f5f20a8e13fbaf4994ca182e200da9; actual CI checkout-log proof awaits terminal logs. Uploaded ZIP manifests remain uninspected; no denied retrieval route was retried. Handler fixture diagnosis and exact evidence.

Supported writer scope is Linux/POSIX local filesystems honoring locks/rename/fsync with local PostgreSQL. Installed native packages remain operator-managed; no full dependency-closure, model-quality, distributed-writer, semantic Git merge, automatic fetch/install/regeneration or user-data migration claim is made.

Canonical plan and conditional acceptance, implementation review, native performed/failed/unrun evidence, public contract. Required exact-head checks and their CI artifacts supply terminal acceptance without a status-only source commit.

@coderabbitai

coderabbitai Bot commented Oct 22, 2026 •

Copy link
Copy Markdown

Important

  • 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration
  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: defaults
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 8afe6fef-417a-41a7-a566-8ff4e44fac8e
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

Add fixed data/material/history/producer codecs, verified artifact custody, inert cold projections, explicit current source bindings and complete nonexecuting public metadata routes. Retain scoped real PostgreSQL/process/client/browser evidence and resolved independent review. Execution, public Save As and extraction remain fenced for M3-M5.
@MrScripty MrScripty changed the title feat(project): implement external project save/load plan External project save/load: canonical files and inert reconstruction (M1–M2; draft) Oct 22, 2026
@MrScripty MrScripty changed the title External project save/load: canonical files and inert reconstruction (M1–M2; draft) External project save/load: canonical files, durable execution and independent forks Oct 22, 2026
@MrScripty
MrScripty marked this pull request as ready for review October 23, 2026 06:19
@MrScripty
MrScripty merged commit 9ae37f7 into main Oct 23, 2026
5 checks passed
@MrScripty
MrScripty deleted the feat/external-project-save-load-20261022 branch October 23, 2026 06:19
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant