Repository navigation
MangoThonny v0.4.3
MangoThonny v0.4.3 Release Notes
Date: 2026-09-10
Status: Stable
Persistent Assignment and learning-process release
MangoThonny v0.4.3 extends the v0.4.2 Learning Log workflow into a persistent Assignment Workspace that can follow one learner task across multiple runs, editor restarts, days, and computers while preserving the existing Event Schema 1.1 contract.
The release also adds complete Host Python and High-Level MicroPython run lifecycle capture, compact high-rate sensor projection, and cross-process summary reconciliation so continuous sensor polling remains useful for research without expanding student-facing assignment history into thousands of raw records.
Added / changed
- Added MangoBox Assignment Workspace with
student_id,task_id, persistentassignment_id, workspace-local metadata, snapshots and resumable assignment state; - added
帶走作業transfer ZIP for continuing the same assignment on another computer; - added
匯入作業with path rebasing while preserving the originalassignment_idand prior learning history; - added
完成並匯出作業for final submission-oriented learning-history packaging; - added ordinary desktop/Host Python run lifecycle capture with
run_start,code_snapshot,program_output,runtime_errorand terminal run status; - added Host frontend lifecycle fallbacks for Thonny configurations where
CommandAcceptedis missed, including assignment rediscovery from the current source-file path; - added High-Level MicroPython lifecycle parity, including Pico backend detection, code snapshots, student-visible stdout/stderr capture, stderr-only exception inference and shared
run_idwith Student API hardware events; - hardened Host/High-Level backend separation for
RP2040,RPiPico,GenericMicroPython, MicroPython and CircuitPython backend identifiers; - added assignment-scoped external-edit gap detection so edits made outside MangoThonny can be represented without pretending the missing editing steps were observed;
- added compact learning projection with
KEEP,SUMMARIZEandTECHNICAL_ONLYpolicies; - added bounded background projection queue (
4096) and constant-memory, run-scoped continuous-sensor aggregation; - added cross-process run-end reconciliation so Host Python frontend lifecycle events can finalize sensor aggregates produced by the child runtime process;
- continuous sensor polling is summarized as
hardware_summaryrather than copied as one learning record per sample; - normalized Host polling semantics for Light, Sound, Joystick, PIR, IR, Ultrasonic and Obstacle Student API reads while preserving true device/environment transitions as semantic events;
- Ultrasonic
distance()summaries retaindistance_cmand logical sensor identity; - Ultrasonic
is_near()summaries retaindistance_cm,threshold_cm, logical sensor identity andtrue_count; - transfer/submission exports merge/deduplicate projection shards while raw high-frequency technical records remain outside student-facing bundles;
- builder/release regression now gates Host frontend lifecycle, Assignment context/propagation/import/assets, High-Level MicroPython lifecycle, learning projection, high-rate aggregation and cross-process summary reconciliation.
Compatibility retained
- Thonny base:
4.1.7; - bundled Python:
3.10.11; - Host Student API:
0.4.6; - canonical Host import:
mangobox; - legacy Host import:
mangox2; - Event Schema:
1.1; - Standard Teaching Packages v1;
- Mango App Exporter;
- Windows AppUserModelID / taskbar identity from
v0.4.1; v0.4.2Learning Log visibility and per-user canonical store behavior.
Runtime firmware remains versioned and released independently. The matching current Runtime baseline remains MangoX2 v0.2.7 and MangoLite v0.6.2.
Functional validation
Recorded validation on the v0.4.3 development/release line includes:
- complete Learning Log / Assignment / projection regression gate: PASS;
- Host frontend lifecycle fallback and backend-separation regression: PASS;
- High-Level MicroPython lifecycle regression: PASS;
- 100,000 direct high-rate sensor events retained as 5 constant-memory groups: PASS;
- 50,000 worker submissions remained bounded and produced only terminal learning records plus hardware summary: PASS;
- High-Level MicroPython physical assignment lifecycle including code snapshot, Student API events, stdout, runtime error and
run_endstatus: PASS; - Host Python physical Assignment lifecycle and transfer export: PASS;
- Light high-rate physical test (
m.light()loop) summarized into onelight/read_summarywith numeric statistics and no raw polling leakage: PASS; - Ultrasonic physical
distance()aggregation with 100 readings and distance statistics: PASS; - Ultrasonic physical
is_near(30)aggregation with 100 readings, threshold statistics andtrue_count: PASS; - cross-computer Assignment transfer/import/continue/re-export with stable
assignment_id, preserved prior history and distinct computer identity: PASS; - exact Portable build/package verification: PASS;
- exact Installer build: PASS;
- installed-mode install/launch/shortcut/metadata/uninstall-preserve/reinstall smoke: PASS.
The functional code-freeze baseline before release-document commits is:
9b2a1673f2d5dcf3f21b5b2fa6b0eb728cd5c8e3
Release-document commits after this baseline do not change MangoThonny runtime behavior.
Exact release artifacts
Built locally from source commit:
f5f0b5db8e23ab968b071193e522f474e12fd28f
MangoThonny-Portable-v0.4.3.zip
SHA-256: 5f1edb5ba895a044477d788ed8ff4257ed8109b449aebb1f247397eafb04991b
MangoThonny-Setup-v0.4.3.exe
SHA-256: f79cf8411721edcbfd2c84fd8c9ff2590ffd4e51e215a1fa9e9fb1d02660246d
Size: 92.2 MB
Language: Traditional Chinese bundled overlay
The Installer was built from the exact verified Portable ZIP above and re-verified that Portable artifact before packaging.
Installed metadata reports:
mangothonny_version = 0.4.3
base_thonny_version = 4.1.7
bundled_python_version = 3.10.11
host_library_version = 0.4.6
source_commit = f5f0b5db8e23ab968b071193e522f474e12fd28f
Promotion status
Completed:
- exact
v0.4.3Portable Build + Package Verification: PASS; - exact Portable SHA-256 recorded;
- Installer built from that exact verified Portable artifact: PASS;
- Installer build PASS and SHA-256 recorded;
- install / launch / shortcut / upgrade-install / uninstall-preserve / reinstall smoke: PASS;
- installed metadata reports the expected
v0.4.3artifact identity: PASS; - release validation approved.
v0.4.3 is approved as the stable MangoThonny release line. Public GitHub Release publication and asset upload, if performed, are separate distribution steps.
See V0.4.3_VALIDATION.md for detailed evidence.