This release adds 3 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Topics
+4 more
Summary
AI summaryEdit transaction layer improved for deterministic recovery and safer handling of unsafe matches.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Medium |
Adds deterministic exact-text recovery for edit transactions. Adds deterministic exact-text recovery for edit transactions. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Feature | Medium |
Distinguishes unrecovered edit‑mechanics failures as blocking and recovered failures as non‑poisoning. Distinguishes unrecovered edit‑mechanics failures as blocking and recovered failures as non‑poisoning. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Feature | Medium |
Adds generic project toolchain layer and Python .venv detection as a scoped adapter. Adds generic project toolchain layer and Python .venv detection as a scoped adapter. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Feature | Low |
Python validation now prefers the project .venv interpreter when available. Python validation now prefers the project .venv interpreter when available. Source: granite4.1:30b@2026-06-05-audit Confidence: low |
— |
| Feature | Low |
Runs `pytest`, `ruff`, and `mypy` only when they are available, requested, or relevant to the project. Runs `pytest`, `ruff`, and `mypy` only when they are available, requested, or relevant to the project. Source: granite4.1:30b@2026-06-05-audit Confidence: low |
— |
| Feature | Low |
Blocks global installs by default unless explicitly approved, enforcing use of project‑local environments. Blocks global installs by default unless explicitly approved, enforcing use of project‑local environments. Source: granite4.1:30b@2026-06-05-audit Confidence: low |
— |
| Feature | Low |
Improves live Worker summary so task completion is visible immediately without a reload. Improves live Worker summary so task completion is visible immediately without a reload. Source: granite4.1:30b@2026-06-05-audit Confidence: low |
— |
| Feature | Low |
Adds protection so syntax‑highlighter failures do not spam the console or destabilize the UI. Adds protection so syntax‑highlighter failures do not spam the console or destabilize the UI. Source: granite4.1:30b@2026-06-05-audit Confidence: low |
— |
| Bugfix | Medium |
Improves handling of newline‑normalized, trimmed‑whitespace, and surrounding‑context matches in edit transactions. Improves handling of newline‑normalized, trimmed‑whitespace, and surrounding‑context matches in edit transactions. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Bugfix | Medium |
Causes ambiguous or unsafe matches to fail clearly without mutating files. Causes ambiguous or unsafe matches to fail clearly without mutating files. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Bugfix | Medium |
Applies same‑file edits against an updated in‑memory buffer instead of stale disk state. Applies same‑file edits against an updated in‑memory buffer instead of stale disk state. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Bugfix | Medium |
Includes structured details in edit‑failure reports (operation, reason, candidate count, context). Includes structured details in edit‑failure reports (operation, reason, candidate count, context). Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Bugfix | Medium |
Reports actual failed edit operation on recovery exhaustion instead of vague noise. Reports actual failed edit operation on recovery exhaustion instead of vague noise. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Bugfix | Medium |
Requires follow‑up for files changed without validation and deduplicates modified file reporting. Requires follow‑up for files changed without validation and deduplicates modified file reporting. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Bugfix | Medium |
Reports missing local tools/dependencies as environment setup issues, not fake code validation failures. Reports missing local tools/dependencies as environment setup issues, not fake code validation failures. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Bugfix | Low |
No longer treats missing global `pytest` as a code validation failure. No longer treats missing global `pytest` as a code validation failure. Source: granite4.1:30b@2026-06-05-audit Confidence: low |
— |
| Bugfix | Low |
Prevents non‑Python projects from receiving Python‑specific venv behavior or pytest/py_compile assumptions. Prevents non‑Python projects from receiving Python‑specific venv behavior or pytest/py_compile assumptions. Source: granite4.1:30b@2026-06-05-audit Confidence: low |
— |
| Bugfix | Low |
Enhances bottom‑of‑chat spacing and scroll behavior around the input panel and active spec cards. Enhances bottom‑of‑chat spacing and scroll behavior around the input panel and active spec cards. Source: granite4.1:30b@2026-06-05-audit Confidence: low |
— |
| Bugfix | Low |
Reduces UI weirdness related to final summaries, spec card cleanup, and bottom scrolling. Reduces UI weirdness related to final summaries, spec card cleanup, and bottom scrolling. Source: granite4.1:30b@2026-06-05-audit Confidence: low |
— |
Full changelog
Aura IDE v1.4.6
This is a reliability and workflow release focused on making Aura feel smoother, stricter, and less noisy during real development.
Highlights
Safer edit mechanics
Aura’s edit transaction layer is now more resilient when applying model-generated changes.
- Added deterministic exact-text recovery for edit transactions.
- Improved handling for newline-normalized, trimmed-whitespace, and surrounding-context matches.
- Ambiguous or unsafe matches now fail clearly without mutating files.
- Same-file edits are applied against an updated in-memory buffer instead of stale disk state.
- Edit failures now include structured details like failed operation, reason, candidate count, and context.
- Recovery exhaustion now reports the actual failed edit operation instead of vague follow-up noise.
Better Worker outcome truthfulness
Worker results now distinguish more clearly between different kinds of failures and caveats.
- Unrecovered edit-mechanics failures remain blocking.
- Recovered edit failures no longer poison later successful writes.
- Files changed without validation now correctly require follow-up.
- Modified file reporting is deduped.
- Environment caveats are no longer misclassified as edit failures.
- Missing local tools/dependencies are reported as environment setup issues, not fake code validation failures.
Project-local toolchain support
Aura now respects project-local development environments instead of assuming global tools.
- Added a generic project toolchain layer.
- Added Python
.venvdetection as a Python-specific adapter. - Python validation now prefers the project
.venvinterpreter when available. pytest,ruff, andmypyare only run when available/requested/relevant.- Aura no longer treats missing global
pytestas a code failure. - Global installs are blocked by default unless explicitly approved.
- Non-Python projects are not treated as Python projects.
Language-agnostic workflow preservation
Aura remains a general coding harness, not a Python-only IDE.
- HTML/CSS/JS projects do not get Python venv behavior.
- Non-Python projects avoid pytest/py_compile assumptions.
- Project-local toolchains are handled generically, with Python support scoped behind the Python adapter.
- Validation behavior now depends on the project and task instead of hardcoded Python assumptions.
GUI and workflow polish
- Improved live Worker summary behavior so task completion is visible immediately instead of only after reload.
- Improved bottom-of-chat spacing/scroll behavior around the input panel and active spec cards.
- Reduced UI weirdness around final summaries, spec card cleanup, and bottom scrolling.
- Added protection around syntax highlighting so highlighter failures should not spam the console or destabilize the UI.
Validation
Focused validation passed for:
- edit transaction reliability tests
- Worker edit recovery tests
- project-local Python environment tests
- terminal policy tests
- prompt/toolchain behavior tests
- focused dispatch classification tests
- py_compile on changed Aura modules
Why this release matters
This release is aimed at reducing Worker thrash during normal coding work.
Aura should now be better at:
- applying edits safely
- recovering from edit mismatches
- telling the truth about what happened
- using the right project environment
- avoiding bogus validation failures
- keeping Python behavior scoped to Python projects
- making live task completion easier to trust
In plain terms: Aura should feel less brittle, less noisy.
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About Aura
All releases →Related context
Related tools
Earlier breaking changes
- v1.4.0 All file-writing paths now route through a validation pipeline.
Beta — feedback welcome: [email protected]