Changelog

All notable changes to FixLab are documented in this file.

Unreleased

0.7.1 - 2026-10-05

• Refresh plugin discovery metadata for the GitHub Copilot marketplace with product-neutral React/.NET positioning and focused agent-tooling keywords.

0.7.0 - 2026-10-05

• Add an Agency-like fixlab <repository> interactive entry point for conversational setup, diagnosis, repair, and validation.
• Add fixlab chat as a terminal client for the active dashboard session, including live output, status, evidence, PR, resume, retry, and owned-executor stop commands.
• Let fixlab chat select the repository and start the loopback dashboard automatically when no dashboard is already running.
• Keep the interactive fixlab> input visible and preserve partially typed text while dashboard activity and logs stream.
• Confirm the connected repository when chat starts and allow an alternate repository to launch automatically on the next available loopback port.
• Recognize common chat operations in plain English so dashboard, status, logs, evidence, pull-request, help, and repository switching do not require slash syntax.
• Let new bug <details> and add a bug: <details> create a separate dashboard job instead of commenting on the active job.
• Let users remove waiting jobs by queue position or job ID from terminal chat or the dashboard while preserving cancelled history and cleaning queued screenshot artifacts.
• Buffer background activity while the user is typing so Windows terminal redraws cannot split one instruction into repeated partial comments.
• Document the simplified chat-first installation, onboarding, UI-bug, API-bug, end-to-end validation, recovery, and pull-request workflow.
• Show periodic dashboard activity heartbeats while long-running builds, tests, and browser validation are quiet without resetting the idle watchdog.
• Add a packaged, plan-first fixlab onboard command so users can initialize, review, prepare, authenticate, verify, and start a repository without a machine-specific helper script.
• Add a plan-first, checksum-verified portable Node.js repair for incoherent NVM installations without changing NVM or the permanent system PATH.
• Keep the Node 18 compatibility test suite runnable on the declared minimum runtime while using a newer supported runtime for development Playwright.
• Keep backend-first validation and transient synthesized-Playwright cleanup guidance aligned across dashboard, CLI, packaged-agent, and initialized-agent entry points.
• Promote a scoped learned rule and regression gate for cross-runtime behavioral guidance parity.
• Forward CLI requests to Agency over standard input and verify that backend-first guidance appears only when both an approved environment and non-production read-only data are selected.
• Keep request-free runs interactive in both runtimes by reserving stdin forwarding and autonomous invocation flags for runs that carry a request, and preserve inherited terminal input for interactive use.
• Require code-owner approval for non-admin contributors while retaining an administrator bypass for the repository owner's own validated changes.
• Clarify that request-bearing CLI runs forward prompts through stdin while request-free runs inherit terminal input and remain interactive.
• Add command-aware Node/TAP summaries with persisted context metrics so agents receive compact results while redacted raw evidence remains available.
• Sanitize command labels, report exact emitted-summary bytes, and keep timeout enforcement reliable when a child process never closes its output streams.
• Replace pilot-specific documentation and test fixtures with reusable React/.NET examples suitable for the public repository.

0.6.1 - 2026-09-21

• Keep synthesized authenticated Playwright scenarios out of product pull requests unless permanent browser-test coverage is explicitly requested or required by repository instructions.

0.6.0 - 2026-09-21

• Refresh installation, release, marketplace, availability, and team-rollout guidance to distinguish current GitHub distribution from pending npm, marketplace, and hosted-service work.
• Identify MonaDevAI as the plugin publisher in marketplace-facing metadata.
• Warn when fixlab init preserves an older profile that requires newly mandatory fields, and document the explicit test-synthesis upgrade.
• Explain directly in the dashboard how to correct and resume a failed active job before its paused queue can continue.
• Add an explicit, confirmed failed-job dismissal that preserves the failed history entry and advances the queue without resuming repository work.
• Clear and refocus the bug input after every successful submission, including jobs added to the running queue.
• Use an explicitly selected DEV or SIT read-only backend as the primary data source, with visible synthetic fallback only when required data is unavailable and without claiming backend validation.

0.5.4 - 2026-09-18

• Create tagged GitHub releases independently of optional npm publication and require an explicit repository opt-in before publishing to the npm registry.

0.5.3 - 2026-09-18

• Install the Playwright Chromium runtime before the tagged release validation gate so browser tests can run on a fresh GitHub-hosted runner.

0.5.2 - 2026-09-18

• Recover persisted running jobs as interrupted and resumable after dashboard restart.
• Stop an owned executor after a configurable no-output interval and display the latest output age so stalled jobs do not appear active indefinitely.
• Add first-class Playwright test synthesis with repository-configured synthetic/intercepted data, structured scenario evidence, and a highlighted dashboard summary of data source and mutation behavior.
• Keep unrelated test, lint, type-check, and production-build diagnostics out of the local application-startup stage so independently ready Playwright validation can continue.
• Add an optional direct GitHub Copilot CLI runtime for the CLI and dashboard while retaining Agency as the default compatibility runtime.
• Add repository-configured Playwright authentication status/connect controls and same-session comments for running dashboard jobs.
• Highlight the inferred active roadmap step while a running agent is between explicit stage updates.
• Load bounded Azure DevOps comments and images, support clipboard-pasted screenshot evidence, and expose access warnings without discarding bugs.
• Add a visible 20-job running queue with same-session blocker handling.
• Add privacy-safe period statistics for bugs, execution time, exact Copilot token usage, and input cache reuse.
• Add a repository-installed fixlab-autofix Visual Studio Code agent and /fixlab.bugfix prompt for one-bug reproduce, repair, and verification workflows.
• Add versioned autofix/evidence specifications, enforced repository policy checks, a scheduled GitHub agentic learned-rule review, and automatic revert pull-request proposals for failed trusted pushes.
• Add executable proof-of-bug/proof-of-fix coverage for flaky containment and multi-commit rollback restoration, plus a scheduled proof drill.
• Require code-owner review, stale-approval dismissal, and passing CI/CodeQL checks for normal changes to main.
• Add Agent Plugins 1.0 marketplace release guidance, correct the documented GitHub Copilot agent identifier, and provide a concrete private vulnerability reporting route.

0.5.1

• Add the reusable FixLab logo to the public repository and packaged assets.
• Display the product identity prominently in the GitHub README.

0.5.0

• Add manual or Azure DevOps work-item intake.
• Add bounded local screenshot evidence without storing tokens, request attachments, or screenshot paths in the repository metadata cache.
• Sanitize loaded work-item content and use the developer's local Azure authentication without persisting bearer tokens.

0.4.0

• Add an explicit, dependency-free local dashboard command for any onboarded repository.
• Add bug-fix and risk-scaled small-enhancement request types.
• Add token-efficiency guidance, bounded dashboard logs, and a safe, metadata-only same-repository cache.
• Show repository readiness, staged job progress, and polled local logs.
• Enforce single-job execution, validate-only protections, complete terminal stage markers, safe static serving, and owned-child shutdown.
• Package the dashboard UI and add integration coverage with a fake executor.

0.3.1

• Fix fixlab run and fixlab validate to use Agency's resolved FixLab plugin qualifier.
• Add regression coverage for launching the packaged FixLab agent.

0.3.0

• Add repository-installed FixLab prompt commands.
• Add GitHub Actions for continuous integration and npm releases.
• Add MIT licensing and GitHub contribution templates.
• Add React/.NET onboarding examples and release documentation.

0.2.0

• Add guided repository-local Playwright setup.
• Support npm, pnpm, Yarn, and configured browser channels.
• Verify the configured browser with a real headless launch.

0.1.0

• Package the FixLab CLI and Agency Copilot plugin.
• Add repository profiles, validation-only workflows, and onboarding guides.