Historie doplňku QA Trace - AI Bug Reports, Private & Local - 5 verzí
QA Trace - AI Bug Reports, Private & Local od Anastasia Nesterova
Historie doplňku QA Trace - AI Bug Reports, Private & Local - 5 verzí
Se starými verzemi opatrně! Tyto verze jsou zde dostupné jen pro testovací a referenční účely.Používat byste vždy měli nejnovější verzi doplňku.
Nejnovější verze
Verze 1.1.1
Vydáno 6. zář 2026 – 116,31 KBFunguje s: firefox 142.0 a novější=== New
- Network request tracking (opt-in, per URL). When enabled, every fetch/XMLHttpRequest call is recorded (method, URL, status, request/response headers and bodies), not only failures. Resource loads issued by the browser itself (document, scripts, styles, images, fonts) are not captured.
- Detail view. Errors and network requests now have an Open in new tab button that opens a full-page view with time, tab, message, stack, request/response method, URL, status, headers and bodies, plus Copy and Screenshot actions.
- Actions picker replaces "Number of Actions to Include". The prompt configuration screen lists the actions in scope. Click Start here on an action to include everything from it up to the latest one, or toggle All actions.
- Store full network bodies (disable size reduction). By default request/response bodies are scanned up to 100,000 characters and stored truncated to 12,000. Enable this option under Limits to keep full bodies (up to 9 MB per response). Sensitive fields are still redacted.
- Storage notifications. If browser storage fills up, QA Trace drops the oldest data, keeps recording and shows a notification.
=== Privacy and security
- Bearer tokens and JWTs are now redacted wherever they appear in request or response bodies, not only inside known sensitive fields.
- Binary responses (images, media, fonts, archives, PDFs) are skipped instead of stored.
- Body reading is capped by bytes, so large or streaming responses are never fully buffered. Oversized responses are replaced with a size note.
=== Improvements
- Page hooks are injected at document_start, so requests and errors fired during page load are captured. Previously anything before DOMContentLoaded could be missed.
- UI error screenshots are stored as JPEG instead of PNG to save space. Copying a screenshot still produces PNG.
=== Fixes
- The LLM API Key field on the configuration page showed the webhook password hint.
- Tab scope filtering compared tab ids inconsistently, which could exclude actions or errors from the selected tabs and mis-report the "All tabs" state.Zdrojový kód zveřejněn pod licencí Pouze GNU Affero General Public License v3.0
Starší verze
Verze 1.1.0
Vydáno 5. zář 2026 – 116,35 KBFunguje s: firefox 142.0 a novější=== New
- Network request tracking (opt-in, per URL). When enabled, every fetch/XMLHttpRequest call is recorded (method, URL, status, request/response headers and bodies), not only failures. Resource loads issued by the browser itself (document, scripts, styles, images, fonts) are not captured.
- Detail view. Errors and network requests now have an Open in new tab button that opens a full-page view with time, tab, message, stack, request/response method, URL, status, headers and bodies, plus Copy and Screenshot actions.
- Actions picker replaces "Number of Actions to Include". The prompt configuration screen lists the actions in scope. Click Start here on an action to include everything from it up to the latest one, or toggle All actions.
- Store full network bodies (disable size reduction). By default request/response bodies are scanned up to 100,000 characters and stored truncated to 12,000. Enable this option under Limits to keep full bodies (up to 9 MB per response). Sensitive fields are still redacted.
- Storage notifications. If browser storage fills up, QA Trace drops the oldest data, keeps recording and shows a notification.
=== Privacy and security
- Bearer tokens and JWTs are now redacted wherever they appear in request or response bodies, not only inside known sensitive fields.
- Binary responses (images, media, fonts, archives, PDFs) are skipped instead of stored.
- Body reading is capped by bytes, so large or streaming responses are never fully buffered. Oversized responses are replaced with a size note.
=== Improvements
- Page hooks are injected at document_start, so requests and errors fired during page load are captured. Previously anything before DOMContentLoaded could be missed.
- UI error screenshots are stored as JPEG instead of PNG to save space. Copying a screenshot still produces PNG.
=== Fixes
- The LLM API Key field on the configuration page showed the webhook password hint.
- Tab scope filtering compared tab ids inconsistently, which could exclude actions or errors from the selected tabs and mis-report the "All tabs" state.Zdrojový kód zveřejněn pod licencí Pouze GNU Affero General Public License v3.0
Verze 1.0.2
Vydáno 5. čvn 2026 – 101,25 KBFunguje s: firefox 142.0 a novější- Added an option to redact the origin (protocol, host, and port) from URLs, error messages, stacks, headers, and bodies before data is sent to an LLM or webhook; the locally stored copy keeps full URLs.
- Network requests that finish with a 0 response status (aborted, cancelled, or CORS-blocked) are no longer reported as failed requests, reducing noise in the captured data.
- Updated dependency versions.
Zdrojový kód zveřejněn pod licencí Pouze GNU Affero General Public License v3.0
Verze 1.0.1
Vydáno 23. dub 2026 – 98,86 KBFunguje s: firefox 142.0 a novějšív1.0.1
Added per-URL error monitoring toggle — each allowed URL can now have error detection disabled individually while action tracking remains active
Fixed UI error detection stopping after the first observed mutationZdrojový kód zveřejněn pod licencí Pouze GNU Affero General Public License v3.0
Verze 1.0.0
Vydáno 14. dub 2026 – 98,37 KBFunguje s: firefox 142.0 a novějšíZdrojový kód zveřejněn pod licencí Pouze GNU Affero General Public License v3.0