Changelog
Back to KidQACurrent version: 0.20.0
v0.20.0
2026-06-18- History now groups a visit's questions into one conversation, so you can review a whole back-and-forth together
- Added search and filters to History — find past turns by text, or filter by voice mode, redirected answers, or your thumbs up/down
- Each past answer now shows which voice mode produced it (Realtime, Standard, xAI, or Gemini)
v0.19.0
2026-06-18- Redesigned the History page for readability: each answer now reads as three clear sections — the question and answer, where the image came from, and your feedback
- Your written feedback notes are now bigger and wrap to as many lines as needed, instead of being cut off
- Image details are clearer — the keywords that came back from Pixabay are now labeled “Pixabay tags” (the “Image search” line is still the term KidQ looked up)
- Each past answer now shows your thumbs up/down at a glance, and flags answers that were redirected to a grown-up
v0.18.1
2026-06-18- Fixed the Realtime voice mode occasionally not showing the answer it just spoke — every spoken answer now reliably appears on its own card
- Fixed the Realtime voice mode sometimes logging a wrong (negative) answer-speed number when a question wasn't transcribed or when a child talked over the answer
v0.18.0
2026-06-18- Answer-speed logging now measures the time from when a child stops talking to the first word of the answer — captured the same way across all four voice modes, so response speed can be compared fairly
- Fixed the Gemini voice mode logging the wrong answer-speed number (it had recorded the time to finish the whole answer instead of the time to the first word)
v0.17.0
2026-06-18- Each visit now starts a fresh conversation: refreshing the page or opening a new tab begins a new session and the model starts listening fresh — the same reset the History “Clear” button does
- Within a visit, follow-up answers now remember the whole conversation so far (across all voice modes), instead of just the last few turns
v0.16.0
2026-06-17- The Gemini voice mode now speaks in a warmer voice (Sulafat) instead of the previous breezy default
v0.15.1
2026-06-17- Answers a child hears but interrupts are now captured in answer-quality logging instead of being dropped — tagged with why they were cut short — across all voice modes, including Gemini
- Voice answer logs now record the model that actually answered and order multi-turn sessions accurately for review
v0.15.0
2026-06-16- New Gemini voice mode at /gemini — voice-to-voice Q&A powered by Google's Gemini Live API, with the flashcard headline, image, and thumbs up/down feedback like the other modes
v0.14.0
2026-06-16- You can now give a thumbs up or thumbs down on xAI voice answers, right on the answer screen — matching Standard and Realtime modes
v0.13.0
2026-06-16- The xAI Grok voice mode moved from /sd-voice to /xai and now speaks in the built-in “Ara” voice
- Fixed a spurious “xAI API error” message that could flash on screen during a normal answer
- xAI voice questions are now included in answer-quality telemetry, matching the other two modes
v0.12.0
2026-06-16- Answer logging now records the recent conversation leading up to each answer, so a follow-up question and the answer it produced can be reviewed together instead of in isolation
v0.11.0
2026-06-15- Spoken answers are now complete sentences that stand on their own, instead of short fragments that only made sense alongside the headline on screen
v0.10.0
2026-06-14- You can now give a thumbs up or thumbs down on any answer — right on the answer screen and from the History page — and leave an optional note
v0.9.0
2026-06-14- History now shows the image search term and the keywords Pixabay returned for each picture, so you can see at a glance whether the image actually matched the question
- Each history entry now logs image match details (number of results, dimensions, popularity, source link) and embeds a small text-encoded copy of the picture, so a copied history trace is self-contained
- Darker, higher-contrast text throughout for easier reading
- Renamed the app to KidQA across the interface
v0.8.0
2026-06-05- Standard voice mode now waits for a brief pause before answering, so it no longer cuts kids off mid-question
- Clear on-screen messages when the microphone is blocked or missing, or the browser doesn't support voice — across Realtime, Standard, and Sd-Voice
- Realtime and Sd-Voice now remember earlier questions after a reconnect or reload, and recover cleanly from dropped or stalled connections instead of leaving the mic on
- More reliable answer playback and clearer question audio on iPhone, more pictures shown for everyday questions, and history is no longer lost when multiple tabs or modes are open
v0.7.1
2026-05-28- Fixed homepage Realtime mode — migrated to the OpenAI Realtime GA API after the Beta endpoint was disabled
- New /how-it-works page explaining the image source (Pixabay + keyword safety filter), safety approach, and the differences between Realtime, Standard, and Sd-Voice modes
- Added a How it works link to the footer on every page
v0.7.0
2026-05-11- New /sd-voice page powered by xAI Grok Realtime Voice — voice-to-voice Q&A with the same flashcard headline + image experience
- Shared bottom navigation: every page now links to Realtime, Standard, Sd-Voice, History, Changelog, and About
v0.6.0
2026-04-07- Realtime mode is now the default at / — Standard mode moved to /standard
v0.5.5
2026-04-07- Realtime mode now shows flashcard-style headlines extracted after each answer
- New /api/extract endpoint provides headline and image keywords from Q&A context
- Shorter, punchier answers — prefer 1 sentence over 2-3
- Flashcard-style headlines that adapt to question type (Yes/No, numbers, nouns)
v0.5.3
2026-04-04- Added rate limiting (20 requests/min per IP) across all API routes to protect against abuse
v0.5.2
2026-03-29- Added clear visual distinction between active listening (pulsing green ring, animated sound bars) and paused mic states
- Improved error messages when the answer API fails — now shows the actual server error instead of a generic message
v0.5.1
2026-03-24- Fixed Realtime mode infinite loop on iOS Safari — mic is now muted while the AI speaks to prevent audio feedback triggering new responses
v0.5.0
2026-03-24- Switched answer model to gpt-5.4-nano via the Responses API for faster responses
- Added Realtime mode at /realtime — voice-to-voice via OpenAI Realtime API over WebRTC
- Switched image provider from Wikipedia to Pixabay with safe-search filtering
- Added E2E latency benchmark with model comparison support
- KidQ header now refreshes the page, removed image loading skeleton
v0.4.2
2026-03-24- Fixed voice input on iOS Safari — follow-up questions now work without refreshing the page
v0.4.0
2026-03-22- New black-and-white favicon, cleaner history page layout
- Added Browsee session-replay analytics
v0.3.0
2026-03-22- Center-aligned answer card with larger headline, mobile-optimized layout with always-visible footer
- Moved session history to dedicated /history page, parallelized Wikipedia image lookups for faster display
v0.2.0
2026-03-22- Added version display in footer with link to changelog page
v0.1.0
2026-03-22- Initial KidQ voice Q&A MVP — speak a question, get a spoken answer with text card and Wikipedia image