Versions

Versions

Pre-1.0 numbering: 0.MINOR.PATCH, one minor version per milestone-sized step. Full notes: CHANGELOG.md. Tags: v0.1.0, v0.2.0, v0.3.0, v0.4.0, v0.5.0, v0.5.1, v0.6.0, v0.7.0.

Version	Date	Summary
0.8.0 (unreleased)	2026-10-22	Built offline (robot test next): beamforming over the four head microphones against Pepper's fan noise (#34; offline 24.9 % to 19.5 % recognition errors), a steady camera stream that looks closer when the view changes, and glances towards voices from out of view (#33), keeping every camera frame of a session (RECORD_CAMERA, 2026-10-28), and listening to a microphone on the host, such as a clip-on on Pepper's chest, muted while Pepper speaks (LOCAL_MICROPHONE, 2026-10-28), and a map made with NAOqi's own exploration with named places kept across restarts and driving between them (#36, MAP, 2026-10-28), a Cohere second pass over each utterance (STT_SECOND_PASS, 2026-11-02), and recognising enrolled people by voice and face, with a second voice ending the gate's in-conversation bypass (#14, #25, VOICE_ID, SECOND_VOICE, FACE_ID, 2026-11-05); all off by default (the camera switches are on in the lab's .env from 2026-10-28)
Unreleased	2026-10-20	Fixes from the first robot session with 0.6.0 and 0.7.0: tablet subtitles, sound direction after a bridge restart, the addressee gate's crowd rule, no action from a held fragment, a longer end-of-turn pause, initiative follow-ups after 90 s, an open hand when pointing
0.7.0	2026-10-09	Built offline (robot test next): waiting for the rest of a sentence cut at a pause, shorter replies, pointing and looking at a spot in a photo, a camera fallback for "come to me", a bridge watchdog, and a first opt-in initiative loop (Pepper may speak unprompted)
0.6.0	2026-10-08	Memory, built offline (robot test next): one structure for working memory (Pepper's pose, people remembered out of view, voice direction, scene notes, a timeline and the recall tool) and long-term memory (people who agreed, facts, session episodes; consent checked in code, "forget me"). Bridge: odometry pose, sound localisation
0.5.1	2026-10-08	Fixes from the first robot session with system one: a handshake that holds and notices being taken, routed "look at me", the addressee gate in conversation and when alone, turning to face people, people reports that follow moves, camera thresholds, quiet model outages
0.5.0	2026-10-03	System one in the loop (built offline; robot test 2026-10-08): addressee gate, command router, camera judgements on a local GPU machine, session records and review script, setup guide for cloud-only and lab setups
0.4.0	2026-09-29	Greeting newcomers and head reflexes (tested on the robot); Sonnet 5.5; local models through Ollama; sharp photos, neutral pose after gestures and bridge autostart (built offline, robot check next session)
0.3.0	2026-09-23	First sessions on the robot, hardware fixes, model and speech-to-text choices, world model begins
0.2.0	2026-09-10	Reactive layer and voice input (first labelled v2.2)
0.1.0	2026-09-08	Rewrite of bridge and host (first labelled v2.1)