Sprint 182: Speech Input and Memory Browser — merge feature/CODUX-182-speech-input-and-memory-browser into dev

[numnx]

🚀 Sprint Completion: Sprint 182 · Speech Input and Memory Browser

Automated sprint completion PR opened by Code UX.

📋 Summary

Sprint: Sprint 182 · Speech Input and Memory Browser Tasks: 10/10 completed

🧭 Planning

Planning provider	Codex CLI
Planning model	gpt-5.5

Metric	Value
Input tokens	144,793
Cached input tokens	1,462,016
Output tokens	12,729
Total tokens	1,619,538
Invocations	1
Tool calls	58
Active time	4m 36s

Included usage estimate (subscription/local login): $1.84

🎯 Original Sprint Prompt

Okay then lets implement both options onnx runtime and external api - we will use the user system microfone for speech input and an amazing designed speech button in the chat uis (3d chat and threads) make sure to perfectly define the plan to include electron with support for macos win and linux installation aswell as thru the npm app. We will also need to enhance and refactor the model browser on the memories page. and group TTS and embedding models and make everything easier navigable and a little cleaner and smaller. this should also support providing custom links from huggingface.

✅ Task Checklist

• T01: Add speech settings contracts — codex (PR)
• T02: Implement transcription backend — codex (PR)
• T03: Build shared speech input controls — codex (PR)
• T04: Integrate speech into chat threads — codex (PR)
• T05: Support custom Hugging Face memory models — codex (PR)
• T06: Refactor memory model browser — codex (PR)
• T07: Enable desktop microphone packaging — codex (PR)
• T08: Fix speech request scoping and local ONNX audio decoding — codex (PR)
• T09: Preserve structured speech API errors in the dashboard client — codex (PR)
• T10: Allow audio-only capture in Electron permission checks — codex (PR)

👥 Provider Breakdown

10 by codex

⏱️ Sprint Timing

Started	2026-07-07 13:05:22 UTC
Finished	2026-07-07 14:56:44 UTC
Duration	1h 51m 22s

📊 Aggregate CLI Token Usage

Metric	Value
Input tokens	2,669,229
Cached input tokens	40,175,744
Output tokens	296,797
Total tokens	43,141,770
Invocations	18
Tool calls	960
Active time	2h 0m 1s

Included usage estimate (subscription/local login): $29.98

🕵️ QA Review Summary

Outcome: pass

Static review of the merged sprint found the speech transcription flow, Electron microphone policy, and memory model browser integrated coherently with the expected settings, route, and UI contracts. I could not run Vitest in this workspace because pnpm exec vitest was unavailable locally, so the verdict is based on source inspection and the committed test coverage.

🌿 Branch Info

Base: dev
Head: feature/CODUX-182-speech-input-and-memory-browser

---

🤖 Generated by Code UX