Repository navigation
Sprint Roadmap
igor-c-assuncao edited this page Jun 20, 2026
·
2 revisions
| Sprint | Theme | Status |
|---|---|---|
| Sprint 0 | Foundation | ✅ Complete |
| Sprint 1 | Audio capture | ✅ Complete |
| Sprint 2 | VAD + transcription | ✅ Complete |
| Sprint 3 | AI pipeline | ✅ Complete |
| Sprint 4 | BYOK + settings | ✅ Complete |
| Sprint 5 | UI polish | ✅ Complete |
| Sprint 6 | Window UX | ✅ Complete |
| Sprint 7 | Testing, quality gates, packaged sidecar validation | ✅ Active |
| Sprint 8 | v1.0 release hardening | 🚧 In progress |
| Card | Title | Status |
|---|---|---|
| 7.1 | [Core] Enumerate system audio devices | ✅ |
| 7.2 | [AI/Core] VAD telemetry bridge — RMS in real time | ✅ |
| 7.2b | [UI] Reactive pulse animation on record button | ✅ |
| 7.3 | [UI] Behavior toggles in settings popover | ✅ |
| 7.4 | [UI] Language + Model dropdowns in popover | ✅ |
| 7.5 | [UI] Preferences & Shortcuts buttons in popover footer | ✅ |
| 7.6 | [AI] Dynamic system prompt injection | ✅ |
| Card | Title | Status |
|---|---|---|
| 8.1 | [AI] Speaker diarization in the Python pipeline | ✅ |
| 8.2 | [UI] Transcript with timestamps & per-speaker avatars | ✅ |
| 8.3 | [UI] Participant avatars & tags in expanded header | ✅ |
| 8.4 | [AI/UI] Structured summary — TL;DR, Decisions & Action Items | ✅ |
| 8.5 | [Core] SQLite schema migration & updated Meeting model | ✅ |
| 8.6 | [Core] FTS5 full-text search on meeting history | ✅ |
| Area | Status | Notes |
|---|---|---|
| Windows sidecar packaging | ✅ Active | CPU-only PyInstaller sidecar validates at 405.1 MiB against a 1536 MiB budget |
| Sidecar supervisor | ✅ Active | Rust restarts the Python sidecar with backoff and exposes UI failure/reconnect states |
| Secret handling | ✅ Active | API keys and HF tokens are stored in keychain; Rust logs redact secret payload fields |
| Structured summaries | ✅ Active | Schema v2, evidence-backed claims, and useful fallback if local LLM JSON is invalid |
| Benchmark harness | ✅ Active | Example, QMSum smoke, and AMI ASR reports are generated under benchmarks/reports/
|
| Deployment guardrails | ✅ Active | Latest report is APPROVED for lint/type/prompt/RAG/compliance/MLOps checks |
| Installer signing/distribution | 🚧 In progress | Cross-platform release validation and signing remain release work |
| Representative quality baseline | 🚧 In progress | Larger AMI, CORAA, internal consented fixtures, and LLM-generated QMSum baselines still needed |
| Dataset/run | Result |
|---|---|
| Deterministic example fixture | PASS, weighted score 1.000
|
| QMSum test split, 2 fixtures, mock summary gate | PASS, weighted score 1.000
|
AMI ASR IB4001, Ollama llama3
|
PASS, weighted score 0.924, WER 0.258, CER 0.200, latency 68.428s
|
The example and QMSum mock rows validate scorer/report correctness. The AMI row is a real single-fixture ASR baseline and should not be treated as a production-wide quality score.