```yaml # Defaults: status=open, needs=autonomous, effort=M items: - id: demo-video-recording title: "CLI demo video recorder — VHS tape generation + recording from Python" meta: {tags: [feature, tooling], scope: [present/src/rivus_present/recorder/]} notes: "Use charmbracelet/vhs for scriptable terminal recording. Generate .tape files from Python, render to GIF/MP4. First use case: vario ng model_debate demo." - id: demo-tts-voiceover title: "TTS voiceover generator for demo videos — OpenAI tts-1-hd + edge-tts fallback" meta: {tags: [feature, tooling], scope: [present/src/rivus_present/recorder/], depends: [demo-video-recording]} notes: "Generate narration audio from text using OpenAI TTS API. Combine with VHS recordings via ffmpeg. edge-tts as free fallback for iteration." - id: system-flyover-demo title: "15-minute system flyover demo — script, HTML presentation, preflight checklist" meta: {tags: [feature, demo], needs: human, scope: [present/gallery/demos/], status: done} notes: "5-act narrative touring 8+ systems: Intel→SupplyChain→Draft→Vario→ngstudio→Ideas→Jobs→Doctor→Watch. Script, HTML speaker notes, readiness assessment, preflight checklist." - id: system-flyover-screenshots title: "Capture screenshots for all 10 demo stops" meta: {tags: [polish, demo], effort: S, needs: human, scope: [present/gallery/demos/flyover-screenshots/], depends: [system-flyover-demo]} notes: "Start 3 missing servers (supplychain, vario, ideas), then screenshot each stop. Use appctl screenshot." - id: illustrate-learnable-skills title: "Add illustrations to learnable-skills.html report via rivus_present.illustrator" meta: {tags: [polish], scope: [present/src/rivus_present/illustrator/], status: closed} - id: lyria-presentation-music title: "Build CLI tool to generate 30-60s presentation background tracks with Gemini Lyria" meta: {tags: [feature], scope: [present/]} - id: lyria3-api-availability title: "Track when Lyria 3 (with vocals) gets an API model ID — currently app-only" meta: {tags: [research], effort: S, needs: research, scope: [present/]} - id: vario-benchmark-frontier-models title: "Run vario strategies on frontier models (Opus 4.6, GPT-5.2-Pro, Gemini 3.1-Pro, Grok-4.1)" meta: {tags: [research, data], effort: L, scope: [vario/strategies/experiments/]} - id: vario-benchmark-all-strategies title: "Run all 19 vario strategies — current data covers ~7, run remaining 12" meta: {tags: [research, data], effort: L, scope: [vario/strategies/experiments/]} - id: vario-benchmark-sample-sizes title: "Scale to 100+ problems per strategy x model for statistical significance" meta: {tags: [research, data], effort: L, scope: [vario/strategies/experiments/]} - id: vario-benchmark-cross-domain title: "Run cross-domain benchmarks: MMLU-Pro, GPQA Diamond, WritingBench" meta: {tags: [research], effort: L, scope: [vario/strategies/experiments/]} - id: vario-strategy-tournaments title: "Run head-to-head strategy tournaments on identical problem sets with p-values" meta: {tags: [research], scope: [vario/strategies/experiments/], depends: [vario-benchmark-all-strategies]} - id: vario-cost-profiling title: "Build precise $/problem and $/quality-point-gained cost profiles for every strategy x model" meta: {tags: [research, data], scope: [vario/strategies/experiments/], depends: [vario-benchmark-all-strategies]} - id: vario-latency-profiling title: "Measure wall-clock latency distributions (p50, p95) for all strategy x model combos" meta: {tags: [research], scope: [vario/strategies/experiments/], depends: [vario-benchmark-all-strategies]} - id: vario-ablation-studies title: "Ablation studies: measure contribution of each stage (generation, verification, voting)" meta: {tags: [research], effort: L, scope: [vario/strategies/experiments/], depends: [vario-benchmark-all-strategies]} - id: vario-advantage-update-data title: "Update vario-advantage.html with frontier-model benchmark data" meta: {tags: [polish, data], scope: [present/gallery/rivus_explainer/pitch/], depends: [vario-benchmark-frontier-models, vario-benchmark-all-strategies]} - id: vario-ng-explainer title: "Build pitch/vario-ng.html — explainer for Vario NG CLI redesign" meta: {tags: [feature], scope: [present/gallery/rivus_explainer/pitch/]} - id: presentation-where-software-going title: "Write presentation on where software is headed" meta: {tags: [feature], effort: L, needs: human, scope: [present/]} - id: presentation-vc-pitch title: "Build VC pitch deck for a learning/self-improvement software company" meta: {tags: [feature], effort: XL, needs: human, scope: [present/]} - id: video-podcast-previewer title: "Build best-in-class video podcast previewer — visual timeline, key moments, topic segments, searchable transcripts" meta: {tags: [feature], effort: XL, needs: human, scope: [present/]} - id: principles-to-report-gym title: "Build gym for going from principles + raw data to polished report — extract-apply-score loop" meta: {tags: [research, feature], effort: L, needs: research, scope: [present/, kb/]} # Gallery Curation - id: gallery-curation-pass title: "Curate gallery/explainers/ — prune stale, update outdated, add status/owner metadata" meta: {tags: [polish], effort: S, scope: [present/gallery/]} notes: > 7 small explainers consolidated into gallery/explainers/ during restructure. Need a curation pass: prune stale ones, update outdated content, consider adding a manifest.yaml with status/owner per entry. # Servers - id: present-server title: "Build Gradio UI server for present/ — papers, diagrams, demos, slides" meta: {tags: [feature, infra], effort: L, scope: [present/]} ``` # Present — Reference Notes ## Lyria API Details - **Lyria RealTime** (`lyria-realtime-exp`) — works via Gemini API (`v1alpha`), instrumental only, real-time streaming via WebSocket - **Lyria 3** — launched 2026-02-18 in Gemini app (consumer), supports vocals+lyrics, no separate API model ID yet - **Lyria 2** (`lyria-002`) — available via Vertex AI (not Gemini API) - **Tested working**: `google-genai` SDK, `client.aio.live.music.connect()`, stereo 48kHz 16-bit PCM output - **API key**: `GEMINI_API_KEY` env var, requires `http_options={'api_version': 'v1alpha'}` - **Prompt steering**: `set_weighted_prompts([WeightedPrompt(text="...", weight=1.0)])` then `play()` + `receive()` loop ## Coding Principles Presentation Report: `present/gallery/coding_principles/report.html` Served at: `https://static.localhost/present/gallery/coding_principles/report.html` ```yaml items: - id: principles-enhance-examples title: "Add 2-3 more worked examples from different domains (batch, UI, testing)" meta: {tags: [content], effort: M, scope: [present/gallery/coding_principles/]} notes: "Pull from learning.db instances with highest evidence counts. Show breadth of application beyond vario." - id: principles-quick-ref title: "Add compact 1-page quick reference card — all 30 as printable table" meta: {tags: [content], effort: S, scope: [present/gallery/coding_principles/]} - id: principles-system-diagram title: "Add system diagram: learning.db → materialize → principles files → /recall → applied in context" meta: {tags: [content], effort: S, scope: [present/gallery/coding_principles/]} - id: principles-illustrate title: "Run through rivus_present.illustrator for visual enhancement" meta: {tags: [polish], effort: S, scope: [present/gallery/coding_principles/], depends: [principles-enhance-examples]} - id: principles-publish title: "Publish to static.localhost via inv static.publish" meta: {tags: [deploy], effort: S, depends: [principles-illustrate]} ``` ## Vario Benchmark Runners Runners: `vario/strategies/experiments/run_benchmark_v2.py`, `run_genverify_multimodel.py`, `run_math_strategies.py`. Results → `vario/strategies/data/experiments.db`. After data lands, regenerate SVG charts in the pitch page.