User repository on GitHub
Voice-driven semantic auto-advance for Slidev, powered by Cloudflare Agents and TypeSafe AI Jev
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Repos & SDKs
SlidePilot
Experimental Slidev addon that streams presenter voice through a Cloudflare Worker, where Jev answers four typed questions before TypeScript applies the navigation policy and advances the deck.
- Stars
- 8
- Language
- TypeScript
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- harshil1712
- Last verified
- Sep 2026
About this project
Your voice is the clicker. SlidePilot listens to a presenter, asks TypeSafe AI's Jev whether the current slide is semantically complete, and advances a Slidev deck only when deterministic safety checks pass. The Slidev addon streams microphone audio to a Cloudflare Agent over WebSockets. Workers AI Flux produces interim text and completed utterances.
Quoted from the project's README. Read the full README.
Works forJev for Document processing
Used inJev for Software & DevTools
Highlights
- Jev evaluates four typed questions per slide: natural advancement, completion of the core idea, continued speaking, and natural boundary.
- TypeScript, not the model, applies the navigation policy, with cooldown and stale-decision guards before advancing.
- Without a TypeSafe API key the Worker uses a deterministic mock policy; transcripts and slide content are sent to Jev only when a key is set.
- Manual keyboard and clicker navigation always works, and a pending advance is cancelled if the presenter resumes speaking.
- The npm addon package is prepared but not yet published, so the included demo and Worker must be run from a clone.
Quickstart
bash
pnpm add slidev-addon-slidepilotWatch out
MIT-licensed. Needs Node.js 20+, pnpm, a Cloudflare account with Workers AI access and an optional TypeSafe API key; no application-level authentication ships, and the npm addon package is not yet published.
Questions about this entry
Short answers drawn from the linked project or page.
The model decides. Workers AI Flux transcribes, then Jev answers four typed questions per utterance: whether advancing feels natural, whether the slide's core idea is complete, whether the presenter is still speaking, and whether this is a natural boundary. TypeScript applies the policy, and an explicit 'next slide' is never required.
Continued speaking is a hard veto rather than a tiebreaker: if the presenter is still explaining above stillExplainingCeiling (0.55) the pending advance is cancelled, and a 1000 ms cancellable advanceDelayMs lets a resumed sentence call it off. A 2000 ms cooldown and stale-decision guards sit in TypeScript, not in the model.
Without a key the Worker falls back to a deterministic mock policy, which validates the pipeline and the policy but not Jev's behavior. The addon package is prepared for npm and not yet published, so the demo and Worker must be run from a clone, along with Node.js 20+, pnpm and a Cloudflare account with Workers AI access.
Audio is processed by Workers AI with Deepgram Flux. When a key is configured, completed transcripts plus the current slide's visible content and speaker notes go to Jev, and the key stays in the Worker. There is no application-level authentication, so a public deployment is reachable by other clients and can incur Workers AI and Jev usage.
Whether advancing feels natural, whether the slide's core idea is complete, whether the presenter is continuing, and whether this is a natural slide boundary. TypeScript, not the model, applies the navigation policy from those answers.
advanceThreshold 0.68, completeThreshold 0.65, stillExplainingCeiling 0.55, confidence 0.30, cooldownMs 2000 and advanceDelayMs 1000. The README calls these starting values that should be calibrated with rehearsal transcripts.
The Worker exposes a POST /api/evaluate endpoint: send the slide's page, title, content, notes and thresholds plus a list of utterances and it returns the policy decision. GET /health reports whether the provider is jev or the deterministic mock.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Speech to live slides
X: Jev for real-time speech
X: On-device audio, no LLM in the loop
X: A voice assistant with no wake word
More Jev resources like this
LangChain.js is the TypeScript edition of LangChain, a framework for LLM-powered applications that runs on Node.js, Cloudflare Workers, Vercel and Next.js, Supabase Edge Functions, browsers, Deno, and Bun.
Jev Repos & SDKs#community#catalog-import#cobanov
Community
langchain-ai/langchainjsTypeScript
langchain-ai/langchainjsTypeScriptText-only classifier for 261 federal tax forms that sends each PDF page's text to Jev, which returns a probability over forms and page kinds behind a confidence gate; no model is trained.
Jev Repos & SDKs#community#typescript#classification
Experimental Hono router that matches an incoming HTTP request to a plain-language route description: one yes/no Jev question per description is evaluated in a single call, and the first registered route above the threshold wins.
Jev Repos & SDKs#community#typescript#routing
Community
yusukebe/hono-jev-routerTypeScript
yusukebe/hono-jev-routerTypeScriptLangChain is the agent engineering platform: an MIT-licensed Python framework from the LangChain team for building agents and LLM applications by chaining interoperable components with third-party model, tool, and vector-store integrations.
Jev Repos & SDKs#community#catalog-import#cobanov
Communitylangchain-ai/langchainPython
langchain-ai/langchainPythonComposio is an MIT-licensed SDK monorepo giving AI agents 1,000+ pre-authenticated toolkits with per-user sessions, authentication, triggers, and a sandbox, and it ships TypeScript and Python adapters including one for TypeSafe (Jev).
Jev Repos & SDKs#community#catalog-import#cobanov
Community
composiohq/composioTypeScript
composiohq/composioTypeScriptThe AI Toolkit for TypeScript: Vercel's free, open-source AI SDK for building AI-powered applications and agents, with GitHub topics covering OpenAI, Anthropic, Gemini, React, Svelte, Vue, and Next.js.
Jev Repos & SDKs#community#catalog-import#logicrw
Community
vercel/aiTypeScript
vercel/aiTypeScriptFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.