tutor-app
A 1:1 language-tutor companion that joins your Meet, scores pronunciation, and turns mistakes into kardo-style flashcards.
construit avec meetbot · open source
Huit sample apps fonctionnelles, sous licence MIT. Clonez, échangez vos API keys, livrez en production. Chacune est un vrai projet du vendredi après-midi — pas une capture d'écran marketing.
les huit apps
Choisissez le plus proche de ce que vous construisez. Chaque carte renvoie au repo, à un bouton de déploiement en un clic (le cas échéant) et à un bref doc sur son fonctionnement.
A 1:1 language-tutor companion that joins your Meet, scores pronunciation, and turns mistakes into kardo-style flashcards.
Captures every caption, translates with DeepL or Claude, and (when /chat ships) posts the translation back into the meeting.
A Gong/Chorus-style call coach: deal-stage tagging, talk-ratio, objection-handling scorecards, and rep-level dashboards — all webhook-driven.
Auto-summarizes daily standups into a per-person digest, posts to Slack, and threads blockers as Linear issues.
The simplest possible meetbot integration: webhook → Claude → action items → Postmark email to the organizer. Three languages, one pipeline.
Joins a meeting, listens for a wake-word, drafts a reply with OpenAI, speaks it back through ElevenLabs. Your first multimodal in-meeting agent.
Google Calendar OAuth + a recording-rules engine. Preview which upcoming meetings will be recorded before you flip it on for the org.
An embeddable web player for any meetbot manifest. Mixed and per-speaker tracks, clickable transcript, no proprietary container.
ce qui fait une bonne app meetbot
Nous avons vu suffisamment de samples dériver pour savoir ce qui fonctionne. Les MIT-licensed ci-dessous font tous ces quatre choses ; ceux qui ne le faisaient pas n'ont pas livré.
meetbot's signed `job.finalized` webhook is the contract. Polling /jobs/:id wastes a serverless invocation budget and races against your own retries — don't do it.
Verify the HMAC, drop the job into a queue, return 200 in under 100ms. The actual transcript-pull-and-summarize work belongs on a Trigger.dev / Inngest / SQS task, not in the webhook handler.
Every sample takes `OPENAI_API_KEY` or `ANTHROPIC_API_KEY` from `.env`, never from a SaaS console. Customers want to bring their own — and pick which model they're paying for.
Each sample has unit tests for the SDK call (HTTP mocked), the business-logic helpers, and a happy-path integration against a stub. If you can't `pnpm test` the sample in CI, it's a screenshot, not a starter.
ce que nous construirions si nous avions le temps
Nous ne pouvons pas tout construire. Voici huit apps meetbot non évidentes que nous adorerions voir — choisissez-en une, livrez-la, envoyez-nous le repo et nous la mettrons en avant ici.
Auto-drafted SOAP notes after every tele-therapy session. HIPAA territory — needs a BAA with whichever LLM provider, encrypted-at-rest manifest storage, and a hard "never train on this" promise.
Same shape but for primary care. ICD-10 code suggestions, draft visit summary the clinician edits in 60 seconds. ROI is the #1 reason every health-tech AI scribe exists.
Webhook → transcript → rubric-scored against your interview loop's competencies → ATS comment. Replaces frantic post-interview note re-typing during debrief.
Tone, sentiment, named-feature-complaints, missing-stakeholders flag. Ships a weekly "these accounts are 20% likelier to churn this quarter" email to the CS lead.
Auto-summarize the meeting, email both parents and the teacher in their preferred language, attach a follow-up checklist. Schools currently do this on paper.
One 60-min all-hands becomes ten 4-paragraph digests, each filtered for what matters to that team. Eng skips the marketing rev-share update; sales skips the infra migration.
Junior engineers practice system-design interviews with each other, get scored on rubric (clarification questions asked, trade-offs articulated, complexity analyzed). Could be the next Pramp.
Drop the bot into a Riverside or Squadcast post-production session, get back a YouTube-ready chapter list, episode description, and three pull-quote tweet drafts.
vous en avez construit une ?
PR votre URL de repo à github.com/meetbot-dev/awesome-meetbot avec un pitch d'une ligne, une capture d'écran et un bouton de déploiement. Nous reviewons chaque semaine. Le top 3 chaque trimestre obtient un crédit de 500 $ Stripe sur meetbot, un spotlight Twitter et un slot permanent en haut de cette page.