cd /news/artificial-intelligence/my-friend-s-group-chat-just-got-roas… · home › topics › artificial-intelligence › article
[ARTICLE · art-173538] src=github.com ↗ pub= topic=artificial-intelligence verified=true sentiment=↑ positive

My friend's group chat just got roasted

A developer published Kudos AI, a free demo that turns a group-chat JSON export into roasty friend awards and an AI-generated awards ceremony video that users can post back into the chat. The monorepo uses pnpm workspaces with a Next.js 15 web app, a BullMQ worker, and packages for chat ingestion and awards, backed by Postgres 16 and Redis 7 via docker-compose; subjective award judging runs through the Jev/TypeSafe API and ceremony rendering through Magic Hour with async webhooks. Users pick a gentle, medium or spicy roast level, and the demo accepts native Kudos JSON or auto-detected Telegram, Messenger and Discord exports, validating for size, structure and malicious content before normalizing messages and computing stats.

key takeaway

Kudos AI's demo pipeline validates and normalizes chat JSON, computes stats, sends subjective award judging to the Jev/TypeSafe API, and renders ceremony video asynchronously through Magic Hour webhooks, with Postgres 16 and Redis 7 run locally via docker-compose.

read3 min views7 publishedNov 20, 2026
My friend's group chat just got roasted
Image: Michielbdejong (auto-discovered)

Turn a group-chat JSON export into roasty friend awards and an AI-generated awards ceremony you can drop back into the chat.

Free demo. Product docs are in /docs. Local dev uses the slideshow ceremony unless Magic Hour credentials are set.

  1. User uploads a JSON chat file (nativeKudos format or auto-detected platform export).
  2. We validate for size/structure/malicious content, normalize messages, compute stats, and useJev for subjective award judging.
  3. User picks roast level (gentle / medium / spicy ) and optional avatars.
  4. Magic Hour renders a shareable ceremony video (async + webhooks).
  5. User gets a link and/or MP4 to post in the group.

Monorepo (pnpm workspaces):

Path Role
apps/web Next.js 15 app (UI + Route Handlers API)
apps/worker BullMQ worker (analyze + ceremony jobs)
packages/chat-json JSON ingest, adapters, validation
packages/awards Stats, deterministic awards, Jev orchestration
packages/shared Shared types/utilities

Local dependencies: Postgres 16 and Redis 7 via docker-compose.yml.

Environment: copy .env.example to .env at the repo root (the web app loads it from there).

Variable Default / notes
DATABASE_URL postgres://kudos:kudos@localhost:5432/kudos
REDIS_URL redis://localhost:6379
OBJECT_STORE local (default) orr2
OBJECT_STORE_DIR .data whenOBJECT_STORE=local
APP_ORIGIN Extra CSRF origins, comma-separated (the request host is always allowed)
R2_ACCOUNT_ID ,R2_ACCESS_KEY_ID ,R2_SECRET_ACCESS_KEY ,R2_BUCKET Required when OBJECT_STORE=r2
SESSION_SECRET Required for signed session cookies
TYPESAFE_API_KEY Jev / TypeSafe API (analyze)
MAGIC_HOUR_API_KEY Ceremony render
MAGIC_HOUR_WEBHOOK_SECRET Magic Hour webhook verification

Prerequisites: Node 22, pnpm 9, Docker (for Postgres + Redis).

pnpm install
cp .env.example .env   # fill secrets as needed
docker compose up -d
pnpm --filter @kudos/web db:push

Web (dev):

pnpm dev

Worker (dev, separate terminal):

pnpm --filter @kudos/worker dev

Schema is managed with Drizzle (apps/web/src/db/schema.ts). For local/dev, apply the current schema to Postgres:

pnpm --filter @kudos/web db:push

Uses DATABASE_URL from the root .env (see apps/web/drizzle.config.ts). Do not commit .env.

Production build for the Next.js app:

pnpm --filter @kudos/web build

From the repo root:

pnpm test          # vitest in all workspaces
pnpm typecheck     # tsc --noEmit in all workspaces
pnpm secrets:scan  # gitleaks (matches CI)

CI runs test, typecheck, and secrets:scan on push/PR (.github/workflows/ci.yml).

Doc Contents
01-product-brief.md What Kudos AI is and what the demo includes
02-user-flows.md Export guides, upload, awards, and sharing
03-awards-system.md Award catalog, roast levels, and how winners are chosen
04-data-and-privacy.md Consent, retention, and third parties
Area Choice
Ingestion .json upload only at demo MVP; adapters for Telegram / Messenger / Discord JSON
Jev TypeSafe direct POST https://api.typesafe.ai/v1/systemone (pinned model); gateway optional for dev
Ceremony Magic Hour async APIs ( text-to-video ,ai-talking-photo , TTS)
Pricing Free for demo
Roast dial Gentle / medium / spicy
Fairness Max 4 awards per person
/docs/                 Product guides
/apps/web/             Next.js UI + API routes
/apps/worker/          Background jobs (BullMQ)
/packages/chat-json/   Schema, validators, platform adapters
/packages/awards/      Stats + Jev orchestration
/packages/shared/      Shared types
/packages/storage/     Local disk and R2 object storage
/fixtures/             Sample & malicious JSON for tests

TBD.

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @kudos ai 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/my-friend-s-group-ch…] indexed:0 read:3min 2026-11-20 · —