Drop a transcript. Get a summary, action items, and Slack push — all powered by AI.

---

What's this?

Ever had a meeting recording but no time to go through the whole thing? Actionify takes a .txt meeting transcript and runs it through a 5-stage AI pipeline to give you:

• Speaker-labeled transcript — who said what
• TL;DR + key points + decisions — one-line summary at a glance
• Action items — who needs to do what by when, in a neat table
• Slack push — all of the above posted straight to your channel

All in a few seconds.

---

How it works

You drop a .txt file
        │
        ▼
  ┌─ Extract Text ──┐   ← reads your file from disk
  └────────┬────────┘
           ▼
  ┌── Diarize ──────┐   ← AI labels speakers (Speaker A, Speaker B...)
  └────────┬────────┘
           ▼
  ┌── Summarize ────┐   ← AI writes TL;DR, key points, decisions
  └────────┬────────┘
           ▼
  ┌ Extract Actions ┐   ← AI pulls out assignee/task/deadline
  └────────┬────────┘
           ▼
  ┌── Push to Slack ┐   ← formatted Block Kit message
  └─────────────────┘

Built with LangGraph — each stage is a node in a state graph that passes data to the next. If you don't set up Slack, it just skips that step.

---

Tech

What	What we used
Backend	NestJS 11, TypeScript
Frontend	React 19, Vite 8, Tailwind CSS 4
AI pipeline	LangChain.js, LangGraph 1.3
LLM	Google Gemini 2.0 Flash or local LM Studio — your call
Validation	Zod for AI output, class-validator for API
Slack	@slack/web-api with Block Kit
Uploads	Multer 2 (.txt only, 10MB max)
State/HTTP	TanStack React Query 5, Axios

---

Project layout

├── backend/
│   ├── src/
│   │   ├── langgraph/
│   │   │   ├── state.ts              ← what flows through the pipeline
│   │   │   ├── meeting.graph.ts      ← stitches the 5 nodes together
│   │   │   └── nodes/
│   │   │       ├── extract-text.node.ts
│   │   │       ├── diarize.node.ts
│   │   │       ├── summarize.node.ts
│   │   │       ├── extract-actions.node.ts
│   │   │       └── slack-push.node.ts
│   │   ├── modules/
│   │   │   ├── upload/               ← file upload endpoint
│   │   │   ├── meeting/              ← process endpoint + orchestrator
│   │   │   ├── slack/                ← Slack message builder
│   │   │   └── llm/                  ← picks Gemini or LM Studio
│   │   ├── app.module.ts
│   │   └── main.ts
│   └── .env.example
│
├── frontend/
│   ├── src/
│   │   ├── components/
│   │   │   ├── UploadForm.tsx        ← drag & drop area
│   │   │   ├── SummaryCard.tsx       ← shows TL;DR, points, decisions
│   │   │   └── ActionItemList.tsx    ← table with checkboxes
│   │   ├── api.ts                    ← talks to the backend
│   │   ├── App.tsx
│   │   └── main.tsx
│   └── vite.config.ts
│
└── sample-transcript.txt

---

Get it running

You'll need Node.js 18+ and optionally a Slack bot token + Gemini API key.

# backend
cd backend
npm install
cp .env.example .env        # then edit .env with your keys
npm run start:dev

# frontend (new terminal)
cd frontend
npm install
npm run dev

Open http://localhost:5173, drop a .txt file, and you're off.

---

Config

Variable	Default	What it does
LLM_PROVIDER	lm-studio	lm-studio or gemini
LM_STUDIO_URL	http://localhost:1234/v1	your local LLM endpoint
GEMINI_API_KEY	—	key for Gemini
SLACK_BOT_TOKEN	—	your Slack bot token
SLACK_CHANNEL	#meeting-notes	where to post results
PORT	3000	backend port

API

POST /upload — upload a .txt file (field name: file)

POST /meeting/process

{ "filePath": "/path/to/file.txt", "slackChannel": "#meeting-notes" }

Returns the full pipeline output: raw + diarized transcript, summary, and action items.

---

Running tests

cd backend
npm run test        # unit tests
npm run test:e2e    # e2e tests

---

Made with ☕ and LangGraph