WorkLog AI

AI-powered daily worklog manager. Type what you worked on in natural language, and AI extracts tasks, decisions, meetings, and notes into structured, searchable markdown files.

Disclaimer: This project is intended for personal use only and should be hosted locally. It is not production-grade software. The entire codebase was vibe coded with Claude by Anthropic.

Features

• Natural language input — describe your day and AI structures it automatically
• Screenshot processing — upload or paste (Ctrl+V) screenshots for AI text extraction
• Daily markdown files — human-readable, one file per day, works outside the app
• Semantic search — find decisions, tasks, and notes by meaning, not just keywords
• Dark / light theme — toggle between themes, respects system preference
• Calendar navigation — browse any day's worklog or log entries to a future date
• Inline editing — click the pencil icon on any entry to edit it in place
• Meeting details — attendees and notes shown as sub-items under each meeting
• Task time tracking — tasks can carry a time (e.g. 15:00) shown in the UI
• Starred highlights — mark entries as important with the star toggle, shown in sidebar
• Work summary report — select a date range to view aggregated stats, highlights, decisions, meetings, and pending tasks with optional AI narrative summary and HTML download

• AI usage metrics — sidebar tracks daily AI request counts and estimated tokens (persisted across restarts)
• Processing indicators — animated progress bar and success summary on submission
• Dashboard — today's tasks, meetings, pending items, overdue items at a glance
• Multiple AI providers — Ollama (local, default), OpenAI, or Claude
• User authentication — JWT-based login and registration with bcrypt password hashing
• Fully containerized — one docker compose up to start everything

Quick Start

Prerequisites

• Docker and Docker Compose

Run

git clone <repo-url> worklog-ai
cd worklog-ai
docker compose up --build

Open http://localhost:3000 in your browser.

Authentication

On first launch you'll see a registration page — create a username and password to get started. Subsequent visits show a login screen. Credentials are stored locally in the data/ directory (bcrypt-hashed passwords, auto-generated JWT secret). Tokens expire after 7 days.

Ollama setup

By default the backend connects to Ollama on the host machine at http://host.docker.internal:11434. Install Ollama natively and pull the required models:

ollama pull llama3.2          # text processing
ollama pull llava             # screenshot/image OCR
ollama pull nomic-embed-text  # embeddings for semantic search

To run Ollama inside Docker instead, start with the with-ollama profile:

docker compose --profile with-ollama up --build

A setup banner on the dashboard shows model download progress.

Development

Backend:

cd backend
npm install
npm run dev        # starts on port 4000
npm test           # run tests

Frontend:

cd frontend
npm install
npm run dev        # starts on port 5173, proxies /api to :4000
npm test           # run tests

Architecture

┌──────────┐     ┌──────────┐     ┌──────────┐
│ Frontend │────▶│ Backend  │────▶│  Ollama  │
│ React    │     │ Express  │     │  LLM     │
│ :3000    │     │ :4000    │     │ :11434   │
└──────────┘     └────┬─────┘     └──────────┘
                      │
                      ├──▶ Markdown files (./data/)
                      │
                 ┌────▼─────┐
                 │ ChromaDB │
                 │ Vectors  │
                 │ :8000    │
                 └──────────┘

Data flow:

1. User types free text or uploads screenshots
2. Backend sends to AI for structured extraction
3. Structured data is saved to daily markdown file and indexed in ChromaDB
4. Dashboard shows today's entries; search queries ChromaDB for semantic matches

Configuration

Use the Settings page in the UI, or create a config.json in the project root:

Setting	Default	Description
dataPath	./data	Where markdown files are stored
aiProvider	ollama	AI provider: ollama, openai, or claude
ollamaUrl	http://host.docker.internal:11434	Ollama server URL
openaiApiKey	(empty)	OpenAI API key (required if using OpenAI)
claudeApiKey	(empty)	Claude API key (required if using Claude)
chromaUrl	http://chromadb:8000	ChromaDB URL

Environment variables OLLAMA_URL, CHROMA_URL, and DATA_PATH override the corresponding config values (set automatically in Docker).

Worklog Format

Each day produces a markdown file (YYYY-MM-DD.md):

# Worklog - 2026-04-18

## Tasks
- [x] 15:00 Reviewed PR #142.
- [ ] Update API documentation. (due: 2026-04-21)

## Decisions
- Migration timeline pushed to Q3

## Meetings
- 10:00 Daily standup.
  - Attendees: Alice, Bob, Charlie
  - Discussed sprint blockers on payment API.
  - Agreed to prioritise hotfix for checkout latency.

## Notes
- Check Grafana dashboard for latency spikes

## Highlights
- Migration timeline pushed to Q3

## Tags
#sprint-12 #migration #api

Docker Images

Pre-built images are available on Docker Hub:

Image	Description
shri32msi/worklog-ai-backend	Node.js/Express API server
shri32msi/worklog-ai-frontend	React frontend served via Nginx

Pull and run directly:

docker pull shri32msi/worklog-ai-backend:latest
docker pull shri32msi/worklog-ai-frontend:latest

Or use docker compose up --build to build from source.

Tech Stack

• Frontend: React 19, Vite 8, Tailwind CSS 4, TypeScript
• Backend: Node.js 22, Express, TypeScript
• AI: Ollama (llama3.2, llava, nomic-embed-text), OpenAI (gpt-4o), Claude (sonnet)
• Search: ChromaDB with vector embeddings
• Infrastructure: Docker Compose

License

This project is licensed under the MIT License — see the LICENSE file for details.

---

Made with Claude by Anthropic.