Daymark

Tagline: Make the next step obvious.
Amazon Developer Hackathon Submission
Primary Track: Alexa+
Mini-Challenges: AWS Builder & Open Source

1. What is Daymark?

Students and early-career professionals face continuous task thrashing, looming exams, and competing deadlines. Existing productivity systems demand constant manual organization, while general-purpose AI chat assistants offer isolated suggestions disconnected from real, persistent task databases.

Daymark is an AI-powered personal productivity assistant that bridges natural language and deterministic execution:

1. Understands natural requests: e.g., "I have three hours tomorrow. Plan my tasks." or "Add revise calculus for 45 minutes."

2. Executes real mutations: Persists tasks in SQLite with zero hallucinated data or code execution risks.

3. Generates realistic schedules: Enforces strict time budgets, inserts 10-minute cognitive rest buffers, and flags overflow items that do not fit.

4. Exposes genuine MCP tools: Includes a Streamable HTTP Model Context Protocol (MCP) server so external autonomous agents can interface directly with Daymark.

5. Simulates Alexa+: Designed from the ground up to showcase an agentic voice-and-screen productivity flow for the Alexa+ era.

Honesty & Compliance Notice: This project uses the verified simulated Alexa+ experience route coupled with a real Streamable HTTP MCP server to demonstrate the integration path. It does not falsely claim an unreleased official Alexa+ production deployment.

Related MCP server: project-tasks-mcp

2. Quick Start (No Docker Required)

Daymark is designed for immediate local execution on macOS or Linux without Docker or paid infrastructure.

Prerequisites

• Node.js 18+ (tested on v22)

• Python 3.10+ (for optional Python CLI / MCP smoke tests)

1. Install & Run

# Clone the repository
git clone https://github.com/your-username/daymark.git
cd daymark

# Launch development server (starts on http://localhost:3000)
./scripts/run-dev.sh

Or run via npm:

npm install
npm run dev

2. Open the Dashboard

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

• Overview: Executive dark dashboard with real metrics and schedule timeline.

• My Tasks: Filter and sort tasks with persistent SQLite updates.

• Daily Plan: Configurable time budget, non-overlapping blocks, and overflow detection.

• Alexa+ Assistant: Conversational natural-language interface with real tool dispatches.

• MCP Inspector: Live JSON-RPC 2.0 tool runner and protocol validator.

3. Running Acceptance Tests

Run the automated test suite covering CRUD persistence, scheduler math, natural language intents, and MCP tools:

npm run test

Verified Test Results:

--- DAYMARK ACCEPTANCE TEST SUITE ---
[Suite 1: Task Management & Validations]
  ✓ Create valid task
  ✓ Reject invalid duration (< 5 mins)
  ✓ Reject invalid duration (> 480 mins)
  ✓ Reject empty title
[Suite 2: Task Mutations & Error Handling]
  ✓ Setup test task
  ✓ Update task attributes
  ✓ Complete task
  ✓ Delete task
  ✓ Throw error on unknown task ID
[Suite 3: Deterministic Scheduling Engine]
  ✓ Schedule strictly within fixed budget without exceeding available time
  ✓ Detect overflow tasks that do not fit
  ✓ Ensure no overlapping blocks & proper chronological sequence
[Suite 4: Natural Language AI Engine]
  ✓ Parse "Add revise calculus for 45 minutes"
  ✓ Parse "I have three hours tomorrow. Plan my tasks."
  ✓ Bedrock mode fallback disclosure when unconfigured
[Suite 5: Model Context Protocol (MCP) Server]
  ✓ Exposes all 8 required MCP tools
  ✓ MCP execute create_task tool
  ✓ MCP execute list_tasks tool
  ✓ MCP execute generate_plan tool
  ✓ MCP reject invalid arguments gracefully

TEST RESULTS: 20 Passed, 0 Failed

4. Real MCP Server (Alexa+ Track)

Daymark implements a self-hosted Model Context Protocol (MCP) server adhering to the current Streamable HTTP (JSON-RPC 2.0) standard.

• Endpoint: http://localhost:3000/mcp

• Supported Methods: initialize, notifications/initialized, tools/list, tools/call, ping

Exposed Tools:

Run MCP Smoke Test

python3 scripts/smoke-test-mcp.py

5. Amazon Bedrock (AWS Builder Mini-Challenge)

Daymark supports a dual-mode AI engine:

1. Mode A — Local Demo Mode (Default): Zero network calls, zero API keys, 100% deterministic intent parsing and planning.

2. Mode B — Amazon Bedrock Mode: Integrates with Amazon Bedrock via AWS credentials and configurable model ID (e.g. anthropic.claude-3-haiku-20240307-v1:0 or amazon.titan-text-express-v1).

Configuring Bedrock

In .env:

AI_PROVIDER=bedrock
AWS_REGION=us-east-1
AWS_ACCESS_KEY_ID=your_access_key
AWS_SECRET_ACCESS_KEY=your_secret_key
BEDROCK_MODEL_ID=anthropic.claude-3-haiku-20240307-v1:0

Graceful Fallback Guarantee

If AWS credentials are unset or the Bedrock call is unreachable, Daymark automatically falls back to Local Mode and explicitly discloses the fallback in the response:

(AWS Bedrock credentials not configured or unreachable; smoothly executed via deterministic Local Engine).

6. Open Source Contribution

This project is licensed under the permissive MIT License.

Open Source Contribution during Hackathon Window

As part of our commitment to the open-source developer ecosystem:

• Modular MCP Planning Engine: The deterministic scheduling logic in src/server/scheduler.ts and backend/app/services/scheduler.py is architected as an independently importable, zero-dependency utility that any MCP server or AI agent framework can adopt.

• Specification-Compliant Streamable HTTP MCP Implementation: Provides an open reference implementation for developers building web-hosted MCP servers compliant with the November 2024 specification.

7. 3-Minute Demo Video Flow

Refer to docs/DEMO_SCRIPT.md for the complete timestamped presentation guide:

• 0:00–0:20: Problem & Dashboard Introduction

• 0:20–0:50: Natural-Language Task Creation (Add revise calculus for 45 minutes)

• 0:50–1:25: Deterministic Planning (I have 3 hours tomorrow. Plan my tasks.)

• 1:25–1:55: Real State Mutation & Audit Trail

• 1:55–2:25: Live MCP JSON-RPC 2.0 Server Demonstration

• 2:25–2:45: AWS Bedrock Integration & Transparent Fallback

• 2:45–3:00: Architecture & Open-Source Closing

8. Acceptance Criteria Checklist

• 1. The dashboard runs locally without Docker (./scripts/run-dev.sh).

• 2. Tasks persist in local SQLite store across restarts.

• 3. Natural-language requests perform actual supported actions.

• 4. Scheduling uses real task data and respects time constraints.

• 5. The activity feed reflects real application operations.

• 6. A genuine Streamable HTTP MCP server exposes 8 working tools.

• 7. MCP tools and the web application use the same data store.

• 8. Optional Bedrock integration is implemented and fails gracefully.

• 9. Automated tests are included and report 20/20 passing.

• 10. The README enables another developer to reproduce the demo.

• 11. The project has an MIT license and no committed secrets.

• 12. The demo can be recorded in under three minutes.