project-manager

🇬🇧 EN | 🇨🇳 中文

A lightweight, local-first requirement management and output tracking system. SQLite as the sole data source, AI agent handles all data operations, static HTML provides daily visualization.

Read about the design philosophy behind this project: What is an AI-Native Data System?

Why use this?

• Zero infrastructure: Single SQLite file, no server, no signup, no cloud
• Agent-native: Designed to be operated by AI agents (like opencode)—just talk to it
• Unified workflow: Manage requirements, schedules, deliveries, and cover outputs in one place
• Beautiful dashboards: Auto-generated HTML with calendar view, task tracking, and interactive ECharts statistics
• Data sovereignty: Everything stays on your machine

Talk to an agent for CRUD and scheduling; open the HTML dashboards yourself.

Pages

Four auto-generated HTML pages (under your data dir’s html/, or demo/html/ for the sample set):

Calendar

Monthly schedule view with color-coded owners and holiday markers.

Dashboard

KPI metrics, monthly stats, top requesters, type breakdown, fiscal year comparisons (powered by ECharts).

Recent Tasks

In-progress tasks and recently completed items with thumbnails and status.

History

Full archive of completed requirements with screenshot card grid.

Customization

The dashboard is built around a specific set of requirement types (UI design, data analysis, course production, internal efficiency) and a fiscal year starting in April. If your workflow differs, ask the agent to adapt render_html.py, render_queries.py, render_components.py, and schema.sql—cover value formula, KPI metrics, chart labels, and type colors.

Configuration

Constant	Default	Purpose
COVER_VALUE_MULTIPLIER	20	Cover image value multiplier (cover_count × multiplier)
FY_START_MONTH	4	Fiscal year start month
FY_END_MONTH	3	Fiscal year end month

Defined in scripts/config.py. Confirm calendar dates yourself before schedule changes.

---

Created by Victor42 & Agent Vik

---

For Agent

Stand up this local SQLite project-manager, wire the skill, then operate requirements/schedules/HTML via CLI or skill.

Prerequisites

• Python 3
• pip install Pillow
• Write access to a data directory (default: repo demo/)

Install / first-time setup

1. Clone and enter the repo; install Pillow.
2. Sample data: run the commands below, then open demo/html/dashboard.html to verify.

python3 scripts/init.py
python3 scripts/pm.py render-html

3. Real data: copy .env.example → .env, set PM_DATA_DIR, re-run init/render. Read .env before every later operation.
4. Install skill: copy skills/project-manager.md and/or skills/SKILL.md into the agent skills directory; fix paths to this clone.

Usage

Entrypoint: python3 scripts/pm.py COMMAND ... (optional --db-path).

Commands:

• doctor — DB / model consistency
• compute-periods — recompute stat period dates
• render-html — regenerate dashboard/calendar HTML
• stats — local stats
• requirement create / deliver / insert / thumbnail — requirement writes
• schedule add / adjust / move — schedule mutations
• holiday ... — public holiday / personal leave

Also follow the installed skill for conversational CRUD. Before any schedule mutation, confirm “today” with a real clock/date command (skill red flag).

After data changes that should appear on dashboards: python3 scripts/pm.py render-html.

Hand off to the human

• Business meaning of projects/owners/types
• Viewing HTML in a browser
• Dashboard taxonomy customization requests

Red lines

• Do not point PM_DATA_DIR at the wrong path or overwrite production pm.db without confirmation
• Do not invent schedule dates from chat memory
• Do not cp pm.db into ad-hoc .bak snapshots. backup.sql re-exports after every pm.py command; snapshot rotation rules are in skills/project-manager.md
• Schema/ops: notes.md and skills/