Project Health (ph) 🚀

The unified CLI & IDE tool for deep codebase intelligence

project-healthy (ph) is an advanced, all-in-one CLI binary and IDE extension designed to run deep, automated analyses on any software repository. It combines eight parallel analysis modules and an AI intelligence layer powered by MegaLLM to provide actionable insights, unified health scores, and conversational code interaction.

📖 Executive Summary

Developers often juggle between multiple tools to evaluate codebase health—CI dashboards, linters, security scanners, documentation wikis, and PR reviewers. ph solves this fragmentation by offering one command, one score, and one conversation.

Running ph scan orchestrates eight analysis modules in parallel to generate a comprehensive 0-100 codebase health score, severity-ranked findings, and an actionable top-3 task list. An embedded AI layer powers conversational code interrogation, PR reviews, onboarding document generation, and intelligent documentation updates.

🎯 Primary Goals

• Single Binary: Fast, global installation (npm install -g project-healthy).
• Parallel Analysis: Eight distinct modules run concurrently (under 60s for typical projects).
• Unified Scoring: 0–100 weighted health score with actionable insights.
• MegaLLM-powered AI: High-quality LLM integrations routed through a hosted backend (supporting Claude, GPT, Gemini, etc.).
• Zero-Config AI: Install from npm and start using AI features immediately — no API keys, no authentication, no setup.
• IDE & CI/CD Support: VS Code extension included, alongside CI pipeline integration (e.g., block merges if score drops below a threshold).

🏗 Architecture Overview

1. Analysis Pipeline

The execution pipeline seamlessly integrates multiple insights: User runs: ph scan → Analysis Orchestrator (Runs M01 to M08 concurrently) → Score Aggregator (Computes HealthReport) → Output Layer (Displays Terminal | JSON | HTML)

2. Local Context Cache (.ph-cache/)

The tool leverages a local workspace cache located at the project root to support AI grounding and temporal state checks:

• ast-index.json: Maps symbol names to locations.
• docs-index.json: Maps source files to their referencing doc sections.
• last-scan.json: Context injector for AI queries.
• sessions/: Chat history files for persistent context.

3. AI Backend Architecture

AI calls are routed through a hosted backend at https://project-healthy.vercel.app/v1. The backend interacts with the MegaLLM API using secured keys on the server side. No third-party LLM API keys run on the developer's local machine. The CLI connects directly to the hosted backend with no authentication required.

CLI (ph) → https://project-healthy.vercel.app/v1 → MegaLLM API
          (hosted, public, no auth)               (server-side API key)

🧩 Core Analysis Modules (M-01 to M-08)

ph calculates your health score using eight diverse, weighted modules.

| ID | Module Name | Weight | Primary Purpose | How it works | | :------- | :---------------------- | :----: | :------------------------------------------ | :---------------------------------------------------------------------------- | | M-05 | Dependency Security | 20% | Detect CVEs & license risks. | Analyzes npm audit/pip-audit and license compatibility. | | M-02 | Code Quality | 18% | Linting, complexity, duplicates, dead code. | Runs ESLint, computes cyclomatic complexity, checks via ts-prune & jscpd. | | M-01 | CI/CD Pipeline | 15% | Build reliability & CI bottlenecks. | Analyzes CI YAML and fetches run history from GitHub/GitLab. | | M-04 | Test Flakiness | 14% | Detects inconsistently failing tests. | Parses JUnit XML reports cross-referenced against Git commit history. | | M-07 | Env Integrity | 13% | .env drift and secret leakage prevention. | Scans git history for secrets, diffs .env.example, checks Dockerfiles. | | M-08 | Build Performance | 10% | Caching efficiency & build bottlenecks. | Parses CI logs to find duration anomalies and cache misses. | | M-03 | Docs Freshness | 6% | Identifies stale documentation. | Correlates source code modification dates against related markdown sections. | | M-06 | PR Complexity | 4% | Flags PRs too large or complex for review. | Computes lines changed, cross-module impact, and review turnaround times. |

Health Score Labels

• 90–100 (EXCELLENT): No immediate action needed.
• 75–89 (GOOD): Address low-severity issues during sprints.
• 60–74 (MODERATE): Remediation sprint required within 2 weeks.
• 40–59 (HIGH RISK): Immediate action required on critical issues.
• 0–39 (CRITICAL): Block new features; dedicated remediation needed.

🤖 AI Features (Powered by MegaLLM)

All AI features are powered by the hosted backend and maintain grounded context via the local .ph-cache/. No API keys or authentication required.

1. ph ask (Natural Language Interrogation) Ask plain-English questions about the codebase. Replaces generic advice with specific file:line citations.
2. ph review (PR Review Co-pilot) Senior-level review of a branch or PR. Highlights bugs, security gaps, untested paths, and complexity spikes based on actual test coverage.
3. ph brief (Onboarding Briefing Generator) Automatically creates an ONBOARDING.md containing architecture summaries, ownership maps, entry points, etc.
4. ph chat (Conversational Codebase REPL) A persistent terminal chat session.
5. ph fix (Self-Healing Codebase Engine) Automatically remediate findings from your last scan using MegaLLM patch generation. It triages, fixes, and validates findings, supporting auto-generated fixes (complexity refactoring, dead exports, secret leaks, etc.). Modes: --auto (fully automated fix loop), --interactive, --dry-run, --ai.
6. Git Hook (Commit Doc Updater) A surgical post-commit hook that automatically amends (or opens a PR to update) documentation sections that reference code you just changed.

🛠 Tech Stack

Strictly TypeScript & Node.js ecosystem (Zero Java)

• CLI: commander, chalk, ora, execa, pkg
• AI Integration: openai (configured to point to hosted backend)
• Security Checkers: license-checker, js-yaml, dotenv
• Code Analyzers: @typescript-eslint/typescript-estree, complexity-report-es, ts-prune, jscpd
• APIs & Data Gathering: @octokit/rest, simple-git, fast-xml-parser

🚀 Getting Started

Installation (NPM)

Zero configuration needed! Just install from NPM and start using:

npm install -g project-healthy
cd your-project-folder
ph init

The CLI automatically uses our hosted backend (https://project-healthy.vercel.app/v1) for AI features. No API keys, no auth tokens, no environment variables to set.

1. Initialization / Interactive Shell

cd your-project-folder
ph init

ph init acts as your interactive CLI shell! It activates an interactive, continuous command loop. You can input commands straight from the prompt, remaining active until you press Ctrl+C or close the terminal. Additionally, it generates the .ph-cache/ context folder, builds your project-health.config.ts, and installs the smart git hooks.

2. Running a Scan

ph scan --format terminal

Runs the 8 parallel modules and logs the score to stdout. Use --format html for an interactive report, --format sarif for security tool integration, or --fail-under 80 in CI/CD pipelines.

3. Web Dashboard Integration

ph dashboard --port 8080

Launches a robust data integration Web Dashboard with responsive design updates via flexible wrapping utilities, ensuring metrics won't overlap.

4. Interactive Repository Explorer

ph explore

Launches a local web UI at http://localhost:7878 with a file tree with activity heat maps, commit history for each file, click-to-expand diffs inline, and module health impact visualization.

5. Self-Healing Auto-Fix Engine Pipeline

# Auto-fix all fixable findings automatically (Full pipeline)
ph fix --auto

# Interactive selection of findings to fix
ph fix --interactive

# Preview AI-powered fix patches without applying
ph fix --dry-run

# AI-powered fixes (uses hosted backend, no API key needed)
ph fix --ai

6. Additional Commands

# General AI assistance
ph ask "What are the main entry points?"

# Fast CI check with 30s timeout (exits 1 if below threshold)
ph ci-check --fail-under 80

# Compare current branch against base
ph diff --base main

# Print latest health score from cache
ph score

# Historical score trend analysis
ph trend

# AI Codebase Chat
ph chat

💻 Running the CLI Locally (From Source)

If you'd like to extend project-health or run the uncompiled source code yourself:

Step 1. Clone & Build

git clone https://github.com/Sanjay-Shahyml/Project-health.git
cd Project-health
npm install
npm run build

(Optional) Create a global link for the local binary:

npm link

Step 2. Optional Environment Variables

The CLI works out of the box with the hosted backend. You only need .env for optional integrations like GitHub/GitLab:

# Optional: Override the hosted backend URL (e.g., for local development)
# PROJECT_HEALTH_BACKEND_URL="https://project-healthy.vercel.app/v1"

# Optional: GitHub/GitLab tokens for M-01 (CI/CD) and M-06 (PR Complexity)
GITHUB_TOKEN=ghp_your_github_token_here
GITLAB_TOKEN=glpat_your_gitlab_token_here

# Optional: Snyk token for M-05 (Dependency Security)
SNYK_TOKEN=your_snyk_token_here

🖥️ Self-Hosting the AI Backend

The AI backend is deployed at https://project-healthy.vercel.app/v1 and works for all users. If you want to self-host your own instance:

Step 1: Deploy the ai-proxy

cd ai-proxy
npm install
npm run build

Deploy to Vercel, Railway, or any Node.js hosting provider.

Step 2: Configure Environment Variables

Set these on your hosting platform:

MEGALLM_API_KEY=sk-your-megallm-api-key
MEGALLM_BASE_URL=https://ai.megallm.io/v1
MEGALLM_MODEL=openai-gpt-oss-120b
MEGALLM_MAX_TOKENS=120000
MEGALLM_TEMPERATURE=0.7
PORT=3000
RATE_LIMIT_RPM=60

Step 3: Point CLI to Your Instance

ph chat "hello" --proxy https://your-deployed-backend.com/v1

Or set in .env:

PROJECT_HEALTH_BACKEND_URL="https://your-deployed-backend.com/v1"

https://www.youtube.com/@walecriss