🚀 Intern AI-Coding Workbench

A multi-tenant, dependency-aware platform orchestrating AI-assisted software engineering.

🎓 The DBERT Internship Program

This platform was proudly conceptualized, engineered, and developed under the DBERT Internship Program.

DBERT bridges the critical gap between academic learning and industry-grade software engineering. By empowering emerging tech talent to tackle real-world architecture challenges, the program cultivates innovation, rigorous development standards, and the hands-on experience necessary to build robust, scalable solutions. This Workbench stands as a testament to the high-caliber engineering fostered within the DBERT ecosystem.

---

🌟 Platform Overview

The Intern AI-Coding Workbench is an advanced environment designed to manage complex software projects. Admins bulk-upload Markdown project plans, and "Interns" claim tasks, code within isolated Git workspaces, and collaborate with a built-in AI assistant to seamlessly push Pull Requests to GitHub.

✨ Core Capabilities

• 🗄️ Multi-Repo Architecture: The backend dynamically provisions isolated Git bare-mirrors for every project. Manage 50 different tasks across 10 different GitHub repositories from a single dashboard.
• 🔗 Dependency-Aware Workflows: The system maps out dependency graphs (DependsOn:). Interns receive a "Sync from Main" button to automatically pull upstream dependencies into their active git worktree, and the AI is contextually blocked from writing final code until prerequisites are met.
• 🧠 "Bring-Your-Own-Key" AI Chat: Integrated with OpenRouter. Intern API keys are securely encrypted at rest (AES-256-GCM). The platform dynamically fetches live LLM access lists (GPT-4o, Claude 3.5, etc.) specifically tailored to the intern's credentials.
• 🤖 Smart Bulk-Import Engine: Admins upload standard .md files. A Two-Pass Regex Parser extracts TaskID, AssignTo, Repo, and Branch tags, safely performing data "upserts" to prevent database conflicts when correcting plans.
• ⚡ Zero-CLI Git Operations: Interns never touch a terminal. The system orchestrates git fetch, git worktree add, git commit, and git push entirely via the UI, backed by a background poller that syncs Pull Request statuses.

---

⚡ New Features & Engineering Enhancements

The platform has been enhanced through 3 foundational phases and 3 advanced enterprise options:

1. 🗂️ Phase 1: Editor Polish & Stability

• Multi-Tab Monaco Editor: Full multi-file tab navigation with file-type syntax detection.
• Dirty State Tracking: Real-time unsaved changes indicator (●), unsaved close confirmations, and Ctrl+S / Cmd+S shortcuts.
• FileTree CRUD Engine: Direct creation, renaming, and deletion of files/folders in the worktree with immediate UI updates.
• SQLite Concurrency & WAL Mode: Tuned SQLite with Write-Ahead Logging (WAL), 30-second busy timeouts, and serialized connection pools to prevent write locks under load.

Multi-Tab Code Editor with dirty indicators and syntax highlighting

---

2. 🌿 Phase 2: Git Worktree Lifecycle & Conflict Resolution

• Isolated Git Worktrees: Every intern task operates in an isolated, dedicated git worktree on disk (backend/data/workspaces/{intern}__{branch}_{id}).
• In-App Merge Conflict Resolution: Visual amber conflict banner with files affected, manual conflict marker resolution, Abort Merge (git merge --abort), and Complete Merge (git commit).
• One-Click Workspace Reset: Discard untracked and uncommitted changes instantly (git reset --hard && git clean -fd).
• Disk Worktree Pruning: Admins can prune stale or completed worktrees directly from the Admin Board to reclaim disk space.

In-App Merge Conflict Resolution Banner with Abort / Complete controls

---

3. 🤖 Phase 3: AI Assistant & Context Engine

• AI Context Inspector Modal: Inspect real-time prompt payloads, attached files, character count, and estimated token usage.
• Direct FileTree Context Attachment: One-click paperclip icon to attach files directly from the file tree into the prompt context bundle.
• Workbench Local Fallback Model: Built-in, offline-capable assistant producing streaming code edits without requiring an external API key.
• Monaco Side-by-Side Diff Editor: Visual side-by-side diff review modal comparing original vs. suggested AI edits before accepting changes.
• One-Click Diff Application: Writes changes directly to the worktree on disk, auto-opens the tab, and marks the suggestion as applied.

Monaco Side-by-Side Diff Review Modal with colored additions and deletions

---

4. 📊 Option A (Phase 4): Admin Analytics & Project Plan Importer

• Admin Analytics Dashboard: 4 live KPI telemetry cards for Active Worktrees, Pipeline Velocity %, Git Commits & PRs, and AI Token Usage.
• Multi-Track Progress Bar: Real-time distribution visualization across Full-stack, AI / ML, Data, and RPA tracks.
• Interactive Markdown Plan Importer: Drag-and-drop .md plan importer with a live dry-run parser previewing milestones, tasks, and assignees before committing to SQLite.

Admin Analytics & Telemetry Dashboard with KPI cards and track progress bar

Interactive Project Plan Importer Modal with live dry-run parser preview

---

5. 🧪 Option B (Phase 5): In-Workspace Test Runner & Terminal Drawer

• Subprocess Test Runner Backend: POST /api/workspaces/{id}/run-tests executes tests directly inside the isolated worktree with a 25-second execution timeout guard.
• Collapsible Terminal Drawer: Integrated test drawer with presets for pytest, pytest -v, python -m unittest, and active Python script runs.
• Status & Duration Telemetry: High-contrast PASS (Exit 0) or FAIL (Exit 1) status badges and millisecond duration timer.
• One-Click "Fix with AI": Automatically pipes failing terminal tracebacks and error messages into the AI Assistant prompt for instant remediation.

In-Workspace Test Runner executing pytest with PASS (Exit 0) badge and console report

Failing test run with traceback and instant "Fix with AI" error prompt forwarding

---

6. 📋 Option C (Phase 6): Interactive Kanban Board & DAG Dependency Graph

• Sprint Kanban Board: Toggle between Table and Kanban views with 4 workflow columns (Unassigned, In Progress, PR Open, Merged).
• Quick-Move Status Actions: Transition tasks across columns and reassign interns directly from the Kanban cards.
• Dependency Blocker Badges: Visual lock pill (🔒 Blocked: Waiting on [Prerequisites]) on tasks with unmerged dependencies.
• Interactive Dependency Graph (DAG) Modal: SVG visualizer with topological depth, colored directed arrows, and an interactive Node Inspector sidebar.
• Intern Dashboard Blocker Awareness: Informs interns of blocking prerequisites before starting tasks.

Interactive Sprint Kanban Board with 4 workflow columns and dependency blocker badges

Interactive Task Dependency Graph (DAG) Modal with topological SVG curves and Node Inspector

---

🛠️ System Architecture

Layer	Technologies
Frontend	React 18, Vite, Tailwind CSS, Monaco Editor (@monaco-editor/react), Lucide Icons
Backend API	Python 3.10+, FastAPI, SQLAlchemy, Uvicorn, Httpx, Pytest
Database	SQLite3 (Configured with WAL mode & busy timeout)
VCS / Shell	Native git subprocesses & worktrees, PyGithub for REST PRs
Testing Suite	Automated Playwright Chromium E2E Test Suites (backend/tests/test_phase*.py)

---

🚀 Getting Started (Local Deployment)

Prerequisites

• Node.js (v18+)
• Python (3.10+)
• System-level git installed on the host machine.

1. Backend Initialization

Open a terminal and configure your Python environment:

cd backend
python -m venv venv
.\venv\Scripts\activate   # On Windows (or source venv/bin/activate on Unix)
pip install -r requirements.txt

Configure Environment Variables (.env):

SECRET_KEY=your_fastapi_jwt_secret
MASTER_KEY=your_base64_32byte_encryption_key
ADMIN_EMAIL=admin@example.com
ADMIN_PASSWORD=internpass123
GITHUB_TOKEN=your_github_pat_for_opening_prs

Seed the SQLite Database: This generates the initial schema and creates your Admin/Intern test accounts:

python seed.py

Start the API Server:

uvicorn app.main:app --reload --port 8000

2. Frontend Initialization

In a separate terminal:

cd frontend
npm install
npm run dev

Navigate to http://localhost:5173 to access the Workbench.

---

🧪 Automated Testing & Verification

Run the full end-to-end Playwright test suite to verify all platform subsystems:

cd backend
# Phase 1: Editor & FileTree CRUD
.\venv\Scripts\python.exe tests\test_phase1_e2e.py

# Phase 2: Git Worktree Lifecycle & Conflicts
.\venv\Scripts\python.exe tests\test_phase2_e2e.py

# Phase 3: AI Assistant & Monaco Diff Editor
.\venv\Scripts\python.exe tests\test_phase3_e2e.py

# Phase 4 (Option A): Admin Analytics & Plan Importer
.\venv\Scripts\python.exe tests\test_phase4_e2e.py

# Phase 5 (Option B): Test Runner & Terminal Drawer
.\venv\Scripts\python.exe tests\test_phase5_e2e.py

# Phase 6 (Option C): Kanban Board & Dependency DAG
.\venv\Scripts\python.exe tests\test_phase6_e2e.py

---

📝 Admin: Markdown Plan Format

Admins can instantly generate database tasks by uploading .md files in this structure:

# FinTech Core Platform
Repo: https://github.com/your-org/your-repo.git
Branch: main

## Milestone 1: Data Architecture
### [data] Financial Data ETL Ingestion
TaskID: financial-etl
AssignTo: intern1@example.com

Streaming pipeline for real-time market quotes and transactions.

### [full-stack] Setup Auth Microservice
TaskID: auth-microservice
DependsOn: financial-etl
AssignTo: intern2@example.com

Configure OAuth2 and JWT session validation modules.

---

Engineered with ❤️ by the DBERT Engineering Team.