NebulaCode Enterprise Agent

# NebulaCode Enterprise Agent NebulaCode is a web-based autonomous coding agent for enterprise software delivery. It moves beyond a copilot by planning work, routing between specialist agents, executing inside a GitHub Codespace, validating results, reviewing code, and persisting project memory. ## Hackathon Positioning - Track: Agentic Workflows, Intelligent Reasoning, Collaborative Systems, Enterprise Utility, and Multimodal Intelligence. - Enterprise use case: managers, founders, and technical teams can turn workflow ideas, docs, screenshots, or voice notes into working internal tools. - Measurable value: shorter request-to-prototype cycles with an auditable decision trace, safer tool execution, automated validation, and persistent project memory. - Deployment target: Vultr VM backend with a public web frontend. ## Architecture - Frontend: React, Vite, Tailwind, Monaco editor, xterm.js, WebSocket session streaming. - Backend: Express, LangGraph JS, Gemini, Featherless review tool, Speechmatics-compatible voice route, PostgreSQL persistence. - Agent graph: Planner -> Supervisor -> Coder / Researcher / Tester / Code Reviewer / Repo Analyst / Memory Writer. - Runtime systems: GitHub OAuth, GitHub Codespaces, PostgreSQL checkpointer, session embeddings, WebSocket telemetry. - Enterprise telemetry: each run emits a hackathon track profile, value metric, workflow steps, integrations, and supervisor decision traces. - Repo intelligence: large repositories are summarized into `REPO_INTELLIGENCE.md` with framework, directories, key files, symbols, imports, and file responsibilities so the repo reviewer can inspect architecture without loading every source file. - Voice: users can record from chat; the backend transcribes through Speechmatics batch mode now, with a documented path to real-time WebSocket transcription.