🧠 UltraCodeAI – IntelliJ AI Assistant (Ultra-fast, Local + Cloud LLMs)

"An IDE-first AI copilot for IntelliJ IDEA with chat, diagnostics, and project-aware insights."

---

🚀 Overview

UltraCodeAI is an IntelliJ IDEA plugin that embeds an AI coding copilot directly in the IDE:

• Multi-model chat with streaming responses
• Real‑time Diagnostics tab powered by IntelliJ inspections
• Project‑aware Insights (RAG-ready structure)
• One clean, native, tabbed tool window: Chat • Diagnostics • Insights

It works with both local models (privacy-first) and cloud models (frontier capabilities), and is built 100% in Java for deep IntelliJ integration.

---

✨ Features

• 🤖 Chat tab with streaming responses (token-by-token)
• 🔍 Diagnostics tab with live IntelliJ error/warning highlights
• 💡 Insights tab scaffold for project-aware guidance (RAG-ready)
• 🧩 Local + Cloud model support via a unified execution service
• 🧭 Native tool window (right dock), theme-aware, keyboard-friendly
• 🧱 Pluggable services (execution strategies, RAG/indexer, MCP/logs)
• 🧰 Context controls: add files/snippets/directories from project view

---

🧩 Architecture

• Tool Window UI: com.devoxx.genie.ui.window.UltraCodeAIToolWindowFactory
• Prompt Pipeline: com.devoxx.genie.service.prompt.PromptExecutionService
• Streaming Interface: com.devoxx.genie.service.prompt.response.ResponseListener
• Diagnostics Listener: com.devoxx.genie.service.diagnostics.UltraCodeAIDiagnosticsService
• Settings Panels: multiple projectConfigurable entries in plugin.xml
• Optional Panels: MCP Logs (bottom tool window), Appearance, Costs, Web Search
• Extensibility: RAG indexer, providers, threading, cancellation, memory

---

Folder hints (typical layout): ---bash---

Folder hints (typical layout):

UltraCodeAI/src/main/java/com/ ├─ ui/ │ ├─ window/ # Tool windows (UltraCodeAI, MCP logs) │ ├─ panel/ # Prompt/response panels (if present) │ └─ settings/ # Settings pages ├─ service/ │ ├─ prompt/ # Execution, memory, threading, strategies │ ├─ diagnostics/ # Live analyzer listener (Diagnostics tab) │ ├─ rag/ # Indexer & retrieval (optional, pluggable) │ └─ mcp/ # MCP execution/logging (optional) └─ chatmodel/ # Local & cloud provider services ---bash---

📦 Installation

From source: bash git clone https://github.com/vishnupriyanpr/UltraCodeAI.git cd UltraCodeAI ./gradlew buildPlugin

text

Install into IDE:

• Open IntelliJ → Settings → Plugins → Install from Disk
• Select zip from build/distributions/

Dev run: bash ./gradlew runIde

Requirements:

• IntelliJ IDEA 2023.3.4+ recommended
• Java 17+

---

⚙ Configuration

Open Settings → UltraCodeAI:

• Select provider (local or cloud)
• Optionally set API keys for cloud providers
• Adjust appearance, cost estimation, context window preferences
• Enable/disable MCP, Web Search, RAG (if available in your build)

Tool window:

• View → Tool Windows → UltraCodeAI (right side)

---

🔧 Usage

Chat:

1. Open UltraCodeAI tool window
2. Type a prompt (optionally include code context)
3. Watch streamed response in real time

Add context:

• Right‑click code editor/project files → “Add To Conversation”
• Add directory or calculate tokens for large context

Diagnostics:

• Open Diagnostics tab
• Edit code; issues auto‑populate from IntelliJ analyzer
• (Optional) Click entries to navigate (if wired in your build)

Insights:

• Open Insights tab
• See contextual recommendations (when connected to your RAG/provider flow)

---

🧠 Key Components (Glue Code Highlights)

• UltraCodeAIToolWindowFactory
  • Builds the 3-tab UI
  • Hooks Chat send-button to PromptExecutionService.executePrompt(...)
• PromptExecutionService
  • Primary pipeline: command processing → strategy selection → execution → cancellation → cleanup
  • Overload provided for simple String + ResponseListener Chat integration
• ResponseListener
  • Minimal interface: onTokenReceived, onComplete
• UltraCodeAIDiagnosticsService
  • Listens to IntelliJ daemon analyzer completion
  • Feeds current file highlights to Diagnostics tab UI callback

---

🛠 Building Blocks You Can Extend

• Providers: add your own local/cloud services under chatmodel/
• Strategies: register new PromptExecutionStrategy variants
• RAG: wire your indexer/retriever to enrich Insights tab
• WebView: swap JTextArea for JCEF/HTML+Prism for rich rendering
• Actions: add editor/project view actions to push context to chat

---

🧪 Tips

• For streaming UIs, append tokens on EDT using SwingUtilities.invokeLater(...)
• For large projects, avoid blocking EDT; use background tasks
• Prefer project services (project.getService(...)) for scoped lifecycles
• Keep plugin.xml single source of truth for factories, services, actions

---

🧭 Roadmap (suggested)

• Click-to-navigate diagnostics entries (offset → line → OpenFileDescriptor)
• AI fix suggestions inline in Diagnostics (one-click apply where safe)
• Rich chat rendering via JCEF + Markdown + code highlight
• Provider presets and smart model recommendations by task
• Full RAG wiring with local embeddings + retrieval panel

---

🤝 Contributing

PRs welcome! Suggested flow:

1. Fork and branch: feature/
2. Code + tests
3. ./gradlew buildPlugin
4. Open PR with a clear description and screenshots/GIFs for UI changes

---

📜 License

MitLicense-2.0 (see LICENSE in the repo)

---

🙌 Credits

Vishnupriyan P R.
Vivek K K.
Akshaya K.
Sanjit M.

Crafted by the MeshMinds. Inspired by modern agentic IDE workflows and built with the IntelliJ Platform SDK + Java.