365 Architect

Code Nest 365

CodeNest365 is a Visual Studio 2022+ VSIX extension that integrates multiple large language models — Anthropic Claude, DeepSeek R1, Ollama (local), and OpenAI-compatible endpoints — directly into the IDE. It functions as an intelligent engineering copilot: inline ghost-text completions, streaming conversational chat, context-aware code actions, unit test generation, Git commit messages, PR assistant, and persistent per-workspace memory — all with enterprise-grade security controls.

Features

Multi-Backend AI Routing Engine

  • Priority-based provider selection with automatic failover across Claude, DeepSeek, Ollama, and any OpenAI-compatible backend
  • Configurable per-provider timeouts (ghost-text 5 s, chat 30 s)
  • Extensible provider registry — add custom backends by implementing IAiProvider, IChatProvider, or IStreamingProvider

Inline Ghost-Text Completions

  • Single-line mode — real-time soft suggestions accepted with Tab, dismissed with Escape
  • Multi-line mode — configurable 1–10 line block completions stopping at logical boundaries (method bodies, closing braces)
  • Context-aware prompts using 20 lines of surrounding code + cursor position

Streaming Conversational Chat

  • Dockable WPF chat tool window with full conversation history
  • Live token-by-token streaming via Server-Sent Events (Anthropic & OpenAI SSE formats)
  • Stop button for mid-stream cancellation; non-streaming fallback for incompatible backends

Context-Aware Code Actions

Four right-click commands operating on the active text selection:

Action Description
Explain Clear, concise explanation of selected code's purpose and logic
Fix Issue Bug identification + corrected code with root-cause explanation
Refactor Architecture improvements for clarity, performance, and maintainability
Generate Tests Complete compilable NUnit test file with [TestFixture] / [Test], covering typical usage, edge cases, null inputs, and exceptions

Git Workflow Automation

  • Commit Message Suggester — reads staged or HEAD diffs and generates conventional-commit messages (<type>(<scope>): <summary>)
  • PR Assistant — diffs current branch vs main/master, generates structured PR description with Summary, Changes, Test Plan, Breaking Changes, and interactive Review Checklist

Per-Workspace AI Memory

  • Persisted in .codenest365/memory.json at solution root (team-shared via source control)
  • Auto-detects test framework (NUnit / xUnit / MSTest), LangVersion, and Nullable settings from .csproj and Directory.Build.props
  • Injected into every AI prompt as a ## Project Context block
  • Commands: Remember Selection and View/Edit Memory (DataGrid dialog)

Enhanced Ollama Integration

  • Health check, auto-start lifecycle, and model discovery via GET /api/tags
  • Model pull dialog with real-time progress bar and cancellation
  • Auto-pull fallback — silently downloads missing models on first request

Enterprise Security & Privacy

  • API keys stored in Windows Credential Manager (no plain-text exposure)
  • Full local-only mode — zero network calls when using Ollama
  • CI Mode Guard — automatically disables AI features in automated build environments
  • Opt-in telemetry only — no data transmitted without explicit consent

How It Works

CodeNest365 is a Visual Studio VSIX package (.NET Framework 4.7.2) that registers as an AsyncPackage. On initialization it:

  1. Loads user settings (API keys, provider selection, model preferences) from Tools → Options → CodeNest365
  2. Registers AI providers via ProviderRegistry — each provider is a factory so model changes take effect without restarting VS
  3. Creates a BackendRouter that implements IAiProvider, IChatProvider, and IStreamingProvider with priority ordering and automatic fallback
  4. Loads per-workspace memory from .codenest365/memory.json
  5. Registers all commands: chat window, code actions, commit message, PR assistant, and memory management

The router iterates through providers in priority order — if the primary provider times out or fails, the next provider is tried automatically. Providers communicate with their respective backends via REST/SSE, and the extension never blocks the VS UI thread (all AI calls run on background threads with JoinableTaskFactory).

Supported Providers

Provider Backend Use Case
DefaultProvider Anthropic Claude API Cloud-powered, highest quality completions and chat
DeepSeekProvider DeepSeek R1 API Cloud reasoning model with chain-of-thought traces
OllamaProvider Local Ollama server Air-gapped, offline, or cost-sensitive environments
OpenAiCompatibleProvider Any OpenAI-compatible endpoint Custom or self-hosted backends, Azure OpenAI, etc.

Getting Started

  1. Install the extension — download the .vsix from the releases page and double-click, or install via Visual Studio Marketplace
  2. Configure a provider — go to Tools → Options → CodeNest365:
    • For Claude: enter your Anthropic API key (stored securely in Windows Credential Manager)
    • For Ollama: ensure Ollama is running, select from discovered models, or pull a new one
    • For DeepSeek: enter your DeepSeek API key
  3. Start coding — ghost-text completions appear inline as you type (Tab to accept, Esc to dismiss)
  4. Use the chat panel — open via View → Other Windows → CodeNest365 Chat
  5. Explore code actions — right-click any selection for Explain / Fix / Refactor / Generate Tests
  6. Automate Git — right-click in Solution Explorer for Suggest Commit Message or PR Assistant

Requirements

  • Visual Studio 2022 (17.0+) — Full IDE (Community, Professional, or Enterprise)
  • .NET Framework 4.7.2
  • For cloud providers: valid API key (Claude, DeepSeek) or endpoint URL (OpenAI-compatible)
  • For local models: Ollama installed and running with a code-capable model (e.g., deepseek-coder, codellama, qwen2.5-coder)
★★★★★
Share on LinkedIn