Glyph — Micro AI Helper

A privacy-first, backend-agnostic AI assistant for code generation, documentation, and brainstorming in VS Code.

Architecture

Glyph uses a thin client + heavy server architecture:

┌─────────────────────────────────┐         ┌──────────────────────────┐
│    VS Code Extension (Client)   │         │   glyph-server (Docker)  │
│                                 │         │                          │
│  ServerClient ─────────REST────────────→  │  Express REST API        │
│  (HTTP + Socket.IO)                       │  Socket.IO               │
│                ─────Socket.IO──────────→  │  LLM Service (AI SDK)    │
│                                 │         │  LanceDB (Vector Index)  │
│  Background Indexer             │         │  Tool Registry           │
│  Brainstorm Chat UI             │         │  WorkspaceManager        │
│  Status Bar                     │         │                          │
└─────────────────────────────────┘         └──────────────────────────┘

The extension is a lightweight client (< 3 MB) that handles:

• VS Code UI (commands, webviews, status bar)
• Symbol extraction via VS Code's language services
• Configuration management (settings + secret storage for API keys)

The server handles all compute-intensive operations:

• LLM inference via multiple providers (Ollama, LM Studio, OpenAI, Anthropic, Google, OpenRouter)
• Vector embeddings + LanceDB indexing
• Tool execution (file read/write, terminal commands)
• RAG-based codebase context retrieval

Prerequisites

• VS Code ≥ 1.110.0
• Docker (recommended) or Node.js ≥ 22 for the server

Quick Start

1. Start the Server

# Using Docker (recommended):
docker compose up -d

# Or manually:
cd glyph-server
npm install && npm run build && npm start

2. Install the Extension

Install from the VS Code Marketplace or build from source:

npm install
npm run compile
npx vsce package
# Install the .vsix file in VS Code

3. Configure

The extension auto-activates on VS Code startup. On first launch:

1. The status bar shows connection status
2. If the server is running, you're ready to go
3. If not, click the status bar → "Start Docker"

Features

🤖 Code Generation

Select code → Right-click → Glyph: Code Generate → Enter prompt

📝 Documentation

Select function → Right-click → Glyph: Docs Generate

💬 Brainstorm Chat

Open command palette → Glyph: Brainstorm

Features:

• Real-time streaming via Socket.IO
• Tool use (file read/write, terminal commands)
• HITL permission flow for write operations
• Structure-aware context
• Chat memory with configurable limits

🔍 Background Indexing

On activation, the extension:

1. Scans all workspace source files
2. Extracts symbols using VS Code's language services
3. Sends them to the server for embedding + vector indexing
4. Watches for file changes and re-indexes incrementally

Settings

Setting	Default	Description
glyph.serverUrl	http://localhost:9741	URL of the Glyph server
glyph.serverAuthToken	""	Bearer token for remote servers
glyph.modelName	""	Active model identifier
glyph.providerType	"Ollama"	LLM provider (Ollama, LM Studio, OpenAI, etc.)
glyph.base_url	http://localhost:11434	Provider API base URL
glyph.embeddingModelName	""	Model for vector embeddings
glyph.chat.memoryLimit	15	Max messages in chat history
glyph.chat.structureAware	false	Include directory tree in chat context
glyph.chat.codebaseAware	false	Enable RAG context retrieval
glyph.chat.toolsEnabled	false	Allow AI to read/write files
glyph.reasoning.budgetTokens	10000	Token budget for reasoning models
glyph.agent.requireToolPermission	true	Require confirmation for write tools
glyph.autoStartServer	false	Auto-start Docker on activation
glyph.autoSave	false	Auto-save after code generation

Commands

Command	Description
Glyph: Code Generate	Generate or modify code
Glyph: Docs Generate	Generate documentation comments
Glyph: Brainstorm	Open interactive chat panel
Glyph: Model Select	Switch active model
Glyph: Cloud Provider Orchestrator	Configure cloud AI provider
Glyph: Setup Custom Model & API Key	Configure custom model endpoint
Glyph: Manage / Delete API Keys	Manage stored API keys
Glyph: Run Diagnostics	Check server + provider health
Glyph: Start Server (Docker)	Start glyph-server via Docker
Glyph: Reload Config	Reload configuration

Supported Providers

Provider	Type	Default Endpoint
Ollama	Local	http://localhost:11434
LM Studio	Local	http://localhost:1234
OpenAI	Cloud	https://api.openai.com
Anthropic	Cloud	https://api.anthropic.com
Google	Cloud	https://generativelanguage.googleapis.com
OpenRouter	Cloud	https://openrouter.ai

Development

# Install dependencies
npm install

# Build
npm run compile

# Watch mode
npm run watch

# Package VSIX
npm run release

# Lint
npm run lint

License

MIT