@neohub/core

Core functionality for NeoHub AI assistant - Ollama integration, model management, and intelligent task routing.

Features

• OllamaClient: Full-featured client for Ollama API with streaming support
• ModelSupervisor: Intelligent model selection based on task type and context
• ConversationManager: Manages chat history and context windows
• CodeAnalyzer: Analyzes code complexity and generates context for AI models

Installation

npm install @neohub/core

Usage

import { OllamaClient, ModelSupervisor } from '@neohub/core';

// Initialize Ollama client
const client = new OllamaClient({
  baseUrl: 'http://localhost:11434',
  model: 'deepseek-coder:33b',
});

// Use Model Supervisor for intelligent model selection
const supervisor = new ModelSupervisor(client);
const model = await supervisor.selectModel({
  taskType: 'code_generation',
  codeComplexity: 'medium',
  userPreference: 'auto',
});

Testing

# Run unit tests (fast, no external dependencies)
npm test

# Run integration tests (requires Ollama with large models)
npm run test:integration

Note: Integration tests require:

• Ollama running on http://localhost:11434
• At least one chat-capable model installed (e.g., deepseek-coder:33b, codellama:34b)
• Sufficient resources (large models need significant RAM and CPU)

Integration tests are excluded from the default test suite and CI pipeline due to resource requirements and execution time (30+ seconds per test).

Development

# Build TypeScript
npm run build

# Watch mode for development
npm run dev

# Clean build artifacts
npm run clean

License

MIT