pycall-node 🚀

Call Python ML/AI functions from Node.js as native async functions. Seamlessly bridge the gap between Node.js and Python for heavy-duty inference tasks.

Install

npm install pycall-node

Quick Start (sklearn)

const { PyBridge } = require('pycall-node');
const py = new PyBridge({ autoInstall: true });

await py.start();
const prediction = await py.inference.predict({
  modelPath: './model.pkl',
  framework: 'sklearn',
  input: [[5.1, 3.5, 1.4, 0.2]]
});
console.log(prediction);
await py.stop();

Tier 1 Features

1. Core Bridge

Low-latency communication over NDJSON stdin/stdout.

• Unique Call IDs: Parallel calls are matched via UUIDs.
• Auto-Restart: Process crashes are handled with exponential backoff.
• Timeouts: Configurable per-call or global timeouts.

2. Model Inference Bridge

Native support for leading ML frameworks (sklearn, torch, tensorflow, yolov8, transformers).

3. Streaming Output

Perfect for LLM token streaming. Supports AsyncIterator and EventEmitter.

Tier 2 Features (GenAI & Vision)

4. RAG Pipeline Connector

Bridge Node.js directly into LangChain or LlamaIndex RAG pipelines.

// Query a Chroma RAG pipeline
const answer = await py.rag.query({
  backend: 'langchain-chroma',
  vectorstorePath: './chroma_db',
  embeddingModel: 'sentence-transformers/all-MiniLM-L6-v2',
  llm: 'ollama/llama3',
  query: 'What is the attention mechanism?'
});

// Stream RAG answer
const stream = py.rag.stream({
  backend: 'langchain-chroma',
  vectorstorePath: './chroma_db',
  query: 'Summarize the document'
});
for await (const token of stream) {
  process.stdout.write(token);
}

5. Embedding Generator

Generate vector embeddings from any provider (sentence-transformers, OpenAI, Ollama, FastEmbed).

const vector = await py.embeddings.encode({
  provider: 'sentence-transformers',
  model: 'all-MiniLM-L6-v2',
  text: 'Semantic search is powerful'
});

const results = await py.embeddings.search({
  provider: 'fastembed',
  model: 'BAAI/bge-small-en-v1.5',
  query: 'machine learning',
  corpus: ['deep learning', 'cooking pizza', 'neural nets']
});

6. Vision Model Bridge

Bridge Node.js to Python vision pipelines (YOLOv8, Transformers, OCR, DeepFace).

const detections = await py.vision.detect({
  framework: 'yolov8',
  modelPath: './yolov8n.pt',
  image: './scene.jpg'
});

const text = await py.vision.ocr({
  image: './invoice.png',
  languages: ['en']
});

Supported Frameworks

| Category | Libraries | |----------|-----------| | Core ML | sklearn, torch, tensorflow, onnxruntime, xgboost, catboost | | GenAI | langchain, llama-index, transformers, sentence-transformers, fastembed | | Vision | ultralytics (YOLO), opencv-python, pillow, deepface, easyocr, sam2 | | Providers | openai, ollama, cohere, huggingface_hub |

Requirements

• Node.js >= 16.0.0
• Python 3.8+
• pip (for auto-install)

License

MIT © sanjeevni112