Hangarx
Hangarx
← Back to Sprout Docs

Introduction to Sprout

Hangarx Sprout is an enterprise-grade AI research assistant with advanced reasoning capabilities, built on a serverless architecture with extended compute limits, featuring multi-modal file processing, integrated authentication, billing, and multiple AI model backends.

Platform Foundation

Deployment PlatformVercel Serverless Functions
RuntimeNode.js 20.x
Compute ModelFluid Compute (extended execution)
Timeout Limit800 seconds (13.3 minutes)
Memory AllocationUp to 3008 MB

AI Model Stack

Primary Reasoning Engine

Kimi K2 Thinking Model (Moonshot AI)

  • • Context window: Up to 200K tokens
  • • Optimized for deep reasoning and research tasks
  • • Supports chain-of-thought and tool use

Secondary Inference

Huggingface Inference Endpoints

  • • Text generation: Llama 3, Mixtral, custom models
  • • Image processing: ViT, CLIP, Salesforce BLIP
  • • Vision Models: Integrated for OCR and image understanding

Development Framework

Sprout is built on the Vercel AI SDK v3.0+ with the following key features:

  • Streaming responses with StreamingTextResponse
  • Tool use and function calling
  • Multi-modal input handling
  • Built-in prompt management
  • Edge runtime compatibility

Built-in Tools

tavily_search

Web research with citations

tavily_extract

Deep content extraction from URLs

file_processor

Handle uploaded documents

image_analyzer

OCR and visual understanding

calculator

Mathematical computations

code_interpreter

Safe sandboxed code execution

Next: Quick Start →