How Huolala’s Wukong Platform Accelerates Enterprise AI with Low‑Code & Zero‑Code

Background

Huolala is an internet logistics technology company operating in 11 global markets, covering 363 cities in mainland China with 900,000 active drivers and 12 million active users as of the end of 2023. The emergence of pre‑trained large models has sparked rapid growth in AI applications, prompting the company to integrate AIGC solutions into its technology and operations.

Huolala’s Large‑Model Application Status and Insights

Huolala has deployed large‑model applications across marketing, invitation, customer service, and outbound calling, while internal R&D, testing, and design teams also seek to leverage these models. Challenges include duplicated effort, lack of a unified knowledge base, and mismatched understanding between business needs and model capabilities.

Wukong Platform – One‑Stop Large‑Model Development

The Wukong platform provides a low‑code and zero‑code development environment built on large‑model core capabilities, offering diverse interactive entry points and a visual SOP for efficient, stable, and scalable AI application development.

Layered Development System

User Layer: All end‑users and business scenarios that consume large‑model capabilities, such as marketers, operation staff, and developers.

Interaction Layer: Integrates browsers, mini‑programs, Feishu workbench, web pages, bots, and Open API, enabling one‑click publishing to required channels.

Development Layer: Core visual AI tool platform that connects Huolala’s proprietary models and external models, offering unified low‑code and zero‑code AI application building and publishing.

Model Layer: Supports Huolala’s self‑developed freight model and integrates multiple open‑source and commercial foundation models.

Infrastructure Layer: Underlying services such as databases, knowledge bases, gateways, and third‑party dependencies.

Core Functionalities

Zero‑Code AI Applications

Users interact with a conversational AI assistant that gathers requirements, dynamically adjusts configurations, and automatically incorporates tools, knowledge, or models needed to fulfill the task.

Low‑Code AI Applications

Low‑code development uses a workflow‑oriented visual canvas where each component node provides a specific function (e.g., AI Agent, Chain, Memory, Prompt, Knowledge Base, Tool, Output Parsing). Users can drag, drop, and connect components to create executable AI pipelines.

Knowledge Management

Provides a knowledge base that supplies context during inference, reducing hallucinations without requiring model fine‑tuning. The backend includes data extraction, embedding models, vector engines, re‑ranking, and RAG pipelines.

AI Tools

Custom tools can be registered to expose internal services (e.g., weather, web search, image generation, code interpreter). These tools extend AI capabilities, enabling tasks such as automated activity creation or real‑time data retrieval.

Application Monitoring

Offers macro‑level metrics for all AI applications and detailed dashboards for individual apps, including call volume, cost, token usage, inference traceability, and root‑cause analysis.

Lalabot – Enterprise AI Assistant

Lalabot builds on the Wukong platform to provide a multi‑modal AI assistant across browsers, mini‑programs, and web, offering chat, writing, image generation, summarization, and custom agents.

Key Features

Internet Search: AI‑driven query analysis and multi‑source result synthesis.

Custom Agents: Deploy AI agents built on Wukong to handle simple to complex dialogues, shareable across the organization.

Intelligent Summarization: Generates concise summaries of long documents, meeting notes, or chat histories.

Writing & Drawing: Generates outlines, drafts, and images from brief user prompts.

Real‑World Deployments

Wukong has been applied in over 14 business units and 50+ real scenarios, including:

Intelligent Invitation Training: AI assistant with ASR, TTS, and digital humans for interactive training and quality inspection.

Professional Assistant: AI‑powered Q&A, automated weekly reports with charts, and document generation.

AI Fault Analysis: Uses AI to diagnose Kubernetes deployment failures and filter false‑positive alerts.

Summary & Outlook

As Huolala’s AI coverage expands, business demands evolve toward multimodal models and multi‑agent technologies. The Wukong platform continues to explore and integrate emerging AI capabilities to further empower Huolala’s operations.