Product Definition
Tencent Cloud DataBuddy, the Big Data Agent Workbench is an Agent-Native, fully managed Data + AI integrated data intelligence platform launched by Tencent Cloud. DataBuddy integrates data computing and AI agent capabilities. Through a unified metadata and semantic layer, it provides enterprises with end-to-end, full-pipeline capabilities spanning from data ingestion, data engineering, and data science to data analysis and data governance.
DataBuddy enables big data platforms to evolve from "humans operating tools" to "AI working, humans overseeing." Users can complete complex tasks such as data ingestion, ETL pipeline construction, metric development, anomaly attribution, and visual dashboard creation using natural language, thereby lowering the barrier to entry and reducing Ops costs for data platforms.
Main Features
DataBuddy covers the entire data lifecycle. Its core capabilities can be summarized as "three major Agents, DIOps, a compute and storage foundation, and a governance framework":
Three Major Agents (For Business Scenarios)
Agent | Target audience | Core Capabilities |
Engineering Agent (DataBuddy-Engineering) | Data engineer | Data ingestion, data warehouse tiered architecture design, ETL task generation and orchestration, workflow scheduling, intelligent Ops |
Governance Agent (DataBuddy-Governance) | Data administrator / Governance team | AI-based identification and tiered classification of sensitive fields, intelligent recommendation of quality rules, data warehouse quality diagnosis, metadata completion, root cause diagnosis |
Analysis Agent (DataBuddy-Analysis) | Analyst / Business personnel | Natural language data query, analysis report generation, indicator fluctuation attribution, AI-powered dashboard creation, proactive push |
DIOps Integrated Engineering Capabilities
DIOps (Data Intelligence Ops) is DataBuddy's integrated engineering framework for the entire "Data + AI" lifecycle. It encompasses the three major Ops capabilities of DataOps, MLOps, and AIOps:
Ops Capabilities | Core Modules | Benefit |
DataOps | Studio (development), Workflow (orchestration and scheduling), CI/CD (release), Ops (monitoring) | Closed-loop big data development and Ops across the entire pipeline, with data, code, and configuration versioned. |
MLOps | Feature Management, Model Experimentation, Model Management, Model Serving, AI Gateway | Full lifecycle management for machine learning models, reducing the end-to-end model deployment cycle from 30 days to 7 days. |
AIOps | AI Intelligent Diagnosis, Root Cause Analysis, Automatic Repair, Intelligent Ops | Transforming from "manual troubleshooting" to "AI proactive diagnosis", with 60%+ of issues self-healed by AI. |
Computing and Storage Foundation
Data Lake Compute: Handles SQL, Python, and Notebook compute workloads, and supports elastic resources measured in CU (Compute Unit, where 1 CU is approximately equal to 1 vCPU + 4 GB of Memory).
Lakehouse Storage: A managed storage layer that hosts tables, volumes, and other multimodal data.
A Governance Framework
Unity Semantics: Provides ontology modeling, metrics, dimensions, and semantic models to reduce the risk of Agent hallucinations.
Data Quality: Provides full-pipeline quality rules, anomaly detection, and data profiling.
TCCatalog, the Multimodal Data Catalog: Manages structured data, unstructured data, ML models, and external data sources in a unified manner.
Agent Security Protection: Provides prompt injection detection, strict OBO (On Behalf Of) permission constraints, high-risk SQL interception, and precise "three-element" interception.
Product Architecture
DataBuddy adopts a three-layer architecture of interaction layer, platform layer, and engine layer (featuring integrated three-layer fusion and being a fully managed product):
Interaction Layer: Agent (engineering, analysis, and governance), GUI (full-featured visualization), and Notebook (interactive analysis and data science)
Platform Layer: Unified semantics, Data platform (integration, development, and scheduling), AI platform (ML, GenAI, and inference), and unified governance (data governance + AI governance).
Engine Layer: Data lake compute, lakehouse storage, integrated Data + AI, and multimodal support.