Agent Development at Ant Afu
A healthcare AI app serving 100M+ users runs an agent-driven build-and-test loop on Fork Database, with isolated sandboxes for each agent.
Meet OceanBase AI Database, the unified database for operational data, real-time analytics, and AI. Explore ->
Meet OceanBase AI Database, the unified database for operational data, real-time analytics, and AI. Explore ->
The unified database for operational data, real-time analytics, and AI.
Combine the openness of a data lake with the structure and online serving of a database - one architecture for multimodal data, real-time analytics, and AI agents.
Structured, semi-structured, and unstructured data in one architecture — no copies, no silos.
One metadata
Govern every data type through a unified metadata layer.
One copy of data
Analytics and AI run on the same data — no movement, no duplication.
One architecture
Unite lake storage and database serving to simplify the AI data stack.
Store fields, JSON, full-text, vectors, and large objects as native columns of one table.
Adaptive storage
Small values stay inline, large ones move to object storage.
AI columns
Embeddings run as table columns — retryable, consistent with data.
Unified governance
One set of permissions, lineage, and lifecycle across every data type.
Vector, full-text, and structured filtering in a single query.
Real-time retrieval
Search fresh operational and multimodal data without external sync.
Higher relevance
Semantic + keyword recall, ranked in-engine with strong consistency.
Simpler stack
No separate vector database or search engine to run and sync.
Persistent memory, context, and state for Agents — all in one place.
Fork database
Create isolated sandbox in milliseconds, rollback when needed.
Per-agent boundaries
Give each agent its own data scope on shared resources.
Scale elastically
Grow from one agent to large concurrent agent workloads.
Connect storage and compute engines without rebuilding the stack.
Open storage
Support S3-compatible object storage and Iceberg table formats.
Open compute
Run SQL, Spark and Ray on the same data.
Lower cost
Multi-tenancy and advanced compression cut storage cost up to 70–90%.
Bring transactions, user behavior data, chats, and logs into one consistent view for real-time risk blocking, follow-up reviews, and historical investigations.
Analyze live operational data the moment it lands, so dashboards, reports, and pricing always reflect the current moment.
Run one agent or a massive fleet on shared infrastructure, each with persistent memory and its own isolated sandbox.
Resolve customer questions from a single, real-time view of business data and knowledge — no conflicting answers.
Match users to the right items in milliseconds, drawing on one profile that unifies their attributes, behavior, and preferences.
Replace fragmented database, search, vector, and analytics tools with one engine — cutting complexity and storage cost.
From healthcare AI to large-scale agent platforms, teams run on OceanBase to simplify data stack, scale agent workloads, and reduce infrastructure costs.
A healthcare AI app serving 100M+ users runs an agent-driven build-and-test loop on Fork Database, with isolated sandboxes for each agent.
A consumer AI platform sustains massive 24/7 agent concurrency while consolidating a multi-system stack onto one engine.
China Unicom uses RAG to bring real-time operational knowledge into AI workflows, improving response speed and answer accuracy for database operations.
Lalamove brought AI into logistics with OceanBase: loss-risk code detection and a data-warehouse Q&A assistant on a simplified stack.
Open formats and native connectors mean LakeBase works with your frameworks, engines, and tools — no rip-and-replace.