Live from the Community Carnival! Open Source, Open Ecosystem, Shared Success — a 13-Year-Old Takes Second Place in AI Coding!
“We firmly believe that open source is the key engine driving a product’s continuous evolution. Especially as we explore AI-native scenarios, only by working shoulder to shoulder with the upstream and downstream ecosystem and with developers — creating and advancing together — can we go further.” So said OceanBase CTO Yang Chuanhui on January 31, at the 2026 OceanBase Community Carnival held in Shanghai.
The Community Carnival is OceanBase’s recurring annual flagship event, now in its third year, aimed at building an open, shared platform for technical exchange that connects developers and industry partners worldwide.
This event drew over 260 tech enthusiasts and developers, who through keynotes, panel discussions, AI Coding challenges, a community open mic, and other formats delivered more than ten high-quality talks, fully showcasing the vitality of the community ecosystem and its technical innovation.
Four years of open source, over 100,000 cumulative downloads
OceanBase is a 100%-independently-developed, native distributed database. It has long held to the philosophy of “applications driving technical innovation,” and officially announced its open-sourcing in June 2021.
OceanBase CTO Yang Chuanhui noted: “Foundational software gets good by being used. By whom? The answer is naturally developers.” He stressed that databases, as digital infrastructure, must grow together with their users and ecosystem.
OceanBase CTO Yang Chuanhui
He shared a set of figures: since going open source, OceanBase has surpassed 100,000 cumulative downloads worldwide, reached a deployment scale of millions of nodes, and attracted over 1,600 external contributors to co-build through code submissions, documentation improvements, bug fixes, and more.
Yang Chuanhui said developers are a key force in bringing technology to the ground and the cornerstone of building an innovative ecosystem. This is also why OceanBase open-sourced its first AI-native hybrid-search database, OceanBase seekdb, in 2025. OceanBase seekdb is built for “out of the box” use: with just three lines of code, developers can quickly build knowledge bases, agents, and other AI applications, and effortlessly handle tens-of-billions-scale multimodal data retrieval. “We are still in the exploration phase, and we look forward to more young developers joining us to advance the fusion of AI and database technology together.”
Open source, open ecosystem, shared success
“Every step of progress in the community is inseparable from the support of developers and co-builders,” said Feng Zhongyan, OceanBase’s open-source ecosystem lead. On stage, under the theme “Every Step with You — the OceanBase Community Carnival,” he shared OceanBase’s open-source philosophy and future plans.
Feng Zhongyan, OceanBase open-source ecosystem lead
Feng Zhongyan noted that OceanBase’s open-source journey has entered its fourth year, and that “open source, open ecosystem, shared success” is not just a simple slogan but a long-held philosophy. Over the past year, with the active participation of community developers and the collaborative push of ecosystem partners, the OceanBase Community Edition has gradually built up a complete set of enterprise-grade database capabilities.
To date, OceanBase has partnered with over 400 independent software vendors to jointly build more than 1,000 joint solutions, has more than 300 reseller partners and over 30 delivery partners, and has completed over 1,600 cumulative technical integrations — continuously empowering the digital transformation of thousands of industries.
Looking ahead to 2026, Feng Zhongyan said OceanBase will continue to deepen cooperation with ecosystem partners and gather more industry forces. On one hand, it will actively embrace AI technology and continue building the developer ecosystem; on the other, it will firmly advance its globalization strategy, joining hands with upstream and downstream ecosystem partners to expand into more use cases and help OceanBase reach a broader global market.
On the day of the event, OceanBase also officially appointed LangChain Ambassador Zhang Haili, Xenera LLM Project Lead Yi Hong, NVIDIA technical expert Cheng Zhiwei, Li Ziyi (a student at the National Cybersecurity College of Wuhan University), and He Wenchao of Shanghai Acmug Information Technology Co., Ltd. as its annual community ambassadors. Feng Zhongyan said he looks forward to walking the path of globalization together with more community ambassadors. At the same time, the 31 community moderators of the 2025 OceanBase community were also announced.
Ecosystem gathering: guests discuss building the AI data foundation
In the AI era, building a solid data foundation depends on the shared participation of the broad developer base and ecosystem collaboration. This event invited guests from various technical communities, who, drawing on their own practice and industry reflections, shared in depth around the path to building the AI data foundation. As an AI-native hybrid-search database, OceanBase seekdb became a high-frequency term in several guests’ talks.
RAGFlow CEO Zhang Yingfeng has personally lived through the leap from traditional search to the AI era. He gave a talk titled “From RAG to Context Engine: Building the Data Foundation for AI Agents.”
RAGFlow CEO Zhang Yingfeng
“In many people’s eyes, RAG may already be outdated technology, but I believe it’s precisely what can become the important foundation an AI-native database needs,” Zhang Yingfeng noted. The AI-native database of the future, he said, should not be merely a stack of models but should take “strong retrieval capability” as its core, building a context-engine architecture that can uniformly manage knowledge, data, and tools. Within this framework, a single RAG technique alone can no longer handle complex interaction scenarios, but it can evolve into a unified context engine that supports agents — through a “retrieval-first + context-optimization” mechanism, achieving comprehensive processing of structured and unstructured data as well as interaction memory.
He stressed that the essence of RAG lies in retrieval. The context engine of the future should be able to provide agents with precise information on demand, and, with the help of the OceanBase seekdb AI-native database, support multimodal, high-frequency hybrid retrieval — ultimately driving the technical leap from single-channel retrieval to all-around context service.
Zheng Li, Dify open-source ecosystem lead
Zheng Li, Dify’s open-source ecosystem lead, gave a talk on the theme “Dify x OceanBase seekdb.” Through concrete practice cases, he introduced Dify’s core capabilities and the path to building an integrated database in its collaboration with OceanBase seekdb.
Zheng Li noted that many multi-agent architectures emphasize AI’s autonomous decision-making and execution, yet actual business advancement still relies heavily on human communication, confirmation, and collaboration, which makes “fully automated” agents hard to land directly in real workflows. To address this, Dify holds to a design philosophy of “augmenting human capability,” letting AI blend into the workflow and boost efficiency rather than replace the human’s role.
In its collaboration with OceanBase seekdb, Dify completed an upgrade from a combination of multiple databases to a transaction-consistent unified data layer. On one hand, based on OceanBase seekdb, Dify has officially supported MySQL since v1.10.1. On the other hand, through a unified storage and retrieval architecture, OceanBase seekdb can simultaneously serve as the metadata database and provide hybrid search of vector and keyword (Hybrid Search), forming an out-of-the-box integrated deployment solution that further lowers deployment and operations complexity.
Miley Fu, DevRel and Founding Member of Second State
Miley Fu, DevRel and Founding Member of Second State, gave a talk titled “Building Customizable Agentic Voice AI: Echokit with OceanBase’s Hybrid Search.”
She introduced that WasmEdge has newly open-sourced an Agentic Voice AI product — Echokit — which emphasizes local deployment, supports fully offline operation, and balances privacy protection, controllability, and a high degree of customization. In this process, Echokit has partnered with OceanBase seekdb, using it as a local database for hybrid search. On why she chose OceanBase seekdb, Miley Fu cited its three major strengths: no CDC latency, native AI support, and good SQL compatibility — enabling atomic updates of vectors, metadata, and text, easy integration with agents, and making it well-suited to real-time voice AI scenarios.
Amy, Datawhale content ecosystem lead
Amy, Datawhale’s content ecosystem lead, took a community and education perspective with a talk titled “Steering Learning Toward Industry Value: Datawhale’s Thinking and Exploration.” As an open-source learning community founded seven years ago, Datawhale has always been committed to lowering the barrier to technical learning and helping developers master cutting-edge skills through hands-on practice. Amy said this philosophy aligns closely with OceanBase’s “open source, open ecosystem, shared success.” Datawhale plans to jointly build an AI + Database Learning Center with OceanBase, lowering the difficulty of getting started with database technology and helping build a healthy, sustainable developer ecosystem.
From knowledge enablement to architectural implementation, open-source tools are driving AI applications toward maturity.
Hu Yuewei, initiator and architect of the TEN Framework, shared his practice in real-time multimodal agent development in his talk “TEN Framework: How to Quickly Build a Low-Latency Conversational AI Agent with Memory.”
Hu Yuewei, TEN Framework initiator and architect
The TEN Framework is an open-source development framework for real-time multimodal AI agents, having earned nearly ten thousand stars on GitHub, with proven real-world deployment capability. The TEN Framework is currently developing a Voice AI Agent product and partnering with OceanBase PowerMem to achieve real-time synchronization and memory management of conversational context, providing underlying support for low-latency, high-concurrency conversational scenarios.
From the evolution of retrieval architecture and the building of an integrated data layer, to the landing of voice AI, open-source education, and framework enablement — the talks by these five guests not only presented the diverse paths to building the AI data foundation but also jointly confirmed the core value of open-source collaboration and ecosystem co-building in driving technology toward maturity.
Panel discussions: from RAG to AI, experts debate future directions
After the wonderful guest talks, two panel discussions on key issues of the AI era further sparked thought-provoking exchanges on site.
Against the backdrop of rapidly evolving artificial intelligence, RAG technology is becoming an important breakthrough point for bringing AI capabilities to the ground, and the related discussion was especially in-depth.
On site, LangChain Ambassador Zhang Haili, RAGFlow CEO Zhang Yingfeng, FastGPT lead Yu Jinlong, Co-founder of Nowledge Labs Gu Siwei, and Ji Jiannan, head of OceanBase’s AI Platform and Applications department, explored the topic “From Prompt to Skills — Is RAG Still Good Enough?”
From the angles of product practice, technical evolution, and system architecture, several industry experts argued that RAG is not outdated; on the contrary, its deep fusion with technologies like Skills, Memory, and databases gives it even more vitality. It is becoming the core infrastructure of context engineering and, through deep fusion with databases, skill systems, and memory mechanisms, is driving AI applications to leap from “Q&A toys” to “production-grade workflows.”
RAGFlow CEO Zhang Yingfeng said that from a RAG engine to a context engine, the technology doesn’t change, but its connotation changes with the times. On whether future RAG should rely more on databases for multi-path retrieval, Ji Jiannan, head of OceanBase’s AI Platform and Applications department, argued that RAG should be combined with the database — which is exactly the core of the “hybrid search” concept OceanBase has put forward. Co-founder of Nowledge Labs Gu Siwei, from a graph-database perspective, pointed out that the index structure should stay close to the essence of knowledge and support dynamic agent retrieval; FastGPT lead Yu Jinlong added an explanation of dynamic retrieval combining scalar and vector.
In the second panel, Xie Xiaoyu, an enterprise instructor for the artificial intelligence course at Nanjing University’s Graduate School, served as moderator, discussing with Sun Tao (core R&D engineer at Eigent and core member of CAMEL-AI), OceanBase Ambassador Cheng Zhiwei, Bian Sikang (head of products for Ant Bailing’s models), and Sun Jiajun (a founding-team member of Fellou) the topic “After the Year of the Agent, What Does Truly Usable AI Look Like?”
As AI technology delves deeper into real-world applications, one key issue is sparking wide discussion: is the barrier for humans to use AI rising? On this question, the experts argued that, although some AI tools still require a certain amount of configuration and learning cost, technical evolution is driving a fundamental shift in interaction. Looking back at the history of human-computer interaction — from DOS commands to the graphical interface — the technical barrier has always kept falling. Especially now, the significant improvement in LLM capabilities is making AI easier to understand and use. More and more products are trying to lower operational difficulty through interface guidance and visual interaction, letting non-technical users complete complex tasks with AI’s help.
This “human-centered” design trend means that in the future, AI will no longer be merely a tool for technical experts but will truly become a widely accessible capability available to everyone. In this process, how to make technology adapt to human habits, rather than making humans adapt to technology, will become an important direction in product evolution.
AI Coding challenge stages a peak showdown: a 13-year-old takes second place
In addition, this event innovatively set up an AI Coding challenge segment. OceanBase Ambassador Yi Hong gave a talk on the theme “Open Source, Agents, and AI Coding,” and live-built a coding agent “by hand” with zero code.
OceanBase Ambassador Yi Hong
In the AI Coding segment, ten awards were presented, including the “Fastest Merge Award,” the “Hardest PR Award,” the “Most Merges Award,” and the “Best Creativity Award.” Among them, OceanBase Ambassador Cheng Zhiwei won the “Best Creativity Award,” and Zhang Tianyu, a 13-year-old eighth-grader from Shanghai, took second place in the AI Coding “Hardest PR Award.”
In the past, taking part in open source often required first spending time getting familiar with a project, then completing the coding, debugging, and submission — a relatively high overall barrier. As AI Coding tools have become more capable, developers can get more assistance in understanding code, generating changes, locating problems, and refining submissions, and the barrier to participating in open source has fallen accordingly.
Before the event, OceanBase had already opened up 83 issues related to OceanBase and its ecosystem in a concentrated way in the OceanBase seekdb GitHub repository, making it easy for community developers to join the discussion and contribute.
Zhang Tianyu, who won second place in the AI Coding “Hardest PR Award,” chose the topic “Add a web dashboard for powermem,” which required developing a statistics API and a frontend page. He completed the frontend independently thanks to two years of React/Vue experience, while leaving the backend to AI-assisted generation. “What surprised me was that the AI-generated backend code ran through on the first try.”
In addition, during the afternoon community open mic, technical experts from FastGPT, CelHive, CAMEL-AI, Refly.AI, Dify, and OceanBase seekdb gave live demos showing how convenient it is to build agent systems and workflows on each AI platform. The most impressive part was that every platform demonstrated how to efficiently build agents and workflows through natural language — practically sounding the trumpet for an Agentic revolution.
For developers, using AI tools to quickly understand and get started with a project while focusing more on realizing ideas and exploring boundaries not only makes development smarter but also makes open-source co-building more sustainable and more creative — and this is the new theme and new opportunity that the AI era brings to the open-source ecosystem.
This Community Carnival used technology as a bond, effectively igniting the community’s innovative vitality. Looking ahead, we sincerely invite more developers and ecosystem partners to join us in expanding the application boundaries and the imaginative space of open-source technology.