BISHENG is an open LLM DevOps platform for building enterprise AI applications โ visual workflow orchestration, retrieval-augmented generation, agents, model management, evaluation and fine-tuning in one self-hosted system. Developed by DataElement and named after Bi Sheng, the inventor of movable-type printing, it has earned about 12,100 stars on GitHub and is used by Fortune 500 organizations.
Enterprises need more than a chat demo: they need governance, permissions and repeatable processes around LLM applications. BISHENG’s workflow engine is its centerpiece โ a single orchestration framework where multi-turn conversation and automated processing coexist, with loops, parallel branches, batch runs and human-in-the-loop steps that pause for input or approval. On top of that sit team knowledge spaces with file-level permissions and cited AI Q&A, a general-purpose task mode where an agent plans multi-step work and delivers reports, presentations and spreadsheets using built-in Office skill packs, and high-precision document-parsing models for printed text, handwriting, tables and seals that can be deployed privately. The platform covers the operational lifecycle too: unified model management, evaluation, supervised fine-tuning, dataset management, RBAC plus relationship-based permissions, SSO/LDAP, approvals and audit logs. It installs with Docker and has stayed in rapid development through 2026.
BISHENG’s open-source platform is free under the Apache-2.0 license and can be self-hosted with Docker. Some capabilities โ including preset analytics dashboards and the DSH Desktop enterprise access client โ are part of DataElement’s commercial edition, so the overall offering is freemium.
BISHENG is for enterprise IT and AI teams โ particularly in China and other markets valuing self-hosting and domestic-stack support โ that need a governed, full-lifecycle platform for building and operating LLM applications, not just a prototype toolkit.