Bisheng

Bisheng is an open source, self-hosted platform for teams building AI applications around business documents and processes. Its visual workflow editor combines automated tasks with human feedback, including intervention during multi-turn conversations. It's suited to document review, support ticket assistance and report generation that need more control than a single chatbot exchange.

Workflows support loops, parallel tasks, batch processing and conditional branches within one orchestration framework. Teams can combine different input and output types, generate reports with fixed layouts, and coordinate multiple agents. The Lingsight agent uses Agent Guidance Language (AGL) to incorporate domain experts' preferences and business rules into task handling.

Alongside workflows, Bisheng includes retrieval-augmented generation (RAG), shared model management, evaluation, supervised fine-tuning and dataset management. Its document parsing models cover printed and handwritten text, tables, page layouts and seals. Teams can deploy those parsing models privately.

For organizational use, the platform provides role-based access control, user groups, SSO/LDAP and traffic controls by group. Monitoring and usage statistics sit alongside support for high availability deployments. Bisheng runs through Docker and Docker Compose, with a browser interface, and uses the Apache 2.0 license. LLM processing depends on the configured model services; requests to cloud providers leave the platform. The stated minimum for the platform is four virtual CPU cores and 16 GB of RAM; the recommended server has 18 virtual cores and 48 GB of RAM.