BISHENG Enterprise LLM DevOps Platform
An open-source platform for building and operating enterprise-grade LLM applications, with workflow orchestration, RAG, agents, model management, and fine-tuning.
- Source repo
- dataelement/bisheng
- Stars
- ★ 12k
- Last updated
- 3d ago
- License
- Apache-2.0
- Primary language
- Python
- FA score
- 0/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changes
- Cost
- Free, no paid service needed
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Enterprise knowledge base creation: use built-in document parsing and RAG to turn unstructured documents into searchable knowledge for internal QA.
- Not a fit if
- Teams without multi-core, high-RAM hardware for the default stack
- Users wanting a lightweight deploy without ES, Milvus and OnlyOffice
- English-first teams that need docs outside the Feishu Wiki
- Source review
- 0/100 · Major gaps 8 safety controls not found
What does this agent do, and when should you use it?
BISHENG is an open-source LLM application DevOps platform designed for enterprise scenarios. It provides a comprehensive suite of capabilities including visual workflow orchestration, RAG, agent building, unified model management, supervised fine-tuning (SFT), dataset management, evaluation, enterprise-level system management (security review, RBAC, SSO/LDAP), and observability. The platform's workflow engine supports loops, parallelism, batch processing, conditional logic, and human-in-the-loop intervention within a single framework. It also bundles high-precision document parsing models (OCR) that can be privately deployed for free. BISHENG has been adopted by industry leaders and Fortune 500 companies, making it suitable for building document review, fixed-layout report generation, multi-agent collaboration, customer service assistance, and meeting minutes generation applications.
BISHENG runs as a Dockerized platform started via docker compose, offering a web UI to orchestrate workflows, manage models, create RAG applications, run evaluations, and perform fine-tuning. Core components include a workflow orchestration engine (with loop, parallel, batch, and conditional capabilities), RAG pipeline, agent orchestration (including Lingsight, a general-purpose agent based on the AGL framework), unified model management, SFT module, dataset management, evaluation module, and enterprise management features (RBAC, user groups, SSO/LDAP, traffic control, monitoring). It also integrates high-precision OCR for document parsing. A typical flow: users design workflows via the UI, configure models and data sources, then the platform executes tasks such as document parsing, retrieval-augmented generation, agent collaboration, while providing monitoring and feedback.
- Enterprise knowledge base creation: use built-in document parsing and RAG to turn unstructured documents into searchable knowledge for internal QA.
- Complex business process automation: design multi-step, conditional workflows via visual interface for approval, report generation, or content review.
- Multi-agent collaboration: leverage AGL-based Lingsight agent to embed domain expertise for tasks requiring professional judgment.
- Customer service support: use models to understand customer conversations and provide real-time response suggestions to agents.
- Model fine-tuning and evaluation: use the built-in SFT module and evaluation tools to adapt open-source models to specific domains.
- Document parsing and governance: apply high-precision OCR to scanned documents and handwritten notes for data governance and archival.
How do you install or deploy this agent?
Requirements: CPU ≥ 4 cores, RAM ≥ 16GB, Docker 19.03.9+ and Docker Compose 1.25.1+. Steps:\n1. Clone repo: git clone https://github.com/dataelement/bisheng.git\n2. Enter install dir: cd bisheng/docker\n3. Start: docker compose -f docker-compose.yml -p bisheng up -d\n4. Open http://IP:3001 in browser and register first user (becomes admin).\nFor more deployment details, refer to the self-hosting guide: https://dataelem.feishu.cn/wiki/BSCcwKd4Yiot3IkOEC8cxGW7nPcHow do you use this agent?
After installation, open http://IP:3001, log in with your registered account. In the UI, you can: 1) choose an application type (workflow, RAG, agent); 2) configure models (e.g., OpenAI API); 3) design workflows or RAG pipelines; 4) add data sources and test; 5) release the app for end users. Administrators can configure RBAC, user groups, SSO/LDAP in the system management section. Detailed usage guide is available in the Wiki: https://dataelem.feishu.cn/wiki/ZxW6wZyAJicX4WkG0NqcWsbyndeWhat are this agent's strengths and limitations?
- Visual workflow supports loops, parallelism, and batch processing intuitively by 'drawing a flowchart', lowering the learning curve.
- Built-in high-precision document parsing (OCR) for printed, handwritten, and rare characters, freely deployable on-premises.
- Comprehensive enterprise features: RBAC, SSO/LDAP, monitoring, traffic control, suitable for enterprise rollout.
- End-to-end LLM lifecycle management: RAG, agent orchestration, SFT, and evaluation in one platform.
- Requires Docker and substantial hardware (recommended 18 cores, 48GB RAM), leading to high deployment overhead.
- Default brings third-party components (ES, Milvus, OnlyOffice) that add resource usage and operational complexity.
- Documentation relies heavily on Feishu Wiki; some content may require external access or login.
- Model support depends on configuration; integrating proprietary models may require extra adaptation.
How does this agent compare with similar options?
The README acknowledges inspiration from langchain, langflow, unstructured, and LLaMA-Factory, but does not explicitly name competitors.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| BISHENG Enterprise LLM DevOps Platform This agent | 0 · Major gaps | Self-hosted serviceFree | ★ 12k | 3d ago | Python | — |
| Astron Agent | 21 · Major gaps | Self-hosted serviceFree + model costs | ★ 8.9k | 5d ago | Java | — |
| EpicStaff Agent Workflow Platform | 57 · Major gaps | Self-hosted service | ★ 359 | 10d ago | Python | — |
| ODS Private AI Server | 71 · Some gaps | CLIFree | ★ 7.6k | today | Python | OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: README claims enterprise-grade features (security review, RBAC, SSO/LDAP) but provides no implementation details or code evidence. SECURITY.md provides a vulnerability reporting process but does not address least privilege, user confirmation, data flow transparency, sensitive data handling, or dependency security. Test files contain hardcoded API keys (e.g., 'sk-b4c85ea486a04aab8041dc9f36d20617'), posing a serious security risk. No evidence of rollback mechanisms or source attribution. Therefore, all trust criteria score 0 due to insufficient evidence or negative evidence (hardcoded keys).
Evidence shows: README claims powerful features but provides no detailed information on self-consistency, dependency availability, or failure messages. CI workflows have hardcoded versions (e.g., 'RELEASE_VERSION=1.3.1'), which may be inconsistent. Test files contain hardcoded keys, indicating poor test quality. No evidence of failure messages or dependency availability. Therefore, all reliability criteria score 0.
Evidence shows: README describes target audience (enterprise) and scenarios (document review, report generation, etc.) but does not clearly define capability boundaries or trigger precision. Environment fit only mentions Docker and hardware requirements, without detailed configuration. Therefore, audience_and_scenarios scores 1 (described but lacking support), and the rest score 0.
Evidence shows: README provides installation instructions (Docker Compose) but lacks information architecture, naming stability, examples/FAQ, known limitations, or versioning/changelog. LICENSE is Apache-2.0, but no versioning or maintenance responsibility evidence. Therefore, install_notes and license score 1, and the rest score 0.
Evidence shows: README claims high-value features but provides no evidence of output usability, marginal value, or cost-benefit. Therefore, all effectiveness criteria score 0.
Evidence shows: Claims in README lack traceability, no cross-source corroboration, and no separation of facts and inferences. Therefore, all verifiability criteria score 0.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- Hardcoded API keys in test files may leak sensitive credentials.
- Hardcoded version numbers in CI workflows may cause release inconsistencies.
- Feature claims in README lack code or documentation support; evaluate with caution.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v3.0.0-beta5 Pre-release
Release notes excerpt
What's Changed * Feat/3.0.0 beta4 by @jieyuhuayang in https://github.com/dataelement/bisheng/pull/2466 * Feat/3.0.0 beta4 by @zgqgit in https://github.com/dataelement/bisheng/pull/2468 Full Changelog: https://github.com/dataelement/bisheng/compare/v3.0.0-beta4...v3.0.0-beta5View on GitHub ↗