BG

bionic-gpt/bionic-gpt

部署与 DevOps
2383 stars 质量 51 趋势 51

Bionic is sovereign Agentic AI for the enterprise — Runs on-premise and can securely work with your sensitive data and systems.

概览

Bionic is an open-source Rust agentic harness for internal AI teams building sovereign AI. Deploy it on-premise, in private cloud, or in air-gapped environments. Connect your own models, data, tools, and internal systems, then build organisation-specific AI workflows without rebuilding the whole platform stack. Open source. Self-hosted. Model independent. Homepage Documentation Go Bionic Contributing Most internal AI teams do not need another generic chat UI. They need a controlled runtime where models can use tools, inspect files, call approved integrations, run code, retrieve knowledge, and produce durable outputs. Bionic provides that foundation. Your team keeps control of the models, infrastructure, integrations, governance rules, and business-specific workflows.

README

Bionic

Why Bionic

Most internal AI teams do not need another generic chat UI. They need a controlled runtime where models can use tools, inspect files, call approved integrations, run code, retrieve knowledge, and produce durable outputs.

Bionic provides that foundation. Your team keeps control of the models, infrastructure, integrations, governance rules, and business-specific workflows.

What Bionic Provides

  • AI workspace and conversation history
  • Model connectivity for hosted, private, and local models
  • RAG and dataset-backed knowledge
  • Built-in tool runtime
  • Sandboxed code and command execution
  • Virtual filesystem for uploads, datasets, skills, and generated outputs
  • Integrations exposed to the runtime
  • Skills for reusable domain workflows
  • Generated artifacts and canvases
  • Identity, teams, permissions, audit, and usage controls
  • Kubernetes-oriented deployment infrastructure

Yes. For a README I’d keep it this tight:

Agentic Runtime and Sandbox

Bionic gives the model a single run_bash tool and exposes its capabilities through a virtual filesystem.

/
├── memory.md
├── attachments/
│   ├── customer-list.xlsx
│   └── contract.pdf
├── conversations/
│   ├── today/
│   └── last-week/
├── datasets/
│   ├── customers.csv
│   └── sales.parquet
├── skills/
│   ├── crm/
│   │   ├── SKILL.md
│   │   └── operations/
│   │       └── crm_list_customers.md
│   ├── spreadsheets/
│   │   └── SKILL.md
│   └── pdf/
│       └── SKILL.md
├── artifacts/
│   ├── report.xlsx
│   └── analysis.pdf
└── tmp/
  • Memory — relevant context from previous work.
  • Attachments — files supplied by the user.
  • Conversations — previous conversations available when more context is needed.
  • Datasets — data available to the agent.
  • Skills — instructions and capabilities, including access to enterprise systems.
  • Artifacts — files produced by the agent.
  • tmp — ephemeral working files and generated code.

The filesystem is virtual, so resources can be discovered without copying everything into the sandbox or loading it into the model’s context.

This architecture is closely aligned with Everything is Context: Agentic File System Abstraction for Context Engineering, which proposes a Unix-inspired filesystem abstraction for exposing memory, knowledge and tools to agents. :chatgpt-content-reference{index=“0”}

This architecture is closely aligned with Everything is Context: Agentic File System Abstraction for Context Engineering, which proposes a Unix-inspired filesystem abstraction for exposing memory, knowledge and tools to agents.

For memory specifically, Filesystem-Based Memory for LLM Agents studies directory-tree memory accessed by agents using generic filesystem tools, and finds that organized filesystem memory can substantially reduce retrieval cost.

Architecture

Run Bionic

For local evaluation and small pilots, use the Docker Compose installation:

Try Bionic with Docker Compose

For production-style local testing, use Kubernetes:

Run Bionic on Kubernetes

For private cloud, on-premise, and air-gapped deployments, start with the production installation docs:

Bionic documentation

Extend Bionic

Bionic is designed for internal AI engineers and platform teams that need to connect real organisational systems.

  • Models: use approved hosted models, private inference endpoints, or local models.
  • Datasets: connect private documents and knowledge sources for grounded workflows.
  • Integrations: expose approved business systems to the runtime.
  • Skills: package instructions, templates, and repeatable domain workflows.
  • Tools: give models deterministic capabilities through the built-in tool runtime.
  • Outputs: persist generated files and artifacts for use in the chat experience.

Security and Control

Bionic is built for customer-controlled deployment environments:

  • Self-hosted infrastructure
  • SSO/OIDC integration
  • Team-based permissions
  • Audit trails
  • Usage controls
  • Postgres-backed persistence
  • Object storage for generated files and documents
  • Local, private, or hosted model support
  • Kubernetes deployment model

Commercial Support

Bionic is open source. Commercial support is available for organisations running it as critical internal infrastructure.

  • Community: free, open-source, self-hosted foundation.
  • Enterprise: production support, SLAs, security response, supported releases, architecture guidance, and upgrade assistance.
  • Deployment Accelerator: help deploying Bionic and delivering a first validated production workflow.

Explore deployment options or talk to us.

Contributing

Contributions are welcome. Start with CONTRIBUTING.md.

License

Bionic is licensed under the Apache License 2.0.

View this README on GitHub

推荐工具

换一个关键词,或者移除筛选条件。

安装

npx skillfish add bionic-gpt/bionic-gpt