Skip to content

BailingHub Documentation

Your business system keeps final authority. BailingHub governs discovery, approval, execution, and audit.

Core 0.9.0 with SDK/DSH 0.7.0: conversation models and generation tools share a USD allowance; periodic sale price and allowance are separate, as are image results and settlement. Existing business authorization, approvals and task controls remain in force. Read the changes and upgrade steps.

Start with the capability map

If entry points, routes, tools, knowledge, approvals, channels, and local runtimes are not yet clear, start with the Capability and configuration map.

I already have a business system

Use the SDK and an ACC capability declaration to expose one real action, such as reading an order or updating a staff profile.

I want a local agent

Use the independent DSH plugin for browser authorization, select accounts for the conversation, and invoke governed business tools. The MCP job entry is configured separately.

I want to verify the full path

Run the public Docker demo or online trial and inspect identity, routes, approvals, traces, and audit records.

  1. Declare one business action that an agent may use.
  2. Create a client and an allowlisted route in BailingHub.
  3. Trigger it from a backend, embedded chat, local agent, or automation platform.
  4. Inspect the tool call, approval, final result, and trace in the console.

Business and delivery teams

Start with the User Guide and turn one scenario into identity, capability, approval, and audit requirements.

Local-agent developers

Read Agent Client for browser authorization, local orchestration, and governance boundaries.

Topic What it answers
Capability and configuration map Find the entry point, route, model, tool, identity, approval, and audit settings for a goal
10-minute path Verify a read, a write, and an approval using deterministic demo data
User Guide Decide what to configure from a business scenario
Docker Demo Run the hub, MySQL, and Demo Business locally
Routes, models, and context Decide where a request goes, which model it uses, and which context it receives
Client API v1 Submit jobs, poll status, and receive signed callbacks
Business Tools Expose only ACC-declared business actions
Model services and plan billing Configure models, shared USD allowances, periodic plans and image tools; inspect original requests and settlement
Business SDKs Build specs, verify calls, sign tickets, and call the hub
Business adapters and community recipes Distinguish SDKs, independent adapters, and reference recipes
Accounts and conversation records Select accounts, inspect actions and synchronized text, and handle outages or missing history
Agent Client Let a locally authorized agent orchestrate governed business tools
Executor vs. Agent Client Separate a Hub-task worker from a local agent carrying a business identity
Widget and SSE Configure embedded chat, the Origin allowlist, default-open behavior, and streaming events
Media and rich output Handle images, audio, files, Markdown, charts, and safe forms
WeCom inbound channel Receive WeCom messages and map platform identity to business identity correctly
Knowledge Synchronize documents idempotently and retrieve them by route
Approval Intents Bind high-risk actions to frozen arguments and business decisions
Operations Self-host, migrate, monitor, and protect secrets
Architecture and fixed-version index Return to the authoritative v0.9.0 architecture, contracts, schemas, and operations references
MCP to safe execution Separate protocol, runtime, governance, and final business authority