xinyin025/LangBot
1
0
Fork 0
mirror of https://github.com/langbot-app/LangBot.git synced 2026-09-16 14:57:15 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat(runner): unify plugin execution across agents and event processors

Browse Source
This commit is contained in:
RockChinQ
2026-09-10 18:04:38 +08:00
parent 8903a40c41
commit f24a7c9bb2
223 changed files with 4091 additions and 3068 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/review/box-session-scope.md
+11 -11
View File
@@ -7,7 +7,7 @@
## 1. Decision
The LangBot Host owns the Box session used by an event run. A Pipeline, Agent,
or AgentRunner cannot choose a global, per-user, per-conversation, or per-query
or Runner cannot choose a global, per-user, per-conversation, or per-query
sandbox mode.
`BoxService.resolve_box_session_id(query)` always returns this shape:
@@ -81,7 +81,7 @@ Host scope or launcher/session identity is also rejected. There is no
## 3. Host execution Query
AgentRunner callbacks need a Host-owned Query view because model/tool loaders
Runner callbacks need a Host-owned Query view because model/tool loaders
already consume that type. The Query is internal and is never exposed as a
Runner-controlled object.
@@ -96,11 +96,11 @@ Runner-controlled object.
This gives Pipeline and pure EBA execution the same Host tool path without
inventing a fake Pipeline for an independent Agent.
## 4. AgentRunner callback paths
## 4. Runner callback paths
AgentRunner implementations may use either callback transport:
Runner implementations may use either callback transport:
1. SDK/Python runners call `AgentRunAPIProxy.call_tool`.
1. SDK/Python runners call `RunnerAPIProxy.call_tool`.
2. External harnesses call the SDK-owned scoped MCP bridge.
Both transports emit the same `PluginToRuntimeAction.CALL_TOOL`. The Host then
@@ -108,8 +108,8 @@ validates the same run authorization, restores the same execution Query, and
dispatches to the same ToolManager and BoxService.
```text
AgentRunner
+-- AgentRunAPIProxy.call_tool --------+
Runner
+-- RunnerAPIProxy.call_tool --------+
| |
+-- SDK-owned scoped MCP bridge -------+--> PluginToRuntimeAction.CALL_TOOL
--> run authorization
@@ -119,7 +119,7 @@ AgentRunner
--> lb-box-<sha256>
```
An AgentRunner is not required to use MCP. Local Python runners can use the SDK
An Runner is not required to use MCP. Local Python runners can use the SDK
directly; code-agent harnesses can use the bridge. The transports do not define
different authorization or sandbox semantics.
@@ -143,16 +143,16 @@ This is separate from the scoped MCP bridge above:
| Path | Purpose | Session rule |
| --- | --- | --- |
| AgentRunner scoped MCP bridge | Call authorized Host tools for one active run | Host-owned `lb-box-<sha256>` from the run execution Query |
| Runner scoped MCP bridge | Call authorized Host tools for one active run | Host-owned `lb-box-<sha256>` from the run execution Query |
| MCP-in-Box stdio server | Keep configured MCP server processes running | Dedicated persistent `mcp-shared` session |
Calling a sandbox tool through the AgentRunner bridge never redirects the run
Calling a sandbox tool through the Runner bridge never redirects the run
workspace into `mcp-shared`. Conversely, an MCP server's managed-process
lifecycle does not inherit the current event scope.
## 7. Configuration and compatibility
There is no Box session scope field in Pipeline metadata, AgentRunner config,
There is no Box session scope field in Pipeline metadata, Runner config,
or the public Pipeline/Runner API. Operators configure the Box subsystem itself
(`box.enabled`, backend/runtime settings, profiles, mount allowlists, quotas,
and workspace roots), not per-Runner session templates.