Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
Python agentagentic-aiagentic-frameworkagentic-workflowaiautomationclaudedeepseekgenaigptllmlow-codemcpnextjsno-codeopenaipythonskillsworkflow
"""Client-safe shell environment helpers for Agent Stub forwarding.
Only user-visible ``shell.run`` commands receive these variables. Internal
lifecycle commands remain free of Agent Stub credentials so workspace setup
and cleanup cannot accidentally inherit user-facing forwarding state. The module stays server-extra-free because the
shell runtime and provider factory use it in sandbox-visible paths.
"""
from __future__ import annotations
from typing import Protocol
from dify_agent.agent_stub.protocol.agent_stub import (
AGENT_STUB_API_BASE_URL_ENV_VAR,
AGENT_STUB_AUTH_JWE_ENV_VAR,
normalize_agent_stub_api_base_url,
)
from dify_agent.layers.execution_context import DifyExecutionContextLayerConfig
class ShellAgentStubTokenFactory(Protocol):
"""Callable boundary for server-side Agent Stub token issuance."""
def __call__(self, execution_context: DifyExecutionContextLayerConfig, *, session_id: str | None) -> str: ...
def build_shell_agent_stub_env(
*,
agent_stub_api_base_url: str | None,
execution_context: DifyExecutionContextLayerConfig | None,
token_factory: ShellAgentStubTokenFactory | None,
session_id: str | None,
) -> dict[str, str] | None:
"""Build the shell-visible Agent Stub environment for one user command."""
if agent_stub_api_base_url is None or execution_context is None or token_factory is None:
return None
return {
AGENT_STUB_API_BASE_URL_ENV_VAR: normalize_agent_stub_api_base_url(agent_stub_api_base_url),
AGENT_STUB_AUTH_JWE_ENV_VAR: token_factory(execution_context, session_id=session_id),
}
__all__ = [
"AGENT_STUB_AUTH_JWE_ENV_VAR",
"AGENT_STUB_API_BASE_URL_ENV_VAR",
"ShellAgentStubTokenFactory",
"build_shell_agent_stub_env",
]