LangChain Plugin: Agent

Overview

FlotorchLangChainAgent builds a LangChain agent from the FloTorch Console config, supports custom tools and optional memory signals in the prompt. It returns a LangChain agent (Runnable) suitable for use with AgentExecutor.

---

Setup

from flotorch.langchain.agent import FlotorchLangChainAgent

API_KEY = "<your_api_key>"
BASE_URL = "https://gateway.flotorch.cloud"
AGENT_NAME = "<your_agent_name>"  # agent name configured in console

---

API

FlotorchLangChainAgent

FlotorchLangChainAgent(
  agent_name: str,  # agent name configured in console
  enable_memory: bool = False,
  custom_tools: list | None = None,
  base_url: str | None = None,
  api_key: str | None = None,
)

get_agent() -> Runnable

Returns a FloTorch agent (LangChain-compatible) with optional config sync.

get_tools() -> list

Returns the active tool list (auto-syncing via ToolsProxy).

---

Usage Example

flotorch_client = FlotorchLangChainAgent(
    api_key=API_KEY,
    enable_memory=True,
    base_url=BASE_URL,
    agent_name=AGENT_NAME,
    custom_tools=tools,
)

agent = flotorch_client.get_agent()
all_tools = flotorch_client.get_tools()

from langchain.agents import AgentExecutor
executor = AgentExecutor(agent=agent, tools=all_tools, memory=memory_services, verbose=False)

---

Notes

• When enable_memory=True, the prompt injects placeholders for session and long-term summaries
• Tools include custom tools and MCP tools from Console config