CrewAI Plugin: Memory

Overview

FloTorch integrates with CrewAI’s dual-memory pattern:

• Short-term conversation context via ShortTermMemory backed by FlotorchCrewAISession storage
• Long-term, persistent external memory via ExternalMemory backed by FlotorchMemoryStorage

---

Setup

from flotorch.crewai.sessions import FlotorchCrewAISession
from flotorch.crewai.memory import FlotorchMemoryStorage
from crewai.memory.short_term.short_term_memory import ShortTermMemory
from crewai.memory.external.external_memory import ExternalMemory

API_KEY = "<your_api_key>"
BASE_URL = "https://gateway.flotorch.cloud"
PROVIDER = "<flotorch_memory_provider>"  # configured in FloTorch Console
USER_ID = "user_123"
APP_ID = "app_123"

---

Memory setup

Short-term memory (conversation context)

short_term_storage = FlotorchCrewAISession(
  api_key=API_KEY,
  base_url=BASE_URL,
)

short_term_memory = ShortTermMemory(storage=short_term_storage)

External memory (long-term persistence)

external_storage = FlotorchMemoryStorage(
  api_key=API_KEY,
  base_url=BASE_URL,
  name=PROVIDER,
  user_id=USER_ID,
  app_id=APP_ID,
)

external_memory = ExternalMemory(storage=external_storage)

---

Features

CrewAI Integration

• Uses CrewAI-native ShortTermMemory and ExternalMemory
• Short-term storage backed by FloTorch Sessions; external storage backed by FloTorch Memory
• Automatic content extraction and robust error handling

---

Usage Example

from crewai import Crew

crew = Crew(
  agents=[agent],
  tasks=[task],
  short_term_memory=short_term_memory,
  external_memory=external_memory,
  verbose=True,
)

---

Notes

• Requires a configured Memory Provider in FloTorch Console
• Use short_term_memory and external_memory parameters on Crew
• Ensure user_id and app_id are set for external memory attribution