Skip to main content

OpenAI Integration

MemexLLM's OpenAI integration adds conversation memory to your existing OpenAI applications with minimal code changes.

Installation​

pip install memexllm[openai]

Basic Usage​

Here's how to enhance your OpenAI client with conversation memory using the FIFO algorithm:

from openai import OpenAI
from memexllm.integrations.openai import with_history
from memexllm.history import HistoryManager
from memexllm.storage import SQLiteStorage
from memexllm.algorithms import FIFOAlgorithm

# Your existing OpenAI code
client = OpenAI()

# Create history manager
history_manager = HistoryManager(
    storage=SQLiteStorage("chat.db"),
    algorithm=FIFOAlgorithm(max_messages=50)  # Keep last 50 messages
)

# Add conversation memory with FIFO algorithm
client = with_history(history_manager=history_manager)(client)

# Use the client as normal - conversation history is handled automatically
response = client.chat.completions.create(
    model="gpt-3.5-turbo",
    messages=[{"role": "user", "content": "Hello!"}],
    thread_id="my-thread"  # Optional, will be created if not provided
)

That's it! The only changes needed are:

  1. Import the with_history wrapper and FIFOAlgorithm
  2. Choose a storage backend
  3. Configure the FIFO algorithm
  4. Wrap your OpenAI client

Async Support​

The integration works the same way with async clients:

from openai import AsyncOpenAI
from memexllm.integrations.openai import with_history
from memexllm.storage import SQLiteStorage
from memexllm.algorithms import FIFOAlgorithm

# Your existing async OpenAI code
client = AsyncOpenAI()

# Add conversation memory with FIFO algorithm
client = with_history(
    storage=SQLiteStorage("chat.db"),
    algorithm=FIFOAlgorithm(max_messages=50)  # Keep last 50 messages
)(client)

# Use as normal
response = await client.chat.completions.create(
    model="gpt-3.5-turbo",
    messages=[{"role": "user", "content": "Hello!"}],
    thread_id="my-thread"
)

Next Steps​