Google Gemini

Memori integrates with Google Gemini via the google-genai SDK. Register the GenerativeModel instance and all generate_content() calls are automatically captured.

Quick Start

Gemini Integration

import os
from memori import Memori
import google.generativeai as genai

genai.configure(api_key=os.getenv("GOOGLE_API_KEY"))
client = genai.GenerativeModel("gemini-2.0-flash-exp")  # Use any Gemini model

mem = Memori().llm.register(client)
mem.attribution(entity_id="user_123", process_id="gemini_assistant")

response = client.generate_content("Hello!")
print(response.text)

Supported Modes

Mode	Python	TypeScript
Sync	`client.generate_content()`	—
Async	`await client.generate_content_async()`	`await client.models.generateContent()`
Streamed	`stream=True` parameter	`client.models.generateContentStream()`

Additional Modes

Async (Python)

import os, asyncio
from memori import Memori
import google.generativeai as genai

genai.configure(api_key=os.getenv("GOOGLE_API_KEY"))
client = genai.GenerativeModel("gemini-2.0-flash-exp")  # Use any Gemini model

mem = Memori().llm.register(client)
mem.attribution(entity_id="user_123", process_id="gemini_assistant")

async def main():
    response = await client.generate_content_async("Hello!")
    print(response.text)

asyncio.run(main())

Streaming

import os
from memori import Memori
import google.generativeai as genai

genai.configure(api_key=os.getenv("GOOGLE_API_KEY"))
client = genai.GenerativeModel("gemini-2.0-flash-exp")  # Use any Gemini model

mem = Memori().llm.register(client)
mem.attribution(entity_id="user_123", process_id="gemini_assistant")

response = client.generate_content("Hello!", stream=True)
for chunk in response:
    print(chunk.text, end="")

Multi-Turn Conversations

Use start_chat() for multi-turn interactions. Memori tracks the full conversation automatically.

chat = client.start_chat()
response = chat.send_message("My name is Alice.")
print(response.text)

response = chat.send_message("What's my name?")
print(response.text)

TypeScript

Multi-turn conversations in TypeScript use the standard messages array pattern:

import { GoogleGenAI } from '@google/genai';
import { Memori } from '@memorilabs/memori';

const client = new GoogleGenAI({ apiKey: process.env.GOOGLE_API_KEY });

const mem = new Memori().llm.register(client);
mem.attribution('user_123', 'gemini_assistant');

const response = await client.models.generateContent({
  model: 'gemini-2.0-flash',
  contents: [
    { role: 'user', parts: [{ text: 'My name is Alice.' }] },
  ],
});
console.log(response.text);

const response2 = await client.models.generateContent({
  model: 'gemini-2.0-flash',
  contents: [
    { role: 'user', parts: [{ text: 'My name is Alice.' }] },
    { role: 'model', parts: [{ text: response.text ?? '' }] },
    { role: 'user', parts: [{ text: "What's my name?" }] },
  ],
});
console.log(response2.text);