Google Gemini Integration

Automatic tracing for Google Generative AI (Gemini) API calls.

Quick Start

1
2
3
4
5
6
7
8
9
10
11
12
import tracium
import google.generativeai as genai
 
# Enable auto-instrumentation
tracium.trace()
 
# Configure Google AI
genai.configure(api_key="your-api-key")
 
# Use Gemini normally - all calls are traced
model = genai.GenerativeModel("gemini-pro")
response = model.generate_content("Explain AI in simple terms.")

What Gets Captured

Input - Prompts and content sent to Gemini
Model - gemini-pro, gemini-pro-vision, etc.
Output - Generated response
Token usage - Prompt and completion tokens
Safety ratings - Content safety scores
Latency - API call duration

Chat Sessions

1
2
3
4
5
6
7
8
9
10
11
12
import tracium
import google.generativeai as genai
 
tracium.trace()
genai.configure(api_key="your-api-key")
 
model = genai.GenerativeModel("gemini-pro")
chat = model.start_chat(history=[])
 
# Each message in the chat is traced
response = chat.send_message("Hi! Tell me about yourself.")
response = chat.send_message("What can you help me with?")

Multimodal (Vision)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
import tracium
import google.generativeai as genai
from PIL import Image
 
tracium.trace()
genai.configure(api_key="your-api-key")
 
model = genai.GenerativeModel("gemini-pro-vision")
 
# Load an image
image = Image.open("image.jpg")
 
# Multimodal calls are traced
response = model.generate_content([
    "What's in this image?",
    image
])

Streaming

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
import tracium
import google.generativeai as genai
 
tracium.trace()
genai.configure(api_key="your-api-key")
 
model = genai.GenerativeModel("gemini-pro")
 
# Streaming is supported
response = model.generate_content(
    "Write a long story about space exploration.",
    stream=True
)
 
for chunk in response:
    print(chunk.text, end="")

← Anthropic LangChain →