LlamaIndex LLMs Gemini

0.6.2 verified Fri May 01 auth: no python

LlamaIndex integration for Google Gemini LLMs. Provides a Gemini class that wraps the Google Generative AI SDK for use within LlamaIndex workflows. Current stable version is 0.6.2, released regularly alongside LlamaIndex ecosystem updates.

pip install llama-index-llms-gemini

Common errors

error ModuleNotFoundError: No module named 'llama_index.llms.gemini' ↓

cause The package is not installed or the import path is incorrect.

fix

Run pip install llama-index-llms-gemini (or switch to llama-index-llms-google) and use from llama_index.llms.gemini import Gemini.

error google.api_core.exceptions.PermissionDenied: 403 ↓

cause Invalid or missing API key.

fix

Ensure GOOGLE_API_KEY is set correctly. If using Vertex AI, check your GCP service account permissions.

error ValueError: Model name 'gemini-pro' is not valid ↓

cause Model name must include the 'models/' prefix.

fix

Use model='models/gemini-pro' instead of 'gemini-pro'.

Warnings

breaking In version 0.2.0+, the import path changed from `llama_index.llms.gemini` to `llama_index.llms.google` (for Google AI) or `llama_index.llms.vertex` (for Vertex AI). This package (`llama-index-llms-gemini`) is deprecated in favor of `llama-index-llms-google`. ↓

fix Switch to `pip install llama-index-llms-google` and use `from llama_index.llms.google import Gemini`.

deprecated The `llama-index-llms-gemini` package is deprecated. New projects should use `llama-index-llms-google`. ↓

fix Use `pip install llama-index-llms-google` and import from `llama_index.llms.google`.

gotcha Google API key must be set either via the `api_key` parameter or the `GOOGLE_API_KEY` environment variable. Not setting it results in an authentication error. ↓

fix Set `GOOGLE_API_KEY` environment variable or pass `api_key` to the Gemini constructor.

Imports

Gemini
wrong
```
from llama_index.llms import Gemini
```
correct
```
from llama_index.llms.gemini import Gemini
```
Gemini is no longer directly exposed from llama_index.llms; must import from the subpackage.

Quickstart

Initialize Gemini LLM with an API key and generate a completion.

import os
from llama_index.llms.gemini import Gemini

api_key = os.environ.get("GOOGLE_API_KEY", "")
llm = Gemini(api_key=api_key, model="models/gemini-pro")
resp = llm.complete("What is the capital of France?")
print(resp)