LlamaIndex Vertex AI Embeddings

0.5.0 verified Fri May 01 auth: no python

LlamaIndex integration for Google Vertex AI embeddings. Version 0.5.0 supports Gemini models (text-embedding-004, text-embedding-005) and legacy Gecko models. Requires Python >=3.10 and a Google Cloud project with Vertex AI enabled. Release cadence: irregular, tied to llama-index core updates.

pip install llama-index-embeddings-vertex

Common errors

error google.auth.exceptions.DefaultCredentialsError: Could not automatically determine credentials. ↓

cause No valid Google Cloud credentials found in environment (missing GOOGLE_APPLICATION_CREDENTIALS or gcloud auth).

fix

Set GOOGLE_APPLICATION_CREDENTIALS to path to service account JSON, or run 'gcloud auth application-default login'.

error ImportError: cannot import name 'VertexEmbedding' from 'llama_index.embeddings.vertex' ↓

cause Outdated library version; import path changed in v0.2.0.

fix

Upgrade: pip install --upgrade llama-index-embeddings-vertex. If version <0.2.0, use old import: from llama_index.embeddings.vertex_ai import VertexEmbedding.

error ValueError: Model 'text-embedding-004' not found or not supported in location 'us-central1'. ↓

cause Specified model is not deployed or not available in the chosen GCP region.

fix

Verify model availability in your region: https://cloud.google.com/vertex-ai/generative-ai/docs/embeddings/get-text-embeddings#supported_models. Use location 'us-central1' for most models.

Warnings

breaking Version 0.2.0 moved VertexEmbedding from llama_index.embeddings.vertex_ai to llama_index.embeddings.vertex. Old import will break. ↓

fix Change 'from llama_index.embeddings.vertex_ai import VertexEmbedding' to 'from llama_index.embeddings.vertex import VertexEmbedding'.

deprecated Legacy Gecko models (textembedding-gecko@001, @002, @003) are deprecated and may be removed in future versions. Use text-embedding-004 or later. ↓

fix Set model_name='text-embedding-004' or 'text-embedding-005' (if available).

gotcha The 'project' parameter is required; omitting it defaults to the project set in gcloud config, which may be unexpected. ↓

fix Always pass 'project' explicitly to VertexEmbedding().

Imports

VertexEmbedding

wrong

from llama_index.embeddings.vertex_ai import VertexEmbedding

correct

from llama_index.embeddings.vertex import VertexEmbedding

Common mistake: using 'vertex_ai' instead of 'vertex'.

VertexEmbedding

wrong

from llama_index.embeddings.google import VertexEmbedding

correct

from llama_index.embeddings.vertex import VertexEmbedding

Incorrect module path: Google provider is separate.

Quickstart

Initialize VertexEmbedding with model name and project. Authentication via GOOGLE_APPLICATION_CREDENTIALS env var (service account) or default credentials.

import os
from llama_index.embeddings.vertex import VertexEmbedding

# Set credentials (recommended: service account JSON path)
os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = "path/to/service-account.json"

embed_model = VertexEmbedding(
    model_name="text-embedding-004",  # or "textembedding-gecko@003"
    project="your-gcp-project-id",
    location="us-central1",
)

embeddings = embed_model.get_text_embedding("Hello, world!")
print(len(embeddings))  # 768 for text-embedding-004