Skip to content

Document understanding

To see an example of document understanding, run the "Document Processing with Gemini" Jupyter notebook in one of the following environments:

Open in Colab | Open in Colab Enterprise | Open in Vertex AI Workbench user-managed notebooks | View on GitHub

You can add documents (PDF and TXT files) to Gemini requests to perform tasks that involve understanding the contents of the included documents. This page shows you how to add PDFs to your requests to Gemini in Vertex AI by using the Google Cloud console and the Vertex AI API.

Supported models

The following table lists the models that support document understanding:

Model Media details MIME types
Gemini 2.5 Pro - Maximum number of files per prompt: 3,000 - Maximum number of pages per file: 1,000 - Maximum file size per file: 50 MB - application/pdf - text/plain
Gemini 2.5 Flash - Maximum number of files per prompt: 3,000 - Maximum number of pages per file: 1,000 - Maximum file size per file for the API or Cloud Storage imports: 50 MB - Maximum file size per file for direct uploads through the console: 7 MB - application/pdf - text/plain
Gemini 2.0 Flash - Maximum number of files per prompt: 3,000 - Maximum number of pages per file: 1,000 - Maximum file size per file: 50 MB - Maximum tokens per minute (TPM) per project1: - US/Asia: 3.4 M - EU: 3.4 M - application/pdf - text/plain
Gemini 2.0 Flash-Lite - Maximum number of files per prompt: 3,000 - Maximum number of pages per file: 1,000 - Maximum file size per file: 50 MB - Maximum tokens per minute (TPM): - US/Asia: 3.4 M - EU: 3.4 M - application/pdf - text/plain

1This is the maximum TPM from document inputs across all requests of a project. Also use the maximum TPM for other modalities.

The quota metric is generate_content_document_input_per_base_model_id_and_resolution.

For a list of languages supported by Gemini models, see model information Google models. To learn more about how to design multimodal prompts, see Design multimodal prompts. If you're looking for a way to use Gemini directly from your mobile and web apps, see the Vertex AI in Firebase SDKs for Android, Swift, web, and Flutter apps.

Add documents to a request

The following code sample shows you how to include a PDF in a prompt request. This PDF sample works with all Gemini multimodal models.

Gen AI SDK for Python

Install

pip install --upgrade google-genai

To learn more, see the SDK reference documentation.

Set environment variables to use the Gen AI SDK with Vertex AI:

# Replace the `GOOGLE_CLOUD_PROJECT` and `GOOGLE_CLOUD_LOCATION` values
# with appropriate values for your project.
export GOOGLE_CLOUD_PROJECT=GOOGLE_CLOUD_PROJECT
export GOOGLE_CLOUD_LOCATION=global
export GOOGLE_GENAI_USE_VERTEXAI=True

from google import genai
from google.genai.types import HttpOptions, Part

client = genai.Client(http_options=HttpOptions(api_version="v1"))
model_id = "gemini-2.0-flash-001"

prompt = """
You are a highly skilled document summarization specialist.
Your task is to provide a concise executive summary of no more than 300 words.
Please summarize the given document for a general audience.
"""

pdf_file = Part.from_uri(
 file_uri="gs://cloud-samples-data/generative-ai/pdf/1706.03762v7.pdf",
 mime_type="application/pdf",
)

response = client.models.generate_content(
 model=model_id,
 contents=[pdf_file, prompt],
)

print(response.text)
# Example response:
# Here is a summary of the document in 300 words.
#
# The paper introduces the Transformer, a novel neural network architecture for
# sequence transduction tasks like machine translation. Unlike existing models that rely on recurrent or
# convolutional layers, the Transformer is based entirely on attention mechanisms.
# ...

REST

After you set up your environment, you can use REST to test a text prompt. The following sample sends a request to the publisher model endpoint.

Before using any of the request data, make the following replacements:

  • PROJECT_ID: Your project ID.
  • FILE_URI: The URI or URL of the file to include in the prompt. Acceptable values include the following:
  • Cloud Storage bucket URI: The object must either be publicly readable or reside in the same Google Cloud project that's sending the request. For gemini-2.0-flash and gemini-2.0-flash-lite, the size limit is 2 GB.
  • HTTP URL: The file URL must be publicly readable. You can specify one video file, one audio file, and up to 10 image files per request. Audio files, video files, and documents can't exceed 15 MB.
  • YouTube video URL:The YouTube video must be either owned by the account that you used to sign in to the Google Cloud console or is public. Only one YouTube video URL is supported per request.

When specifying a fileURI, you must also specify the media type (mimeType) of the file. If VPC Service Controls is enabled, specifying a media file URL for fileURI is not supported.

If you don't have a PDF file in Cloud Storage, then you can use the following publicly available file: gs://cloud-samples-data/generative-ai/pdf/2403.05530.pdf with a mime type of application/pdf. To view this PDF, open the sample PDF file. - MIME_TYPE: The media type of the file specified in the data or fileUri fields. Acceptable values include the following:

Click to expand MIME types

  • application/pdf
  • audio/mpeg
  • audio/mp3
  • audio/wav
  • image/png
  • image/jpeg
  • image/webp
  • text/plain
  • video/mov
  • video/mpeg
  • video/mp4
  • video/mpg
  • video/avi
  • video/wmv
  • video/mpegps
  • video/flv
  • TEXT: The text instructions to include in the prompt. For example, You are a very professional document summarization specialist. Please summarize the given document.

To send your request, choose one of these options:

curl

Note: The following command assumes that you have logged in to the gcloud CLI with your user account by running gcloud init or gcloud auth login , or by using Cloud Shell, which automatically logs you into the gcloud CLI . You can check the currently active account by running gcloud auth list.

Save the request body in a file named request.json. Run the following command in the terminal to create or overwrite this file in the current directory:

cat > request.json << 'EOF'
{
 "contents": {
 "role": "USER",
 "parts": [
 {
 "fileData": {
 "fileUri": "FILE_URI",
 "mimeType": "MIME_TYPE"
 }
 },
 {
 "text": "TEXT"
 }
 ]
 }
}
EOF

Then execute the following command to send your REST request:

curl -X POST \ 
 -H "Authorization: Bearer $(gcloud auth print-access-token)" \ 
 -H "Content-Type: application/json; charset=utf-8" \ 
 -d @request.json \ 
 "https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/global/publishers/google/models/gemini-2.0-flash:generateContent"

PowerShell

Note: The following command assumes that you have logged in to the gcloud CLI with your user account by running gcloud init or gcloud auth login . You can check the currently active account by running gcloud auth list.

Save the request body in a file named request.json. Run the following command in the terminal to create or overwrite this file in the current directory:

@'
{
 "contents": {
 "role": "USER",
 "parts": [
 {
 "fileData": {
 "fileUri": "FILE_URI",
 "mimeType": "MIME_TYPE"
 }
 },
 {
 "text": "TEXT"
 }
 ]
 }
}
'@ | Out-File -FilePath request.json -Encoding utf8

Then execute the following command to send your REST request:

$cred = gcloud auth print-access-token 
$headers = @{ "Authorization" = "Bearer $cred" } 

Invoke-WebRequest ` 
 -Method POST ` 
 -Headers $headers ` 
 -ContentType: "application/json; charset=utf-8" ` 
 -InFile request.json ` 
 -Uri "https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/global/publishers/google/models/gemini-2.0-flash:generateContent" | Select-Object -Expand Content

You should receive a JSON response similar to the following.

Response

{
 "candidates": [
 {
 "content": {
 "role": "model",
 "parts": [
 {
 "text": "This report presents Gemini 2.0 Pro.\n"
 }
 ]
 },
 "finishReason": "STOP",
 "safetyRatings": [
 {
 "category": "HARM_CATEGORY_HATE_SPEECH",
 "probability": "NEGLIGIBLE",
 "probabilityScore": 0.13273923,
 "severity": "HARM_SEVERITY_NEGLIGIBLE",
 "severityScore": 0.08819004
 },
 {
 "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
 "probability": "NEGLIGIBLE",
 "probabilityScore": 0.1046602,
 "severity": "HARM_SEVERITY_NEGLIGIBLE",
 "severityScore": 0.0996453
 },
 {
 "category": "HARM_CATEGORY_HARASSMENT",
 "probability": "NEGLIGIBLE",
 "probabilityScore": 0.15987214,
 "severity": "HARM_SEVERITY_NEGLIGIBLE",
 "severityScore": 0.098946586
 },
 {
 "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
 "probability": "NEGLIGIBLE",
 "probabilityScore": 0.056966383,
 "severity": "HARM_SEVERITY_NEGLIGIBLE",
 "severityScore": 0.075721376
 }
 ]
 }
 ],
 "usageMetadata": {
 "promptTokenCount": 19882,
 "candidatesTokenCount": 336,
 "totalTokenCount": 20218
 }
}

Note the following in the URL for this sample:

  • Use the generateContent method to request that the response is returned after it's fully generated. To reduce the perception of latency to a human audience, stream the response as it's being generated by using the streamGenerateContent method.
  • The multimodal model ID is located at the end of the URL before the method (for example, gemini-2.0-flash). This sample might support other models as well.

Console

To send a multimodal prompt by using the Google Cloud console, do the following:

  1. In the Vertex AI section of the Google Cloud console, go to the Vertex AI Studio page.

Go to Vertex AI Studio 2. Click Create prompt. 3. Optional: Configure the model and parameters:

  • Model: Select a model.
  • Optional: To configure advanced parameters, click Advanced and configure as follows:

#### Click to expand advanced configurations

  • Top-K: Use the slider or textbox to enter a value for top-K.

Top-K changes how the model selects tokens for output. A top-K of 1 means the next selected token is the most probable among all tokens in the model's vocabulary (also called greedy decoding), while a top-K of 3 means that the next token is selected from among the three most probable tokens by using temperature.

For each token selection step, the top-K tokens with the highest probabilities are sampled. Then tokens are further filtered based on top-P with the final token selected using temperature sampling.

Specify a lower value for less random responses and a higher value for more random responses. - Top-P: Use the slider or textbox to enter a value for top-P. Tokens are selected from most probable to the least until the sum of their probabilities equals the value of top-P. For the least variable results, set top-P to 0. - Max responses: Use the slider or textbox to enter a value for the number of responses to generate. - Streaming responses: Enable to print responses as they're generated. - Safety filter threshold: Select the threshold of how likely you are to see responses that could be harmful. - Enable Grounding: Grounding isn't supported for multimodal prompts. - Region: Select the region that you want to use. - Temperature: Use the slider or textbox to enter a value for temperature.

The temperature is used for sampling during response generation, which occurs when topP
and topK are applied. Temperature controls the degree of randomness in token selection.
Lower temperatures are good for prompts that require a less open-ended or creative response, while
higher temperatures can lead to more diverse or creative results. A temperature of 0
means that the highest probability tokens are always selected. In this case, responses for a given
prompt are mostly deterministic, but a small amount of variation is still possible.

If the model returns a response that's too generic, too short, or the model gives a fallback
response, try increasing the temperature.

</li>
<li>**Output token limit**: Use the slider or textbox to enter a value for
the max output limit.

Maximum number of tokens that can be generated in the response. A token is
approximately four characters. 100 tokens correspond to roughly 60-80 words.

Specify a lower value for shorter responses and a higher value for potentially longer
responses.

</li>
<li>**Add stop sequence**: Optional. Enter a stop sequence, which is a
series of characters that includes spaces. If the model encounters a
stop sequence, the response generation stops. The stop sequence isn't
included in the response, and you can add up to five stop sequences.</li>
</ul>
5. Click Insert Media, and select a source for your file.

### Upload

Select the file that you want to upload and click Open.

### By URL

Enter the URL of the file that you want to use and click Insert.

### Cloud Storage

Select the bucket and then the file from the bucket that you want to import and click Select.

### Google Drive

  1. Choose an account and give consent to Vertex AI Studio to access your account the first time you select this option. You can upload multiple files that have a total size of up to 10 MB. A single file can't exceed 7 MB.
  2. Click the file that you want to add.
  3. Click Select.

The file thumbnail displays in the Prompt pane. The total number of tokens also displays. If your prompt data exceeds the token limit, the tokens are truncated and aren't included in processing your data. 6. Enter your text prompt in the Prompt pane. 7. Optional: To view the Token ID to text and Token IDs, click the tokens count in the Prompt pane.

Note: Media tokens aren't supported. 8. Click Submit. 9. Optional: To save your prompt to My prompts, click save_alt Save. 10. Optional: To get the Python code or a curl command for your prompt, click code Build with code > Get code.

Set optional model parameters

Each model has a set of optional parameters that you can set. For more information, see Content generation parameters.

Document requirements

PDF tokenization

PDFs are treated as images, so each page of a PDF is tokenized in the same way as an image.

Also, the cost for PDFs follows Gemini image pricing. For example, if you include a two-page PDF in a Gemini API call, you incur an input fee of processing two images.

PDF best practices

When using PDFs, use the following best practices and information for the best results:

  • If your prompt contains a single PDF, place the PDF before the text prompt in your request.
  • If you have a long document, consider splitting it into multiple PDFs to process it.
  • Use PDFs created with text rendered as text instead of using text in scanned images. This format ensures text is machine-readable so that it's easier for the model to edit, search, and manipulate compared to scanned image PDFs. This practice provides optimal results when working with text-heavy documents like contracts.

Limitations

While Gemini multimodal models are powerful in many multimodal use cases, it's important to understand the limitations of the models:

  • Spatial reasoning: The models aren't precise at locating text or objects in PDFs. They might only return the approximated counts of objects.
  • Accuracy: The models might hallucinate when interpreting handwritten text in PDF documents.

What's next