You've already forked adk-python
mirror of
https://github.com/encounter/adk-python.git
synced 2026-07-09 18:19:28 -07:00
1589fcdd86
mainly because http://github.com/robots.txt disallows `/*/raw/` path. using GCS HTTP URIs is more reliable with Gemini model. PiperOrigin-RevId: 811409688
131 lines
5.2 KiB
Markdown
131 lines
5.2 KiB
Markdown
# Static Non-Text Content Sample Agent
|
|
|
|
This sample demonstrates ADK's static instruction feature with non-text content (images and files).
|
|
|
|
## Features Demonstrated
|
|
|
|
- **Static instructions with mixed content**: Text, images, and file references in a single static instruction
|
|
- **Reference ID generation**: Non-text parts are automatically given reference IDs (`inline_data_0`, `file_data_1`, etc.)
|
|
- **Gemini Files API integration**: Demonstrates uploading documents and using file_data
|
|
- **Mixed content types**: inline_data for images, file_data for documents
|
|
- **API variant detection**: Different behavior for Gemini API vs Vertex AI
|
|
- **GCS file references**: Support for both GCS URI and HTTPS URL access methods in Vertex AI
|
|
|
|
## Static Instruction Content
|
|
|
|
The agent includes:
|
|
|
|
1. **Text instructions**: Guide the agent on how to behave
|
|
2. **Sample image**: A 1x1 yellow pixel PNG (`sample_chart.png`) as inline binary data
|
|
|
|
**Gemini Developer API:**
|
|
3. **Contributing guide**: A sample document uploaded to Gemini Files API and referenced via file_data
|
|
|
|
**Vertex AI:**
|
|
3. **Research paper**: Gemma research paper from Google Cloud Storage via GCS file reference
|
|
4. **AI research paper**: Same research paper accessed via HTTPS URL for comparison
|
|
|
|
## Content Used
|
|
|
|
**All API variants:**
|
|
- **Image**: Base64-encoded 1x1 yellow pixel PNG (embedded in code as `inline_data`)
|
|
|
|
**Gemini Developer API:**
|
|
- **Document**: Sample contributing guide text (uploaded to Gemini Files API as `file_data`)
|
|
- Contains sample guidelines and best practices for development
|
|
- Demonstrates Files API upload and file_data reference functionality
|
|
- Files are automatically cleaned up after 48 hours by the Gemini API
|
|
|
|
**Vertex AI:**
|
|
- **Gemma Research Paper**: Research paper accessed via GCS URI (as `file_data`)
|
|
- GCS URI: `gs://cloud-samples-data/generative-ai/pdf/2403.05530.pdf`
|
|
- Demonstrates native GCS file access in Vertex AI
|
|
- PDF format with technical AI research content about Gemini 1.5
|
|
- **AI Research Paper**: Same research paper accessed via HTTPS URL (as `file_data`)
|
|
- HTTPS URL: `https://storage.googleapis.com/cloud-samples-data/generative-ai/pdf/2403.05530.pdf`
|
|
- Demonstrates HTTPS file access in Vertex AI
|
|
- Agent can discover these are the same document and compare access methods
|
|
|
|
## Setup
|
|
|
|
### Setup API Credentials
|
|
|
|
Create a `.env` file in the project root with your API credentials:
|
|
|
|
```bash
|
|
# Choose Model Backend: 0 -> ML Dev, 1 -> Vertex
|
|
GOOGLE_GENAI_USE_VERTEXAI=1
|
|
|
|
# ML Dev backend config
|
|
GOOGLE_API_KEY=your_google_api_key_here
|
|
|
|
# Vertex backend config
|
|
GOOGLE_CLOUD_PROJECT=your_project_id
|
|
GOOGLE_CLOUD_LOCATION=us-central1
|
|
```
|
|
|
|
The agent will automatically load environment variables on startup.
|
|
|
|
## Usage
|
|
|
|
### Default Test Prompts (Recommended)
|
|
```bash
|
|
cd contributing/samples
|
|
python -m static_non_text_content.main
|
|
```
|
|
This runs test prompts that demonstrate the static content features:
|
|
- **Gemini Developer API**: 4 prompts testing inline_data + Files API upload
|
|
- **Vertex AI**: 5 prompts testing inline_data + GCS/HTTPS file access comparison
|
|
|
|
### Interactive Mode
|
|
```bash
|
|
cd contributing/samples
|
|
adk run static_non_text_content
|
|
```
|
|
Use ADK's built-in interactive mode for free-form conversation.
|
|
|
|
### Single Prompt
|
|
```bash
|
|
cd contributing/samples
|
|
python -m static_non_text_content.main --prompt "What reference materials do you have access to?"
|
|
```
|
|
|
|
### With Debug Logging
|
|
```bash
|
|
cd contributing/samples
|
|
python -m static_non_text_content.main --debug --prompt "What is the Gemma research paper about?"
|
|
```
|
|
|
|
## Default Test Prompts
|
|
|
|
The sample automatically runs test prompts when no `--prompt` is specified:
|
|
|
|
**All API variants:**
|
|
1. "What reference materials do you have access to?"
|
|
2. "Can you describe the sample chart that was provided to you?"
|
|
3. "How do the inline image and file references in your instructions help you answer questions?"
|
|
|
|
**Gemini Developer API only:**
|
|
4. "What does the contributing guide document say about best practices?"
|
|
|
|
**Vertex AI only (additional prompts):**
|
|
5. "What is the Gemma research paper about and what are its key contributions?"
|
|
6. "Can you compare the research papers you have access to? Are they related or different?"
|
|
|
|
**Gemini Developer API** tests: `inline_data` (image) + Files API `file_data` (uploaded document)
|
|
**Vertex AI** tests: `inline_data` (image) + GCS URI `file_data` + HTTPS URL `file_data` (same document via different access methods)
|
|
|
|
## How It Works
|
|
|
|
1. **Static Instruction Processing**: The `static_instruction` content is processed during agent initialization
|
|
2. **Reference Generation**: Non-text parts get references like `[Reference to inline binary data: inline_data_0 ('sample_chart.png', type: image/png)]` in the system instruction
|
|
3. **User Content Creation**: The actual binary data/file references are moved to user contents with proper role attribution
|
|
4. **Model Understanding**: The model receives both the descriptive references and the actual content for analysis
|
|
|
|
## Code Structure
|
|
|
|
- `agent.py`: Defines the agent with static instruction containing mixed content
|
|
- `main.py`: Runnable script with interactive and single-prompt modes
|
|
- `__init__.py`: Package initialization following ADK conventions
|
|
|
|
This sample serves as a test case for the static instruction with non-text parts feature using both `inline_data` and `file_data`. |