Skip to content

[docs][integration] Document Gemini chat model integration #898

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
section9-lab wants to merge 2 commits into
base: main
Choose a base branch
from
+98 −1
Open

[docs][integration] Document Gemini chat model integration #898

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8a24590
[docs][integration] Document Gemini chat model integration
section9-lab Jul 14, 2026
78f8fbb
[docs][integration] Mark Vertex AI support as experimental
section9-lab Jul 22, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
98 changes: 97 additions & 1 deletion docs/content/docs/development/chat_models.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -643,6 +643,102 @@ Some popular options include:
Model availability depends on your Azure region and subscription. Always check the official Azure OpenAI documentation for regional availability before implementing in production.
{{< /hint >}}

### Gemini

Google Gemini provides cloud-based chat models through the Gemini Developer API and Vertex AI. The Flink Agents Gemini integration uses the official Google Gen AI SDK and supports text conversations, system instructions, and tool calling.
Comment thread
section9-lab marked this conversation as resolved.

{{< hint warning >}}
Vertex AI support is experimental. The connection path has been smoke-tested during construction but has not yet been verified end to end.
{{< /hint >}}

{{< hint info >}}
Gemini is only supported in Java currently. To use Gemini from Python agents, see [Using Cross-Language Providers](#using-cross-language-providers).
{{< /hint >}}

#### Prerequisites

1. For the Gemini Developer API, create an API key in [Google AI Studio](https://aistudio.google.com/app/apikey)
2. For Vertex AI, enable Vertex AI in your Google Cloud project and configure Google Cloud credentials

#### GeminiChatModelConnection Parameters

{{< tabs "GeminiChatModelConnection Parameters" >}}

{{< tab "Java" >}}

| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `api_key` | String | Required unless `base_url` is set or `vertex_ai` is `true` | Gemini Developer API key |
| `base_url` | String | None | Custom endpoint, such as a proxy that injects credentials |
| `model` | String | None | Default model name, used when no model is supplied per setup |
| `timeout` | int | None | API request timeout in seconds |
| `vertex_ai` | boolean | `false` | Use the experimental Vertex AI backend; not yet verified end to end |
| `project` | String | None | Vertex AI project id |
| `location` | String | None | Vertex AI location |

{{< /tab >}}

{{< /tabs >}}

#### GeminiChatModelSetup Parameters

{{< tabs "GeminiChatModelSetup Parameters" >}}

{{< tab "Java" >}}

| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `connection` | String | Required | Reference to connection method name |
| `model` | String | `"gemini-3.1-pro-preview"` | Name of the chat model to use |
| `prompt` | Prompt \| String | None | Prompt template or reference to prompt resource |
| `tools` | List<String> | None | List of tool names available to the model |
| `temperature` | double | `0.1` | Sampling temperature (0.0 to 2.0) |
| `max_output_tokens` | long | `1024` | Maximum number of tokens to generate |
| `additional_kwargs` | Map<String, Object> | `{}` | Additional Gemini parameters (`top_k`, `top_p`, `stop_sequences`) |

{{< /tab >}}

{{< /tabs >}}

#### Usage Example

{{< tabs "Gemini Usage Example" >}}

{{< tab "Java" >}}
```java
public class MyAgent extends Agent {
@ChatModelConnection
public static ResourceDescriptor geminiConnection() {
return ResourceDescriptor.Builder.newBuilder(ResourceName.ChatModel.GEMINI_CONNECTION)
.addInitialArgument("api_key", System.getenv("GEMINI_API_KEY"))
.build();
}

@ChatModelSetup
public static ResourceDescriptor geminiChatModel() {
return ResourceDescriptor.Builder.newBuilder(ResourceName.ChatModel.GEMINI_SETUP)
.addInitialArgument("connection", "geminiConnection")
.addInitialArgument("model", "gemini-3.1-pro-preview")
Comment thread
section9-lab marked this conversation as resolved.
.addInitialArgument("temperature", 0.1d)
.addInitialArgument("max_output_tokens", 1024)
.build();
}

...
}
```
{{< /tab >}}

{{< /tabs >}}

#### Available Models

Visit the [Gemini models documentation](https://ai.google.dev/gemini-api/docs/models) for the complete and up-to-date list of available models.

{{< hint warning >}}
Model availability and names may differ between the Gemini Developer API and Vertex AI. Always check the official Gemini documentation before implementing in production.
{{< /hint >}}

### Ollama

Ollama provides local chat models that run on your machine, offering privacy, control, and no API costs.
Expand Down Expand Up @@ -1370,4 +1466,4 @@ public class MyChatModelSetup extends BaseChatModelSetup {

The built-in `chat_model_action` listens to `ChatRequestEvent` and `ToolResponseEvent`. To request a chat completion, send a `ChatRequestEvent`. If the model returns a final answer, the action sends a `ChatResponseEvent`.

If the model asks to call tools, `chat_model_action` sends a `ToolRequestEvent` instead of a final `ChatResponseEvent`. After the tools finish, it receives the matching `ToolResponseEvent`, appends the tool results to the chat history, and calls the model again. This loop continues until the model returns a final response. For details on how tools are executed, see [Built-in Events and Actions in Tool Use]({{< ref "docs/development/tool_use#built-in-events-and-actions" >}}).
If the model asks to call tools, `chat_model_action` sends a `ToolRequestEvent` instead of a final `ChatResponseEvent`. After the tools finish, it receives the matching `ToolResponseEvent`, appends the tool results to the chat history, and calls the model again. This loop continues until the model returns a final response. For details on how tools are executed, see [Built-in Events and Actions in Tool Use]({{< ref "docs/development/tool_use#built-in-events-and-actions" >}}).
1 change: 1 addition & 0 deletions docs/content/docs/faq/faq.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -101,6 +101,7 @@ Flink Agents provides built-in integrations for many ecosystem providers. Some i
| [Anthropic]({{< ref "docs/development/chat_models#anthropic" >}}) | ✅ | ✅ |
| [Azure AI]({{< ref "docs/development/chat_models#azure-ai" >}}) | ❌ | ✅ |
| [Azure OpenAI]({{< ref "docs/development/chat_models#azure-openai" >}}) | ✅ | ✅ |
| [Gemini]({{< ref "docs/development/chat_models#gemini" >}}) | ❌ | ✅ |
| [Ollama]({{< ref "docs/development/chat_models#ollama" >}}) | ✅ | ✅ |
| [OpenAI]({{< ref "docs/development/chat_models#openai" >}}) | ✅ | ✅ |
| [Tongyi (DashScope)]({{< ref "docs/development/chat_models#tongyi-dashscope" >}}) | ✅ | ❌ |
Expand Down
Loading