swarmauri_llm_gemini 0.11.0.dev1

Provider-specific Swarmauri import package for Google Gemini generateContent, streaming, function calling, tool use, and async LLM workflows.

Swarmauri Gemini LLM

swarmauri_llm_gemini provides provider-specific imports for Google Gemini models in Swarmauri applications. It exports GeminiProModel for standard generation and streaming workflows and GeminiToolModel for Gemini function calling through Swarmauri toolkits.

The adapters use Google Gemini's generateContent and streamGenerateContent API shape through generativelanguage.googleapis.com/v1beta. They translate Swarmauri conversations into Gemini contents, apply safety settings, preserve system instructions where supported by the implementation, and append generated responses back to the conversation as AgentMessage instances.

Why Use This Package?

• Keep Google Gemini-specific dependencies explicit in provider-aware Swarmauri applications.
• Use the same Swarmauri Conversation, message, and toolkit workflows across Gemini and other LLM providers.
• Support both direct text generation and Gemini function-calling workflows.
• Stream Gemini responses through the same high-level Swarmauri model interface used by other provider packages.

FAQ

What does swarmauri_llm_gemini install?

It installs provider package entry points for GeminiProModel and GeminiToolModel, both re-exported from swarmauri_standard.

Which Gemini API methods does it use?

GeminiProModel calls models.generateContent for standard prediction and models.streamGenerateContent for streaming. GeminiToolModel uses generateContent with Gemini-compatible function declarations converted from a Swarmauri Toolkit.

Does this package support tools?

Yes. Use GeminiToolModel when a conversation should give Gemini access to Swarmauri tools. The adapter converts toolkit schemas with GeminiSchemaConverter, executes returned function calls, and can make a follow-up model call with tool results.

Which models are listed by default?

The current implementation includes Gemini 2.0 Flash, Gemini 2.0 Flash Lite, Gemini 2.0 Pro experimental, Gemini 1.5 Flash, Gemini 1.5 Flash 8B, and Gemini 1.5 Pro in its allowed model list. Confirm current production model availability in Google's Gemini API documentation before pinning deployment defaults.

Can it include usage metadata?

GeminiProModel can attach token usage metadata to the returned Swarmauri AgentMessage when include_usage is enabled and the provider response contains usage metadata.

Features

• GeminiProModel for generation, async generation, streaming, and batch conversation workflows.
• GeminiToolModel for Gemini function calling through Swarmauri toolkits.
• Support for Gemini systemInstruction / system_instruction payloads in the current adapters.
• Safety settings for harassment, hate speech, sexually explicit content, and dangerous content thresholds.
• Configurable generation controls such as temperature and max_tokens.
• Optional usage metadata handling on GeminiProModel.
• Compatibility with Python 3.10, 3.11, 3.12, 3.13, and 3.14.

Installation

uv add swarmauri_llm_gemini

pip install swarmauri_llm_gemini

Prerequisites

Create a Gemini API key in Google AI Studio or Google Cloud and pass it to the model constructor as api_key=....

Usage

from swarmauri_llm_gemini import GeminiProModel
from swarmauri_standard.conversations.Conversation import Conversation
from swarmauri_standard.messages.HumanMessage import HumanMessage

conversation = Conversation()
conversation.add_message(HumanMessage(content="Write a concise project status update."))

model = GeminiProModel(api_key="GEMINI_API_KEY")
result = model.predict(conversation=conversation, max_tokens=200)

print(result.get_last().content)

Streaming

from swarmauri_llm_gemini import GeminiProModel
from swarmauri_standard.conversations.Conversation import Conversation
from swarmauri_standard.messages.HumanMessage import HumanMessage

conversation = Conversation()
conversation.add_message(HumanMessage(content="List three risks for an AI release."))

model = GeminiProModel(api_key="GEMINI_API_KEY")

for token in model.stream(conversation=conversation, max_tokens=200):
    print(token, end="")

Tool Calling

from swarmauri_llm_gemini import GeminiToolModel
from swarmauri_standard.conversations.Conversation import Conversation
from swarmauri_standard.messages.HumanMessage import HumanMessage
from swarmauri_standard.toolkits.Toolkit import Toolkit

conversation = Conversation()
conversation.add_message(HumanMessage(content="Use the available tool if needed."))

toolkit = Toolkit()
model = GeminiToolModel(api_key="GEMINI_API_KEY")
result = model.predict(conversation=conversation, toolkit=toolkit)

print(result.get_last().content)

Related Packages

• swarmauri_llm_anthropic
• swarmauri_llm_cohere
• swarmauri_llm_deepseek
• swarmauri_llm_groq
• swarmauri_llm_mistral
• swarmauri_llm_openai

Foundational Swarmauri Packages

• swarmauri_core
• swarmauri_base
• swarmauri_standard
• swarmauri

Provider Documentation

• Gemini API reference
• Generate content with Gemini
• Gemini function calling

Best Practices

• Store Gemini API keys in environment variables or a secrets manager.
• Confirm model availability against the current Gemini API before pinning model names for production.
• Use GeminiToolModel only when the workflow actually needs tool execution.
• Keep tool schemas small and deterministic so Gemini can select and call tools reliably.

License

Apache-2.0