Clients

Galadriel framework comes with a set of clients out of the box. New core clients are regularly added to the framework. Below is a list of currently supported clients and how to use them.

​

Cron Client

Cron client sends empty message to the agent in order to trigger the agent execution at regular intervals.

Use Cases

• a trading agent which regularly executes its own trading strategy based on market data and updates users’ portfolios
• a twitter agent which distills world news’ every 1 hour and prepares a comprehensive report

Example:

trading_agent = # Create your agent here
cron = Cron(60) # Configure Cron client

agent = AgentRuntime(
    inputs=[cron],  # Pass the Cron client to trigger agent execution regularly
    outputs=[],
    agent=trading_agent,
)

​

Discord Client

This client listens to messages in a Discord server and replies with agent-generated responses.

Use Cases

• an assistant agent which answers questions, moderates conversations, or assists in a Discord community.

Example:

agent = # Create your agent here
DISCORD_GUILD_ID = "<DISCORD_GUILD_ID>" # Add your Discord guild ID
discord_client = DiscordClient(guild_id=DISCORD_GUILD_ID)) # Create Discord client
knowledge_base_output_client = # Create client eg based on [SimpleMessageClient](https://github.com/galadriel-ai/galadriel/blob/main/galadriel/clients/simple_message_client.py) which updates company's knowledge base based on new facts from Discord channels

agent = AgentRuntime(
    inputs=[discord_client], # Pass the Discord client to listen to Discord inputs
    outputs=[knowledge_base_output_client],
    agent=agent,
)

​

Telegram Client

The Telegram client listens to messages in a Telegram server and replies with agent-generated responses.

Use Cases

• An assistant agent that answers questions, moderates conversations, or assists in a Telegram community.

Example:

agent = # Create your agent here
TELEGRAM_TOKEN = "<TELEGRAM_TOKEN>" # Add your Telegram bot token ID
telegram_client = TelegramClient(token=TELEGRAM_TOKEN))

agent = AgentRuntime(
    inputs=[telegram_client],
    outputs=[telegram_client],
    agent=agent,
)

​

Gradio Client

The Gradio client listens to messages from a local Gradio server and replies with agent-generated responses.

Use Cases

• Test and debug your agent without requiring extra tokens or configuration.

agent = # Create your agent here
gradio_client = GradioClient()

agent = AgentRuntime(
    inputs=[gradio_client],
    outputs=[gradio_client],
    agent=agent,
)

​

Terminal Client

The Terminal client listens to messages from the terminal and replies with agent-generated responses.

Use Cases

• Test and debug your agent without requiring extra tokens or configuration.

agent = # Create your agent here
terminal_client = TerminalClient()

agent = AgentRuntime(
    inputs=[terminal_client],
    outputs=[terminal_client],
    agent=agent,
)

​

SimpleMessageClient

A simple client that receives messages when instantiated and pushes them to the agent at a specified interval. The agent’s responses are then printed to the console.

Use Cases

• Test and debug your agent without requiring extra (authentication?) tokens or configuration.

agent = # Create your agent here
client = SimpleMessageClient("What is the capital of Estonia?")

agent = AgentRuntime(
    inputs=[client],
    outputs=[client],
    agent=agent,
)

​

TwitterMentionClient

The TwitterMention client listens to mentions on Twitter and replies accordingly.

Use Cases

• An assistant agent that answers questions, moderates conversations, or assists in a Twitter community.

agent = # Create your agent here
twitter_client = TwitterMentionClient(
    TwitterCredentials(
        consumer_api_key="<TWITTER_CONSUMER_API_KEY>",
        consumer_api_secret="<TWITTER_CONSUMER_API_SECRET>",
        access_token="<TWITTER_ACCESS_TOKEN>",
        access_token_secret="<TWITTER_ACCESS_TOKEN_SECRET>",
    ),
    user_id="<TWITTER_USER_ID>",
)

agent = AgentRuntime(
    inputs=[twitter_client],
    outputs=[twitter_client],
    agent=agent,
)

​

TwitterPostClient

The TwitterPost client is an AgentOutput type that posts new messages on a Twitter account.

Use Cases

• An assistant that posts curated and personalized content on Twitter.

agent = # Create your agent here

agent = AgentRuntime(
    inputs=[Cron(POST_INTERVAL_SECONDS)],
    outputs=[TwitterPostClient()],
    agent=agent,
)

​

Build your own Client

In order to build the client, implement one or both of these interfaces:

​

AgentInput

class AgentInput:
    async def start(self, queue: PushOnlyQueue) -> None:
        pass

start function contains a queue to which the agent should push a Message object.

class Message(BaseModel):
    content: str
    conversation_id: Optional[str] = None
    type: Optional[str] = None
    additional_kwargs: Optional[Dict] = None

In the simplest case, Message has justcontent .

Let’s implement the simplest AgentInput

class SimpleInput(AgentInput):
    async def start(self, queue: PushOnlyQueue):
        await queue.put(Message(content="Hello, Agent!"))

This client, when passed to the runtime will pass a single Hello, Agent! message to the agent.

​

AgentOutput

class AgentOutput:
    async def send(self, request: Message, response: Message) -> None:
        pass

send function receives the response from the client, together with the request which trigger the agent flow.

Here is the simplest implementation of

class SimpleOutput(AgentOutput):
    async def send(self, request: Message, response: Message):
        print(f"Agent Response: {response.content}")

As explained above, the input and output need to be passed to AgentRuntime to start operating.

​

Conclusion

Galadriel provides a diverse set of built-in clients to seamlessly integrate agents with various platforms, including messaging services, social media, and local development environments. Whether you need to trigger agents at regular intervals, process real-time user interactions, or debug locally, these clients offer flexible solutions. Additionally, developers can extend Galadriel’s capabilities by implementing custom clients using the AgentInput and AgentOutput interfaces. This modular approach ensures adaptability to a wide range of use cases, enabling efficient and intelligent agent interactions across different ecosystems.