First Steps with Synthora

First Steps with Synthora#

Here’s how you can get started with Synthora to build powerful and flexible agents:

from synthora.agents import VanillaAgent

agent = VanillaAgent.default()
print(agent.run("Hi! How are you?").unwrap())

Understanding the VanillaAgent#

  • VanillaAgent: This is the simplest agent provided by Synthora. It is an excellent starting point for basic interactions.

  • Return Value: If everything works correctly, the agent’s return value will be Ok(BaseMessage).

  • Accessing Text Content: The BaseMessage contains complete information about the response. If you only need the text content, you can access it directly:

    print(agent.run("Hi! How are you?").unwrap().content)

Improving Output Presentation#

Simply printing the response to the command line might suffice for basic testing, but it can get unwieldy when dealing with more extensive content. Synthora provides predefined callbacks to help you present information in a structured and visually appealing way.

For example, you can use the RichOutputHandler for better output formatting:

from synthora.agents import VanillaAgent
from synthora.callbacks import RichOutputHandler

agent = VanillaAgent.default("Your name is Synthora", handlers=[RichOutputHandler()])
agent.run("Hi! How are you?")

Bingo! You’ll notice that everything looks much more organized and easier to read.

Using Custom System Messages#

You can also customize the system message to guide the agent’s behavior. The system message acts as the “persona” or instructions for the agent, telling it how to respond to user input. For example:

agent = VanillaAgent.default("Your name is Synthora, and you are here to assist with technical queries.", handlers=[RichOutputHandler()])
agent.run("What is Synthora?")

Advanced Agents: CoT, ReAct, and More#

For more complex tasks, the VanillaAgent might not suffice. Synthora makes it easy to switch to advanced agents like CoT (Chain-of-Thought) or ReAct (Reasoning and Action) agents:

from synthora.agents import ReactAgent
from synthora.callbacks import RichOutputHandler
from synthora.prompts.buildin import ZeroShotCoTPrompt

# Using a React Agent
agent = ReactAgent.default("You are a React Agent.")
print(agent.run("Hi! How are you?").unwrap().content)

# Using a Chain-of-Thought Prompt
agent = VanillaAgent.default(ZeroShotCoTPrompt)
print(agent.run("How many letters 'r' are in the word 'strawberry'?").unwrap().content)

Empowering Agents with Tools#

To handle more sophisticated tasks, agents need tools. Synthora makes it simple to add tools to your agents.

Predefined Tools#

Synthora provides several predefined tools, such as file manipulation and web search:

from synthora.agents import VanillaAgent
from synthora.callbacks import RichOutputHandler
from synthora.toolkits.file_toolkit import FileToolkit
from synthora.toolkits.search_toolkit import search_all

write_file = FileToolkit.write_file
list_directory = FileToolkit.list_directory

agent = VanillaAgent.default(
    tools=[search_all, write_file, list_directory],
    handlers=[RichOutputHandler()],
)

Custom Tools#

You can also define your own tools effortlessly using Synthora’s decorator-based approach:

from synthora.agents import VanillaAgent
from synthora.toolkits.decorators import tool
from synthora.callbacks import RichOutputHandler

@tool
def add(a: int, b: int) -> int:
    """Add two numbers together."""
    return a + b

agent = VanillaAgent.default(
    tools=[add],
    handlers=[RichOutputHandler()],
)
agent.run("123 + 321 = ?")

The ability to integrate both predefined and custom tools allows agents to tackle a wide range of real-world problems seamlessly.

With Synthora, building, customizing, and extending intelligent agents is intuitive and efficient. Start simple with VanillaAgent and scale up to more advanced agents and tools as your needs evolve.

Contents