tinyagents 0.1.0

A tiny, lightweight and unintrusive library for orchestrating agentic applications.

TinyAgents

A tiny, lightweight and unintrusive library for orchestrating agentic applications.

Here's the big idea:

1. 😶‍🌫️ Less than 500 lines of code. "I have absolutely no idea what this library is doing under the hood"
2. 😨 Avoid including the entirety of PyPi in the requirements.txt
3. 🚀 Minimise the need for code changes. "The library is no longer being maintained - so let's retire our application with it."

Upcoming:

1. Support for deployment and execution using Ray.
   • easily deploy and scale nodes in your graph as Ray Serve deployments.
2. A set of callbacks for storing traces to Redis, Pub/Sub or disk.
3. A tool for visualising what is happening within your graphs

Installation

pip install https://github.com/adam-h-ds/tinyagents.git

How it works!

Define your graph using standard operators

Parallelisation

from tinyagents import chainable

@chainable
def tool1(inputs: dict):
    return ...

@chaianble
def tool2(inputs: dict):
    return ...

@chainable
class Agent:
    def __init__(self):
        ...

    def run(self, inputs: list):
        return ...

# run `tool1` and `tool2` in parallel, then pass outputs to `Agent`
graph = (tool1 & tool2) | Agent()
executor = graph.compile()

executor.execute("Hello!")

Branching

jailbreak_check = ...
agent1 = ...
agent2 = ...
guardrail = ...

def my_router(inputs: str):
    if inputs == "jailbreak_attempt":
        return agent2.name

    return agent1.name

# check for jailbreaks, then run either `agent1` or `agent2`, then finally run `guardrail`
graph = jailbreak_check | (agent1 / agent2).bind_router(my_router) | guardrail

print(graph)

## jailbreak_check -> ConditionalBranch(agent1, agent2) -> guardrail

Looping

from tinyagents import loop

agent1 = ...
agent2 = ...
agent3 = ...
guardrail = ...

# run at most 3 iterations between `agent1` and `agent2`, then pass to `agent3` and finally `guardrail`.
graph = loop(agent1, agent2, max_iter=3) | agent3 | guardrail

print(graph)

## Recursive(agent1, agent2) -> agent3 -> guardrail