AI agents made simple

I have built many AI agents, and all frameworks felt so bloated, slow, and unpredictable. To change that, I hacked together a minimal library that works with JSON definitions of all steps, allowing you very simple agent flows and reproducibility. It supports concurrency for up to 1000 calls/min which allows you to process entire CSVs or dataframes of tasks.

Install

pip install flashlearn

Perplexity Clone in 10 Lines of Code

Effortlessly clone a perplexity-like feature with a few simple steps.

question = 'When was python launched?'
skill = GeneralSkill.load_skill(ConvertToGoogleQueries, client=OpenAI())
queries = skill.run_tasks_in_parallel(skill.create_tasks([{"query": question}]))["0"]
results = SimpleGoogleSearch(GOOGLE_API_KEY, GOOGLE_CSE_ID).search(queries['google_queries'])
msgs = [
    {"role": "system", "content": "insert links from search results in response to quote it"},
    {"role": "user", "content": str(results)},
    {"role": "user", "content": question},
]
print(client.chat.completions.create(model=MODEL_NAME, messages=msgs).choices[0].message.content)

Learning a New “Skill” from Sample Data

Quickly “learn” a custom skill from your sample data, then apply or store it skill.save('skill.json') for future use.

from flashlearn.skills.learn_skill import LearnSkill
from flashlearn.utils import imdb_reviews_50k

def main():
    # Instantiate your pipeline “estimator” or “transformer”
    learner = LearnSkill(model_name="gpt-4o-mini", client=OpenAI())
    data = imdb_reviews_50k(sample=100)

    # Provide instructions and sample data for the new skill
    skill = learner.learn_skill(
        data,
        task=(
            'Evaluate likelihood to buy my product and write the reason why (on key "reason")'
            'return int 1-100 on key "likely_to_Buy".'
        ),
    )

    # Construct tasks for parallel execution (akin to batch prediction)
    tasks = skill.create_tasks(data)

    results = skill.run_tasks_in_parallel(tasks)
    print(results)

Predefined Complex Pipelines in 3 Lines

Use prebuilt “skills” as specialized transformers, applying them instantly to your data.

# You can pass client to load your pipeline component
skill = GeneralSkill.load_skill(EmotionalToneDetection)
tasks = skill.create_tasks([{"text": "Your input text here..."}])
results = skill.run_tasks_in_parallel(tasks)

print(results)

Single-Step Classification Using Prebuilt Skills

Execute classic classification tasks as easily as calling fit_predict on a ML estimator.

import os
from openai import OpenAI
from flashlearn.skills.classification import ClassificationSkill

os.environ["OPENAI_API_KEY"] = "YOUR_API_KEY"
data = [{"message": "Where is my refund?"}, {"message": "My product was damaged!"}]
skill = ClassificationSkill(
    model_name="gpt-4o-mini",
    client=OpenAI(),
    categories=["billing", "product issue"],
    system_prompt="Classify the request."
)
tasks = skill.create_tasks(data)
print(skill.run_tasks_in_parallel(tasks))

Supported LLM Providers

Easily integrate LLM capabilities in your ML pipeline components.

client = OpenAI()  # Equivalent to instantiating a pipeline component
deep_seek = OpenAI(api_key='YOUR DEEPSEEK API KEY', base_url="DEEPSEEK BASE URL")
lite_llm = FlashLiteLLMClient()  # LiteLLM integration manages keys as environment variables

If you like it, Give us a star Github