aybruhm / agenta

Easily build, version, evaluate and deploy your LLM-powered apps

Home Page | Slack | Documentation

Quickly iterate, debug, and evaluate your LLM apps
The open-source LLMOps platform for prompt-engineering, evaluation, human feedback, and deployment of complex LLM apps.

---

About • Demo • Quick Start • Installation • Features • Documentation • Support • Community • Contributing

---

ℹ️ About

Building production-ready LLM-powered applications is currently very difficult. It involves countless iterations of prompt engineering, parameter tuning, and architectures.

Agenta provides you with the tools to quickly do prompt engineering and 🧪 experiment, ⚖️ evaluate, and 🚀 deploy your LLM apps. All without imposing any restrictions on your choice of framework, library, or model.

Demo

agenta-demo.mp4

Quick Start

Features

Playground 🪄

With just a few lines of code, define the parameters and prompts you wish to experiment with. You and your team can quickly experiment and test new variants on the web UI.

playground_1024_30.07.2023.mp4

Version Evaluation 📊

Define test sets, then evaluate manually or programmatically your different variants.

API Deployment 🚀

When you are ready, deploy your LLM applications as APIs in one click.

Why choose Agenta for building LLM-apps?

• 🔨 Build quickly: You need to iterate many times on different architectures and prompts to bring apps to production. We streamline this process and allow you to do this in days instead of weeks.
• 🏗️ Build robust apps and reduce hallucination: We provide you with the tools to systematically and easily evaluate your application to make sure you only serve robust apps to production.
• 👨‍💻 Developer-centric: We cater to complex LLM-apps and pipelines that require more than one simple prompt. We allow you to experiment and iterate on apps that have complex integration, business logic, and many prompts.
• 🌐 Solution-Agnostic: You have the freedom to use any libraries and models, be it Langchain, llma_index, or a custom-written alternative.
• 🔒 Privacy-First: We respect your privacy and do not proxy your data through third-party services. The platform and the data are hosted on your infrastructure.

How Agenta works:

1. Write your LLM-app code

Write the code using any framework, library, or model you want. Add the agenta.post decorator and put the inputs and parameters in the function call just like in this example:

Example simple application that generates baby names:

import agenta as ag
from langchain.chains import LLMChain
from langchain.llms import OpenAI
from langchain.prompts import PromptTemplate

default_prompt = "Give me five cool names for a baby from {country} with this gender {gender}!!!!"
ag.init()
ag.config(prompt_template=ag.TextParam(default_prompt),
          temperature=ag.FloatParam(0.9))

@ag.entrypoint
def generate(
    country: str,
    gender: str,
) -> str:
    llm = OpenAI(temperature=ag.config.temperature)
    prompt = PromptTemplate(
        input_variables=["country", "gender"],
        template=ag.config.prompt_template,
    )
    chain = LLMChain(llm=llm, prompt=prompt)
    output = chain.run(country=country, gender=gender)

    return output

2.Deploy your app using the Agenta CLI

3. Go to agenta at localhost:3000

Now your team can 🔄 iterate, 🧪 experiment, and ⚖️ evaluate different versions of your app (with your code!) in the web platform.

Support

Talk with the founders for any commercial inquiries.

Disabling Anonymized Tracking

To disable anonymized telemetry, set the following environment variable:

• For web: Set TELEMETRY_TRACKING_ENABLED to false in your agenta-web/.env file.
• For CLI: Set telemetry_tracking_enabled to false in your ~/.agenta/config.toml file.

After making this change, restart agenta compose.

Contributing

We warmly welcome contributions to Agenta. Feel free to submit issues, fork the repository, and send pull requests.

We are usually hanging in our Slack. Feel free to join our Slack and ask us anything

Check out our Contributing Guide for more information.

Contributors ✨

Thanks goes to these wonderful people (emoji key):

Sameh Methnani 💻 📖	Suad Suljovic 💻 🎨 🧑‍🏫 👀	burtenshaw 💻	Abram 💻 📖	Israel Abebe 🐛 🎨 💻	Master X 💻	corinthian 💻 🎨
Pavle Janjusevic 🚇	Kaosi Ezealigo 🐛 💻	Alberto Nunes 🐛	Maaz Bin Khawar 💻 👀 🧑‍🏫	Nehemiah Onyekachukwu Emmanuel 💻 💡 📖	Philip Okiokio 📖	Abhinav Pandey 💻
Ramchandra Warang 💻 🐛	Biswarghya Biswas 💻	Uddeepta Raaj Kashyap 💻	Nayeem Abdullah 💻	Kang Suhyun 💻	Yoon 💻	Kirthi Bagrecha Jain 💻
Navdeep 💻	Rhythm Sharma 💻	Osinachi Chukwujama 💻	莫尔索 📖	Agunbiade Adedeji 💻	Emmanuel Oloyede 💻 📖	Dhaneshwarguiyan 💻
Priyanshu Prajapati 📖	Raviteja 💻	Arijit 💻	Yachika9925 📖	Aldrin ⚠️	seungduk.kim.2304 💻	Andrei Dragomir 💻
diego 💻

This project follows the all-contributors specification. Contributions of any kind are welcome!

Attribution: Testing icons created by Freepik - Flaticon