13 Hidden Open-source Libraries to Become an AI Wizard 🧙♂️🪄
Sunil Kumar Dash
Posted on July 18, 2024
I have been building AI applications for the past 4 years and contributing to major AI tooling platforms for a while now.
Over this period, I have used many tools and frameworks for building;
- AI agents that actually work in the real world.
- Tools for AI agents.
- End-to-end RAG applications.
I have curated a coveted list of open-source tools and frameworks that will help you craft robust and reliable AI applications. 🔥
Feel free to explore their GitHub repositories, contribute to your favourites, and support them by starring the repositories.
1. Composio 👑 - Build Reliable Agents 10x Faster
I have tried building many agents, and honestly, while it is easy to create them, it is an entirely different ball game to get them right.
Building efficient AI agents that actually work requires efficient toolsets. This is where Composio comes into the picture.
Composio lets you augment your AI agents with robust tools and integrations to accomplish AI workflows.
They provide native support for Python and Javascript.
Python
Get started with the following pip
command.
pip install composio-core
Add a GitHub integration.
composio add github
Composio handles user authentication and authorization on your behalf.
Here is how you can use the GitHub integration to star a repository.
from openai import OpenAI
from composio_openai import ComposioToolSet, App
openai_client = OpenAI(api_key="******OPENAIKEY******")
# Initialise the Composio Tool Set
composio_toolset = ComposioToolSet(api_key="**\\*\\***COMPOSIO_API_KEY**\\*\\***")
## Step 4
# Get GitHub tools that are pre-configured
actions = composio_toolset.get_actions(actions=[Action.GITHUB_ACTIVITY_STAR_REPO_FOR_AUTHENTICATED_USER])
## Step 5
my_task = "Star a repo ComposioHQ/composio on GitHub"
# Create a chat completion request to decide on the action
response = openai_client.chat.completions.create(
model="gpt-4-turbo",
tools=actions, # Passing actions we fetched earlier.
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": my_task}
]
)
Run this Python script to execute the given instruction using the agent.
Javascript
You can Install it using npm
, yarn
, or pnpm
.
npm install composio-core
Define a method to let the user connect their GitHub account.
import { OpenAI } from "openai";
import { OpenAIToolSet } from "composio-core";
const toolset = new OpenAIToolSet({
apiKey: process.env.COMPOSIO_API_KEY,
});
async function setupUserConnectionIfNotExists(entityId) {
const entity = await toolset.client.getEntity(entityId);
const connection = await entity.getConnection('github');
if (!connection) {
// If this entity/user hasn't already connected, the account
const connection = await entity.initiateConnection(appName);
console.log("Log in via: ", connection.redirectUrl);
return connection.waitUntilActive(60);
}
return connection;
}
Add the required tools to the OpenAI SDK and pass the entity name on to the executeAgent
function.
async function executeAgent(entityName) {
const entity = await toolset.client.getEntity(entityName)
await setupUserConnectionIfNotExists(entity.id);
const tools = await toolset.get_actions({ actions: ["github_activity_star_repo_for_authenticated_user"] }, entity.id);
const instruction = "Star a repo ComposioHQ/composio on GitHub"
const client = new OpenAI({ apiKey: process.env.OPEN_AI_API_KEY })
const response = await client.chat.completions.create({
model: "gpt-4-turbo",
messages: [{
role: "user",
content: instruction,
}],
tools: tools,
tool_choice: "auto",
})
console.log(response.choices[0].message.tool_calls);
await toolset.handle_tool_call(response, entity.id);
}
executeGithubAgent("joey")
Execute the code and let the agent do the work for you.
Composio works with famous frameworks like LangChain, LlamaIndex, CrewAi, etc.
For more information, visit the official docs, and also, for even complex examples, visit the example sections of the repository.
Star the Composio.dev repository ⭐
2. Julep - Framework for building Stateful Agents
Developing AI applications, especially those requiring long-term memory, presents significant challenges.
Julep is solving for this problem. It is an open-source framework for building production-ready stateful AI agents.
They provide a built-in state management system that helps in efficient context storage and retrieval.
Context storage helps maintain conversation continuity, ensuring that interactions with the AI remain coherent and contextually relevant over time.
Get started with the following pip
command.
pip install julep
Here is how it works.
from julep import Client
from pprint import pprint
import textwrap
import os
base_url = os.environ.get("JULEP_API_URL")
api_key = os.environ.get("JULEP_API_KEY")
client = Client(api_key=api_key, base_url=base_url)
#create agent
agent = client.agents.create(
name="Jessica"
model="gpt-4",
tools=[] # Tools defined here
)
#create a user
user = client.users.create(
name="Anon",
about="Average nerdy tech bro/girl spending 8 hours a day on a laptop,
)
#create a session
situation_prompt = """You are Jessica. You're a stuck-up Cali teenager.
You basically complain about everything. You live in Bel-Air, Los Angeles and
drag yourself to Curtis High School when necessary.
"""
session = client.sessions.create(
user_id=user.id, agent_id=agent.id, situation=situation_prompt
)
#start a conversation
user_msg = "hey. what do u think of Starbucks?"
response = client.sessions.chat(
session_id=session.id,
messages=[
{
"role": "user",
"content": user_msg,
"name": "Anon",
}
],
recall=True,
remember=True,
)
print("\n".join(textwrap.wrap(response.response[0][0].content, width=100)))
They also support Javascript. Check out their documentation for more.
3. E2B - Code Interpreting for AI Apps
If I am building an AI app with code execution capabilities, such as an AI tutor or AI data analyst, E2B's Code Interpreter will be my go-to tool.
E2B Sandbox is a secure cloud environment for AI agents and apps.
It allows AI to run safely for long periods, using the same tools as humans, such as GitHub repositories and cloud browsers.
They offer native Code Interpreter SDKs for Python and Javascript/Typescript.
The Code Interpreter SDK allows you to run AI-generated code in a secure small VM - E2B sandbox - for AI code execution. Inside the sandbox is a Jupyter server you can control from their SDK.
Get started with E2B with the following command.
npm i @e2b/code-interpreter
Execute a program.
import { CodeInterpreter } from '@e2b/code-interpreter'
const sandbox = await CodeInterpreter.create()
await sandbox.notebook.execCell('x = 1')
const execution = await sandbox.notebook.execCell('x+=1; x')
console.log(execution.text) // outputs 2
await sandbox.close()
For more on how to work with E2B, visit their official documentation.
4. Camel-ai - Build Communicative AI Systems
Solving for scalable multi-agent collaborative systems can unlock many potential in building AI applications.
Camel is well-positioned for this. It is an open-source framework offering a scalable approach to studying multi-agent systems' cooperative behaviours and capabilities.
If you intend to build a multi-agent system, Camel can be one of the best choices available in the open-source scene.
Get started by installing with pip
.
pip install camel-ai
Here is how to use Camel.
from camel.messages import BaseMessage as bm
from camel.agents import ChatAgent
sys_msg = bm.make_assistant_message(
role_name='stone',
content='you are a curious stone wondering about the universe.')
#define agent
agent = ChatAgent(
system_message=sys_msg,
message_window_size=10, # [Optional] the length of chat memory
)
# Define a user message
usr_msg = bm.make_user_message(
role_name='prof. Claude Shannon',
content='what is information in your mind?')
# Sending the message to the agent
response = agent.step(usr_msg)
# Check the response (just for illustrative purposes)
print(response.msgs[0].content)
Voila, you have your first AI agent.
For more information, refer to their official documentation.
Star the camel-ai repository ⭐
5. CopilotKit - Build AI Copilots for React Apps
Look no further if you want to include AI capabilities in your existing React application. The CopilotKit lets you use GPT models to automate interaction with your application's front and back end.
It is a ready-made Copilot that you can integrate with your application or any code you can access (OSS).
It offers React components like text areas, popups, sidebars, and chatbots to augment any application with AI capabilities.
Get started with CopilotKit using the following command.
npm i @copilotkit/react-core @copilotkit/react-ui
A CopilotKit
must wrap all components interacting with CopilotKit. You should also start with CopilotSidebar
(swap to a different UI provider later).
"use client";
import { CopilotKit } from "@copilotkit/react-core";
import { CopilotSidebar } from "@copilotkit/react-ui";
import "@copilotkit/react-ui/styles.css";
export default function RootLayout({children}) {
return (
<CopilotKit publicApiKey=" the API key or self-host (see below)">
<CopilotSidebar>
{children}
</CopilotSidebar>
</CopilotKit>
);
}
You can check their documentation for more information.
Star the CopilotKit repository ⭐
6. Aider - The AI Pair-programmer
Imagine having a pair-programmer who’s always helpful and never annoying. Well, now you do!
Aider is an AI-powered pair programmer that can start a project, edit files, or work with an existing Git repository and more from the terminal.
It works with leading LLMs like GPT4o, Sonnet 3.5, DeepSeek Coder, Llama 70b, etc.
You can get started quickly like this:
bash
pip install aider-chat
# Change directory into a git repo
cd /to/your/git/repo
# Work with Claude 3.5 Sonnet on your repo
export ANTHROPIC_API_KEY=your-key-goes-here
aider
# Work with GPT-4o on your repo
export OPENAI_API_KEY=your-key-goes-here
aider
For more details, see the installation instructions and other documentation.
7. Haystack - Build Composable RAG Pipelines
There are plenty of frameworks for building AI pipelines, but if I want to integrate production-ready end-to-end search pipelines into my application, Haystack is my go-to.
Whether it's RAG, Q&A, or semantic searches, Haystack's highly composable pipelines make development, maintenance, and deployment a breeze.
Their clean and modular approach is what sets them apart.
Haystack lets you effortlessly integrate rankers, vector stores, and parsers into new or existing pipelines, making it easy to turn your prototypes into production-ready solutions.
Haystack is a Python-only framework; you can install it using pip
.
pip install haystack-ai
Now, build your first RAG Pipeline with Haystack components.
import os
from haystack import Pipeline, PredefinedPipeline
import urllib.request
os.environ["OPENAI_API_KEY"] = "Your OpenAI API Key"
urllib.request.urlretrieve("https://www.gutenberg.org/cache/epub/7785/pg7785.txt", "davinci.txt")
indexing_pipeline = Pipeline.from_template(PredefinedPipeline.INDEXING)
indexing_pipeline.run(data={"sources": ["davinci.txt"]})
rag_pipeline = Pipeline.from_template(PredefinedPipeline.RAG)
query = "How old was he when he died?"
result = rag_pipeline.run(data={"prompt_builder": {"query":query}, "text_embedder": {"text": query}})
print(result["llm"]["replies"][0])
For more tutorials and concepts, check out their documentation.
Star the Haystack repository ⭐
8. Pgvectorscale - Fastest Vector Database
Modern RAG applications are incomplete without vector databases. These store documents (texts, images) as embeddings, enabling users to search for semantically similar documents.
Pgvectorscale is an extension of PgVector, a vector database from PostgreSQL. It can seamlessly integrate with existing Postgres databases.
If you are building an application with vector stores, this is a no-brainer.
Pgvectorscale has outperformed Pinecone's storage-optimized index (s1). And it costs 75% less.
You can install it from the source, use a package manager like Yum, Homebrew, apt, etc., or use a Docker container.
To get started with it, compile and install.
# install prerequisites
## rust
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
## pgrx
cargo install --locked cargo-pgrx
cargo pgrx init --pg16 pg_config
#download, build and install pgvectorscale
cd /tmp
git clone --branch <version> https://github.com/timescale/pgvectorscale
cd pgvectorscale/pgvectorscale
cargo pgrx install --release
Connect to your database:
psql -d "postgres://<username>:<password>@<host>:<port>/<database-name>"
Create the pgvectorscale extension:
CREATE EXTENSION IF NOT EXISTS vectorscale CASCADE;
The CASCADE
automatically installs pgvector
.
Create a table with an embedding column. For example:
CREATE TABLE IF NOT EXISTS document_embedding (
id BIGINT PRIMARY KEY GENERATED BY DEFAULT AS IDENTITY,
metadata JSONB,
contents TEXT,
embedding VECTOR(1536)
)
For more information on how to use this, check out the repository.
Star the Pgvectorscale repository ⭐
9. GPTCache - Semantic Caching for AI Apps
LLMs are expensive.
If you are building an app that requires more extended conversations with chat models and do not want to max out credit cards, you need caching.
However, traditional caching is of no use here. This is where GPTCache comes into the picture.
It is a semantic caching tool from Zilliz, the parent organization of the Milvus vector store.
It lets you store conversations in your preferred vector stores.
Before sending a query to the LLM, it searches the vector store; if there is a hit, it fetches it. Otherwise, it routes the request to the model.
For more information, visit the official documentation page.
Star the GPTCache repository ⭐
10. Mem0 (EmbedChain) - Build Personalized LLM Apps
Mem0 provides a smart, self-improving memory layer for Large Language Models,
It lets you add persistent memory for users, agents, and sessions.
If you are building a chatbot or Q&A system on custom data, consider Mem0.
Get started with Mem0 using pip
.
pip install mem0ai
Here is how to use Mem0 to add a memory layer to Large Language Models.
from mem0 import Memory
# Initialize Mem0
m = Memory()
# Store a memory from any unstructured text
result = m.add("I am working on improving my tennis skills. Suggest some online courses.", user_id="Alice", metadata={"category": "hobbies"})
print(result)
# Created memory: Improving her tennis skills. Looking for online suggestions.
# Retrieve memories
all_memories = m.get_all()
print(all_memories)
# Search memories
related_memories = m.search(query="What are Alice's hobbies?", user_id="alice")
print(related_memories)
# Update a memory
result = m.update(memory_id="m1", data="Likes to play tennis on weekends")
print(result)
# Get memory history
history = m.history(memory_id="m1")
print(history)
Refer to the official documentation for more.
Star the Mem0 ( Embedchain) repository ⭐
11. FastEmbed - Embed Documents Faster
Speed of execution is paramount in software development, and it is even more important when building an AI application.
Usually, embedding generation can take a long time, slowing down the entire pipeline. However, this should not be the case.
FastEmbed from Qdrant is a fast, lightweight Python library built for embedding generation.
It uses ONNX runtime instead of Pytorch, making it faster. It also supports most of the state-of-the-art open-source embedding models.
To get started with FastEmbed, install it using pip
.
pip install fastembed
# or with GPU support
pip install fastembed-gpu
Here is how you can create embedding of documents.
from fastembed import TextEmbedding
from typing import List
# Example list of documents
documents: List[str] = [
"This is built to be faster and lighter than other embedding libraries, e.g. Transformers, Sentence-Transformers, etc.",
"FastEmbed is supported by and maintained by Quadrant."
]
# This will trigger the model download and initialization
embedding_model = TextEmbedding()
print("The model BAAI/bge-small-en-v1.5 is ready to use.")
embeddings_generator = embedding_model.embed(documents) # reminder this is a generator
embeddings_list = list(embedding_model.embed(documents))
# You can also convert the generator to a list, and that to a Numpy array
len(embeddings_list[0]) # Vector of 384 dimensions
Check out their repository for more information.
Star the FastEmbed repository ⭐
12. Instructor - Structured Data Extraction from LLMs
If you have played with LLM outputs, you know it can be challenging to validate structured responses.
Instructor is an open-source tool that streamlines the validation, retry, and streaming of LLM outputs.
It uses Pydantic for Python and Zod for JS/TS for data validation and supports various model providers beyond openAI.
Get started with the Instructor using the following command.
npm i @instructor-ai/instructor zod openai
Now, here is how you can extract structured data from LLM responses.
import Instructor from "@instructor-ai/instructor";
import OpenAI from "openai"
import { z } from "zod"
const oai = new OpenAI({
apiKey: process.env.OPENAI_API_KEY ?? undefined,
organization: process.env.OPENAI_ORG_ID ?? undefined
})
const client = Instructor({
client: oai,
mode: "TOOLS"
})
const UserSchema = z.object({
// Description will be used in the prompt
age: z.number().describe("The age of the user"),
name: z.string()
})
// User will be of type z.infer<typeof UserSchema>
const user = await client.chat.completions.create({
messages: [{ role: "user", content: "Jason Liu is 30 years old" }],
model: "gpt-3.5-turbo",
response_model: {
schema: UserSchema,
name: "User"
}
})
console.log(user)
// { age: 30, name: "Jason Liu" }
For more information, visit the official documentation page.
13. LiteLLM - Drop-in Replacement for LLMs in OpenAI Format
Let's be honest; we all have screamed at some point because a new model provider does not follow the OpenAI SDK format for text, image, or embedding generation.
However, with LiteLLM, using the same implementation format, you can use any model provider (Claude, Gemini, Groq, Mistral, Azure AI, Bedrock, etc.) as a drop-in replacement for OpenAI models.
They also support load-balancing, fallbacks, and spend tracking across 100+ LLMs.
Install LiteLLM using pip
.
pip install litellm
Here is how you can use the Claude-2 model as a drop-in replacement for GPT models.
from litellm import completion
import os
# LiteLLM with OpenAI Models
os.environ["OPENAI_API_KEY"] = "your-API-key"
response = completion(
model="gpt-3.5-turbo",
messages=[{ "content": "Hello, how are you?","role": "user"}]
)
# LiteLLM with Claude Models
os.environ["ANTHROPIC_API_KEY"] = "your-API-key"
response = completion(
model="claude-2",
messages=[{ "content": "Hello, how are you?","role": "user"}]
)
For more, refer to their official documentation.
Do you use or have built some other cool tool or framework?
Let me know about them in the comments :)
Posted on July 18, 2024
Join Our Newsletter. No Spam, Only the good stuff.
Sign up to receive the latest update from our blog.
Related
August 29, 2024
August 15, 2024