Sapial

Semi-Autonomous Personal Intelligent Agent from Large language models

Sapial is a open framework for building personal agents that can be trained on your own data. It is built on top of Deno, [Transforemrs](), [Langchain](), and [llama-index]().

Why Deno?

Deno is a secure, user-friendly runtime for JavaScript, TypeScript, and WebAssembly, built on V8, Rust, and Tokio. It supports TypeScript natively, provides built-in development tools, and accommodates existing npm modules. Distributed as a single executable, Deno doubles as both a runtime and package manager, using browser-compatible protocols for loading modules. It's an ideal environment for modern programming and a great alternative for utility scripts previously written in Bash or Python.

Architecture

Sapial is broadly divided into three parts:

1. Python backend services, exposed over local FastAPI bridge
2. Deno runtime (middleware), for running the agent
3. Client Interfaces, that connect to the Deno runtime via local HTTP API

Table of contents

• Setup Sapial
• Start building Sapial
• Contributors guide

Setup

1. Install Deno

   To check if you already have it installed, run

   deno --version

   If Deno is installed, you should see a message like deno 1.x.x where x.x represents a version of deno

   If you get command not found proceed with installation according to your OS

   MacOS and Linux

   Using Shell ```bash curl -fsSL https://deno.land/x/install/install.sh | sh ``` or using [Homebrew](https://formulae.brew.sh/formula/deno) ```bash brew install deno ```
   Windows

   Using [Scoop](https://scoop.sh/) ```bash scoop install deno ``` or using [Chocolatey](https://chocolatey.org/packages/deno) ```bash choco install deno ```

2. Install Python and Pip

   To check if you already have it installed, run

   python --version

   If Python is installed, you should see a message like python 2.x.x where x.x represents a version of Python

   If you get command not found proceed with installation according to your OS

   MacOS and Linux

   
   Directly through [python](https://www.python.org/downloads/) website or, using [Homebrew](https://formulae.brew.sh/) ```bash brew install python ```
   Windows

   
   Directly through [python](https://www.python.org/downloads/) website or, using [Scoop](https://scoop.sh/) ```bash scoop install main/python ``` or, using [Chocolatey](https://chocolatey.org) ```bash choco install python --pre ```

   Important!

   In case you didn't have Python installed, make sure to restart your computer before proceeding to the next steps!

   Usually Pip is automatically installed with python

   To check if you already have it installed, run

   pip --version

   If you get command not found proceed with installation according to your OS

   MacOS and Linux

   
   Using cli ```bash python -m ensurepip --upgrade ```
   Windows

   
   # add OpenAI API keys to .env file Using cli ```bash C:> py -m ensurepip --upgrade ```

3. Setup a Python Virtual Environment

   To setup the virtual environment, run this command in the terminal pip install virtualenv

   To use venv in your project, change directory into services by running cd services

   nd then run python -m venv venv

   This will create a vurtual environment inside services folder named venv

   run cd .. in your terminal to get back to the root folder

4. Install Python Dependencies

   To install, run pip install -r ./services/requirements.txt in your terminal

5. Setup your local environmental variables

   In the root of directory, create a new file named .env

   Open .env file using editor of your choice and add your OpenAI API key

   Your .env file should look like this

   OPENAI_API_KEY = "your_API_KEY" 

Installation

Starting an agent

To start an agent, run the following command in your terminal

deno run --allow-all --unstable ./agents/alice.ts

Starting a client

In a separate terminal, start the CLI client by running

cd /clients/cli
deno run --allow-all httpClient.ts

To check if everything is working properly, type any prompt in the latter terminal

Starting a web client

Alternatively, you can start a web client to use a web interface To start a web client, run yarn && yarn dev

By default, the server will spin up on port 3000 Proceed to localhost:3000 and send a message to start interacting with AI

Contribution guide

We appreciate and highly encourage community contributions!

1. Fork the repository (repo)

   If you're not sure, here's how to fork the repo.

2. Clone your fork

   To clone your repo, you can run the command in the terminal of your choice

   git clone git@github.com:[your_github_handle]/sapial.git && cd sapial

3. Make, commit and push changes

   Create new branch and make changes

   git checkout -b new_branch_name

   Make changes where necessary and save them by running

   git add --all
   git commit -m "Use a message that best describes changes"

   Push the changes to your repo

   git push

4. Propose changes by opening a PR

   After your changes are committed to your GitHub fork, submit a pull request (PR) to the main branch of the sapial-ai/sapial repo

   Debug

   Kill the fastAPI server on agent crash lsof -i :8000 -> kill -9

Next Steps Implementation

• add logging
• start implementing basic agent architectures
• add formatting to text
• add a retriever using llama-index
• tools and plugins

Final Sharing

• share notes and repo with Andrew

Next Steps Planning

• define personas
• define personal alignment stratgies

Later

• Refactor http server to be a class
• Add authentication for agents and clients using LibP2P