Skip to content

Latest commit

 

History

History
206 lines (160 loc) · 6.11 KB

README.md

File metadata and controls

206 lines (160 loc) · 6.11 KB

Backend Environment Setup

The backend requires several setup steps, including the conda environment for running, database for storage, and environment variables for configuration.

But don't worry, we have provided a one-click setup script for you to get started quickly, you can simply run:

backend/setup_script.sh

which setups most of the things for you (work for both mac and linux). And then continue from 4. LLM Setup for further setup if needed.

Or you can follow README below to setup manually because there may be unexpected errors when running the script, and you can fix them manually step by step.

We recommend you to use conda virtual environments and install the dependencies:

conda create -n openagents python=3.10
conda activate openagents
pip install -r backend/requirements.txt

As a systematic demo including backend and frontend, we need persistent data management. Here we use MongoDB as the database and Redis as the cache. To setup the database, follow the instructions below:

1. MongoDB setup

For the mac environment, follow the instructions below to install MongoDB. Assume you have brew installed in your mac.

brew tap mongodb/brew
brew update
brew install mongodb-community@6.0
brew services start mongodb-community@6.0

For the linux environment, install like this:

wget -qO - https://www.mongodb.org/static/pgp/server-4.4.asc | sudo apt-key add -
echo "deb [ arch=amd64,arm64 ] https://repo.mongodb.org/apt/ubuntu bionic/mongodb-org/4.4 multiverse" | sudo tee /etc/apt/sources.list.d/mongodb-org-4.4.list
sudo apt-get update
sudo apt-get install -y mongodb-org
sudo systemctl start mongod

Then create collections in mongodb, ready for coming data!

mongosh
> use xlang
> db.createCollection("user")
> db.createCollection("message")
> db.createCollection("conversation")
> db.createCollection("folder")
> show collections

2. Redis Setup

For MAC environment:

brew install redis

For Linux environment:

curl -fsSL https://packages.redis.io/gpg | sudo gpg --dearmor -o /usr/share/keyrings/redis-archive-keyring.gpg

echo "deb [signed-by=/usr/share/keyrings/redis-archive-keyring.gpg] https://packages.redis.io/deb $(lsb_release -cs) main" | sudo tee /etc/apt/sources.list.d/redis.list

sudo apt-get update
sudo apt-get install redis

To start redis: For Mac environment:

brew services start redis

For Linux environment:

systemctl start redis-server &

3. Environment Variables

Set these environment variables in your terminal to use redis & database, otherwise just stored in python global variables. The max redis memory is now set to 500MB, when the memory is full, LRU data will be removed. Redis serves as cache, the persistent data will be stored in MongoDB and will be retrieved when cache miss.
export VARIABLE_REGISTER_BACKEND=redis
export MESSAGE_MEMORY_MANAGER_BACKEND=database
export JUPYTER_KERNEL_MEMORY_MANAGER_BACKEND=database
export MONGO_SERVER=127.0.0.1

4. LLM Setup

Set your OpenAI key (if you use OpenAI API):

export OPENAI_API_KEY=<OPENAI_API_KEY>

Note if you are using Azure OpenAI Service, you should choose:

export OPENAI_API_TYPE=azure
export OPENAI_API_BASE=<AZURE_API_BASE>
export OPENAI_API_VERSION=<AZURE_API_VERSION>
export OPENAI_API_KEY=<AZURE_API_KEY>

If you are starting your backend in docker, you should add these environment variables in docker-compose.yml as well.

Set your Anthropic key (if you use Anthropic API):

export ANTHROPIC_API_KEY=<ANTHROPIC_API_KEY>

5. Others (Optional)

5.1 Kaggle Setup for Kaggle Search Tool (Necessary if you use KaggleSearchTool in data agent)

Please follow Kaggle Public API page and save your API token to your device, which should be save in ~/.kaggle/kaggle.json.

5.2 Auto Tool Selection (Necessary if you use "Auto" in plugins agent)

We also have auto tool selection for plugin system, to use, you can set up the environment use the following commands:

git clone https://github.com/xlang-ai/instructor-embedding
cd instructor-embedding
pip install -r requirements.txt
pip install InstructorEmbedding
cd ..

5.3 Code Execution Docker (Necessary if you want the code executed in sandbox)

If you want to start sandbox docker instead of on local OS to execute Python programs safely and avoid co-current conflicts caused by multi-users/kernels, follow the instructions below.

First, start the Python code interpreter docker we built for the data agent(see instructions under code-interpreter-docker) repository. The default docker port is localhost:8100.

docker pull xlanglab/xlang-code-interpreter-python
docker run -d --rm \
    --name codeint \
    --env-file real_agents/data_agent/.code_interpreter_docker_env \
    --mount type=bind,source=<PATH_TO_PROJECT>/backend/data,target=/home \
    -p 8100:8100 \
    --ip=10.1.1.2 \
    --cap-add SYS_RESOURCE \
    --cap-add SYS_PTRACE \
    --cap-add NET_ADMIN \
    xlanglab/xlang-code-interpreter-python

Next, set code execution mode to docker ("local" by default):

export CODE_EXECUTION_MODE="docker"

🎉🎉Congratulations!

Now you feel free to start the platform on default port 8000, using the following scripts:

export FLASK_APP=backend.main.py
flask run -p 8000