Skip to main content

All-in-One Docker Image

In this document, we will show you how to run TaskWeaver using the All-in-One Docker Image.

danger

Please note that the All-in-One Docker Image is for development and testing purposes only. It is running in the local mode, which means that the execution of the code snippets is done in the same container. Malicious code can be executed in the container, so please be cautious when running the All-in-One Docker Image.

Prerequisites​

You need to have Docker installed on your machine.

For Windows and macOS users, you can use Docker Desktop. You can download it from Docker's official website.

For Linux users, you can install following the instructions in the Docker's official website. Please find the installation guide for your specific Linux distribution.

Run TaskWeaver using the All-in-One Docker Image​

There are two versions of the TaskWeaver All-in-One Docker Image:

  • taskweavercontainers/taskweaver-all-in-one:latest: This version includes the Planner and CodeInterpreter roles only. You can use this container for code generation and execution tasks.
  • taskweavercontainers/taskweaver-all-in-one:latest-ws: This version includes an additional WebSearch role which can search the web for information. As it requires dependencies to the sentence-transformers library, it is larger.

Open a terminal and run the following command to obtain the TaskWeaver image:

docker pull taskweavercontainers/taskweaver-all-in-one:latest
# if you want to use the version with the WebSearch role 
# docker pull taskweavercontainers/taskweaver-all-in-one:latest-ws

Once the image is pulled, you can run the TaskWeaver container using the following command:

docker run -it -e LLM_API_BASE=<API_BASE> \
  -e LLM_API_KEY=<API_KEY> \
  -e LLM_API_TYPE=<API_TYPE> \
  -e LLM_MODEL=<MODEL> \
  taskweavercontainers/taskweaver-all-in-one:latest

If you want to run TaskWeaver in UI mode, you can use the following command:

docker run -it -e LLM_API_BASE=<API_BASE> \
  -e LLM_API_KEY=<API_KEY> \
  -e LLM_API_TYPE=<API_TYPE> \
  -e LLM_MODEL=<MODEL> \
  -p 8000:8000 \
  --entrypoint /app/entrypoint_chainlit.sh \
  taskweavercontainers/taskweaver-all-in-one:latest

Then you can access the TaskWeaver Web UI by visiting http://localhost:8000 in your web browser.

How to run TaskWeaver on your own project directory​

You can mount your local project directory to the container. For example, you can use the following command:

docker run -it -e LLM_API_BASE=<API_BASE> \
  -e LLM_API_KEY=<API_KEY> \
  -e LLM_API_TYPE=<API_TYPE> \
  -e LLM_MODEL=<MODEL> \
#  -e TASKWEAVER_UID=$(id -u) \ # uncomment if your host OS is not Windows
#  -e TASKWEAVER_GID=$(id -g) \ # uncomment if your host OS is not Windows
  --mount type=bind,source=<your_local_project_dir>,target=/app/TaskWeaver/project/ \
  taskweavercontainers/taskweaver-all-in-one:latest

Then you can edit the taskweaver_config.json file in your local project directory to configure TaskWeaver. In addition, you also can customize the plugins and examples in your local project directory. The structure of the project directory can be referred to the taskweaver/project directory.

How to access your local files in the container​

You can mount your local directory to the container. For example, you can use the following command:

docker run -it -e LLM_API_BASE=<API_BASE> \
  -e LLM_API_KEY=<API_KEY> \
  -e LLM_API_TYPE=<API_TYPE> \
  -e LLM_MODEL=<MODEL> \
#  -e TASKWEAVER_UID=$(id -u) \ # uncomment if your host OS is not Windows
#  -e TASKWEAVER_GID=$(id -g) \ # uncomment if your host OS is not Windows
  --mount type=bind,source=<your_local_dir>,target=/app/TaskWeaver/local/ \
  taskweavercontainers/taskweaver-all-in-one:latest

Then you can access your local files in the container by visiting the /app/TaskWeaver/local/ directory. You can load a file under the /app/TaskWeaver/local/ directory in the TaskWeaver CLI with the /load command. For example, you can load a file named example.csv by running the following command:

 TaskWeaver ▶  I am TaskWeaver, an AI assistant. To get started, could you please enter your request?
    Human   ▶  /load /app/TaskWeaver/local/example.csv
    Human   ▶  display the column names of the loaded file