README.md

Speech to Code API 🤗

FastAPI implementation of the web Speech to Code API.

Get started

Prerequisites

This package uses Poetry for dependency management.

Running the API

1. Install the api using poetry install.
2. Run the API locally using poetry run start. (this calls api.main:start)

You will then be able to access the API on http://0.0.0.0:8000/.

Note: You can control the port using the PORT environment variable. :)

Routes

POST /task

This is the entrypoint of the Speech to Code API. It accepts audio file upload via the request body form-data. To use it, hit it with a post request with a WAV-file submitted under the audio key.

This will create a new task, which is immediately queued.

Example request

Using curl, a request can be submitted with:

curl -X POST -F "[email protected]" http://0.0.0.0:8000/task

This will return a HTTP 200, with contents similar to:

{
    "task_id": "c92ecf6e14814c2a9289a3e276797e4c"
}

GET /task/<id>

Once a task is queued, it may take some time for it to start processing, and subsequently to finish. This endpoint gets the current status of the task, and will provide the task result once it's done.

Simply send a GET request to the endpoint corresponding to the task_id received from POST /task to observe the status.

Example request

You can use this endpoint by curl'ing the route. Like this:

curl http://0.0.0.0:8000/task/c92ecf6e14814c2a9289a3e276797e4c

This will respond with a HTTP 200, containing the following (in chronological order):

1. In the queued state.

{
    "status": "queueing",
    "result": null
}

2. When a processing worker has picked up the task.

{
    "status": "processing",
    "result": null
}

3. When the task is done processing. 🎉

{
    "status": "done",
    "result": "<code from speech>"
}

Docker

Building the Docker image must be done from the root (.. from here) context to include the common package. You can build the API image from the root directory containing api/, like this:

docker build -t speech-to-codeapi -f api/docker/Dockerfile .

Note: This will require mounting relevant Azure credentials, or passing environment credentials to the container.