perlucas/challenge-speech-api

Speech recognition test

Speech Recognition Test API

Basic REST API for transcribing and finding specified words within MP3 audio files:

• Based on the Google Speech API solution
• Built in Python 3 and MongoDB
• Built using a Docker multi-container environment

Set up instructions

1. Clone this repository
2. Create a credentials.json file under the root directory, containing the Google Cloud credentials
3. Create a .env file (rename the provided example)
4. Run docker compose up --build
5. Test this API in localhost:85/

API endpoints & examples

POST /scan

Given a MP3 audio file and a set of words, processes the given file and looks up the words in the transcription. Returns the number of ocurrences and timestamps found for each word.

JSON body examples:

{
	"file": "gs://speech-api-test-2344/PartyAnimal.mp3",
	"words": [
		"party",
		"not",
		"people"
	]
}

{
	"file": "gs://speech-api-test-2344/GreetingsandIntroductions.mp3",
	"words": [
		"athlete",
		"architect",
		"people"
	]
}

{
	"file": "gs://speech-api-test-2344/AirPollution.mp3",
	"words": [
		"pollution",
		"think",
		"people"
	]
}

HTTP 200 Response examples:

{
	"results": [
		{
			"count": 5,
			"timestamps": [
				"00:00:12.1",
				"00:00:14.8",
				"00:00:16.3",
				"00:00:33.8",
				"00:00:59.2"
			],
			"word": "pollution"
		},
		{
			"count": 1,
			"timestamps": [
				"00:00:10.6"
			],
			"word": "think"
		},
		{
			"count": 0,
			"timestamps": [],
			"word": "people"
		}
	]
}

HTTP 500 Response examples:

{
	"message": "No such object: speech-api-test-2344/Greetings and Introductions Basi.mp3"
}

GET /logs

Returns the latest logged in actions for this API.

HTTP 200 Response examples:

{
	"logs": [
		{
			"metadata": {
				"endpoint": "GET /logs",
				"user": "tester"
			},
			"timestamp": "2023-01-28 16:06:22",
			"type": "ACTION_APP_ACCESS"
		},
		{
			"metadata": {
				"endpoint": "POST /scan",
				"user": "tester"
			},
			"timestamp": "2023-01-28 16:06:22",
			"type": "ACTION_APP_ACCESS"
		},
		{
			"metadata": {
				"file": "gs://speech-api-test-2344/PartyAnimal.mp3",
				"words": [
					"party",
					"not",
					"people"
				]
			},
			"timestamp": "2023-01-28 16:06:22",
			"type": "ACTION_SCAN_AUDIO"
		},
	],
	"page": 1,
	"page_size": 50
}

TODO's

The following list contains additional features and capabilities that I've decided not to include in the current task's scope so as not to extend its delivery unnecessarily:

• Implement basic data parameters validation:
  • validate that words is a list of strings
  • validate the file parameter
• Add basic authentication by username and password
• Allow API users to configure audio scanning parameters such as audio language and rate in Hertz
• Implement pagination for logs endpoint
• Implement logs filtering on action type, users, timestamp and metadata
• Run this API using an isolated WSGI server running in its own container instead of using the Flask development server