This project contains the Price Finder API developed and deployed to Google Cloud Platform (GCP) which facilitates the image recognition capabilities through a vehicle image classification (CNN) model and market price retrieval of the Price Finder mobile application.
- Intergration of Github Actions CI/CD pipeline for automated deployment to GCP.
- Automated tests using Github Actions, which runs on every push to the develop branch.
- Code test coverage of 89% which includes functional and unit tests.
- Modular project structure to facilitate seamless scalablitiy with flask blueprints and application factory pattern.
- Comprehensive exception handling to gracefully handle exceptions occured due to both client and server side issues.
- Detailed explanation of the code functionality through Docstrings, comments and documentation.
- Code standards maintained in accordance with PEP8 style guide.
Framework / Library | Functionality |
---|---|
Flask - 2.1 | Develop the API functionality |
Tensorflow - 2.9 | Run the image recognition model |
Requests - 4.11 | Retrieve current listings of the identified vehicle |
Beautiful Soup - 4.11 | Retrieve the current market price |
Clone the repository
- Navigate to a folder in which you would like to setup the project.
- Open up a terminal in that folder and enter the command below to clone the repository.
git clone https://github.com/donheshanthaka/Price-Finder-Flask-API.git
Step 01:
- Navigate to the
Price-Finder-Flask-API
folder in terminal using the command below.
cd Price-Finder-Flask-API
Step 02:
- Create the python virtual environment
python -m venv env
Step 03:
- Navigate to the activation path
cd env/Scripts
- Activate the virtual environment. (Run either one, not both)
activate.bat //In CMD
Activate.ps1 //In Powershell
Step 04:
- Move back to the project folder
cd ../../
- Install the required dependencies
pip install -r requirements.txt
Setup environment variables
- Setup flask app
set FLASK_APP=main.py
- Setup flask environment (development is used since the app would be running locally)
set FLASK_ENV=development
📌 To change the environment type : set FLASK_ENV=development/production
use either development
or production
.
Start the local server
flask run --host=192.168.1.100 --port=8000
--host=192.168.1.100
> The server listens to requests on the given IP address--port=8000
> The server will listen on port 8000
Prerequisites:
📌 Note: If you have Windows 10 version 1803 or later, cURL is installed by default
The current version of the api supports indentifying vehicle models and retrieving their current market price which can be accessed through a single end point. An example of that is given below.
Getting vehicle information
- Making a post request uisng
cURL
to the api endpoint/get-vehicle-info
with an image attached to the body of the request. - Use the command below in a terminal on the project root directory.
📍 Make sure the local server is running before running the command below
curl -L -X POST "http://192.168.1.100:8000/get-vehicle-info" -F imageFile=@tests/images/4.jpeg
- A JSON response will be returned with the identified model of the vehicle and it's current market value.
Response:
{"model":"Toyota Aqua 2014", "price":"RS. 6,754,400"}
📌 Note: The "Price" value returned in the response will vary since it is retrieved everytime from the web when the api is being called.
The same endpoint as above accessed using a python script:
- Run this script on the project root folder.
import requests
url = "http://192.168.1.100:8000/get-vehicle-info"
payload={}
files=[
('imageFile',('4.jpeg',open('tests/images/4.jpeg','rb'),'image/jpeg'))
]
headers = {}
response = requests.request("POST", url, headers=headers, data=payload, files=files)
print(response.text)
The HTTP Status Codes used by the Price Finder API.
HTTP Status Code | Description |
---|---|
200 OK | Successfully identified the image and retrieved the market price. |
400 Bad Request | Image file not found in the request. |
404 Not Found | The requested resource was not found. |
405 Not Found | The requested method is not allowed. |
415 Unsupported Media Type | Invalid Image Type. |
502 Bad Gateway | Unable to access price retrieval web server. |
Identify vehicle and retrieve current market price.
POST /get-vehicle-info
Parameter | Type | Description |
---|---|---|
imageFile |
jpeg /png |
Required. Image to be identified |
Coverage Report
File | Stmts | Miss | Cover | Missing |
---|---|---|---|---|
app | ||||
init.py | 13 | 3 | 77% | 13–17 |
error_handlers.py | 25 | 1 | 96% | 16 |
utils.py | 45 | 8 | 82% | 52, 85–86, 99, 114–119 |
TOTAL | 109 | 12 | 89% |
The flask api is tested in both unit tests and functional tests using the pytest framework.
Functional Tests:
Tests the functionality of the api endpoint /get-vehicle-info
which takes in an image as a parameter and return the vehicle model and price as a JSON object.
Functional test modules:
-
test_get_vehicle
-> Given a flask application, when the '/get-vehicle-info' is requested (POST), check that a '200' response code is returned with valid response data. -
test_get_vehicle_without_image
-> Given a flask application, when the '/get-vehicle-info' is requested (POST) without an image attached in the body, then check that a '400' status code is returned. -
test_get_vehicle_invalid_image_type
-> Given a flask application, when the '/get-vehicle-info' is requested (POST) with an invalid image type, then check that a '415' status code is returned. -
test_page_not_found
-> Given a flask application, when an invalid URL / Endpoint is requested (POST), then check that a '404 Not Found' status code is returned. -
test_method_not_allowed
-> Given a flask application, when an invalid request is made to a valid endpoint (GET), then check that a '405 Method Not Allowed' status code is returned.
Unit Tests:
Tests the each individual functions used by the api, such as price retrieval, image recognition and image reshaping for the cnn model.
Unit test modules:
-
test_get_price
-> Given a name of vehicle, when trying to find the current market price, then check the return value is a string (cannot check for an exact value since market value is not constant,therefore checking the return type is the only option available). -
test_predict
-> Given a path to an image of a vehicle, when trying to identify the vehicle, then check the identified vehicle is correct according to the given image. -
test_reshape_image
-> Given a path to an image, when trying to predict the image, then check the returned image tensor is in correct shape.
A code test coverage of 89% is achieved with the implementation of above test cases.
Prerequisites:
-
Navigate to the root folder of the project and activate the python virtual environment created during the setup process.
Step 01:
- Install pytest and coverage.
pip install pytest, coverage
Step 02:
- Run the test modules.
coverage run --source=app -m pytest
output:
================================== test session starts =========================================
platform win32 -- Python 3.9.5, pytest-7.1.2, pluggy-1.0.0
collected 8 items
tests\functional\test_get_vehicle_info.py ... [ 37%]
tests\functional\test_status_codes.py .. [ 62%]
tests\unit\test_get_price.py . [ 75%]
tests\unit\test_predict.py . [ 87%]
tests\unit\test_reshape_image.py . [100%]
================================== 8 passed in 7.28s ============================================
Step 03:
- Checking the test coverage report.
coverage report
output:
Name Stmts Miss Cover
-------------------------------------------
app\__init__.py 13 3 77%
app\error_handlers.py 25 1 96%
app\utils.py 45 8 82%
app\views.py 26 0 100%
-------------------------------------------
TOTAL 109 12 89%
The API developed in this project is deployed in Google Cloud Platform (GCP) to be accessed by the price finder mobile application. Furthermore, the deployment process is fully automated with the use of github actions CI/CD pipleline.
Prerequisites:
- Google Cloud Platform (GCP) account.
📌 Note: You have to activate the billing feauture of the account by providing your credit / debit card details and it will charge you 1-2 USD to verify your account, but that amount will be refunded and you will not have to turn on billing for the project itself and therefore, you will not be charged for the usage during this project and everything will be under the free usage limits. Moreover, you will recieve 300 USD free credits valid for 3 months.
Step 01:
- Go to the GCP console and create a new project (Make sure to give a unique relatable name for your service).
Example: vehicle-price-finder-001
📌 Note: You will not be able to use the example above since every project has to have a unique name therefore, provide a unique name and keep it noted since it will be required in the steps below.
Step 02:
Activate Cloud Run
and Cloud Build
API
- In the Google Cloud console, go to APIs & services for your project.
- Click on the Library page and Search
Cloud Run
on the search box. - Select the API and select Enable.
- Activate
Cloud Build
api with the same steps.
Step 01:
Download and install Google Cloud CLI using this guide by Google and complete the installing and initializing sections.
Step 01:
Build the docker filer in Cloud Build
PROJECT-ID
-> Can be found in the project dashboard in Google Cloud console. Edit this value before executing the command below.
- Run the command below in a terminal in project root directory
gcloud builds submit --tag gcr.io/{PROJECT-ID}/get-prediction
Step 02:
Deploying to Cloud Run
gcloud run deploy --image gcr.io/{PROJECT-ID}/get-prediction --platform managed
Service name
-> Provide a service name for the api, use get-prediction for the current project.region
-> Select a region based on your location with the help of this documentation for the current project i have usedasia-south1
[6].allow unauthenticated invocations
-> select 'y'
After successful deployment, the Service URL will be displayed at the bottom of the terminal. Copy that for future use since that will be the api access point.
Before following the steps below, fork this repository to your own github account.
And then make sure to complete the sections mentioned below
Step 01:
Create a Google Cloud Service Account.
- Go to google cloud console
- Click navigation menu on top left corner.
IAM & Admin
->Service Accounts
->Create Service Account
- Service Account Name:
api-service-account
Step 02:
Grant the Google Cloud Service Account permissions mentioned below to access Google Cloud resources.
Cloud Run Admin
Cloud Run Service Agent
Cloud Build Service Agent
Viewer
Step 03:
Enable the IAM Credentials API.
gcloud services enable iamcredentials.googleapis.com --project "{PROJECT-ID}"
Step 04:
Create a Workload Identity Pool.
Format:
gcloud iam workload-identity-pools create `"WORKLOAD-ID-POOL-NAME"` \
--project=`"PROJECT_ID"` \
--location="global" \
--display-name=`"DISPLAY_NAME_FOR_POOL"`
Example:
gcloud iam workload-identity-pools create "price-finder-pool" --project "{PROJECT-ID}" --location="global" --display-name="price finder pool"
Step 05:
Get the full ID of the Workload Identity Pool.
Format:
gcloud iam workload-identity-pools describe `"WORKLOAD-ID-POOL-NAME"` \
--project=`"PROJECT_ID"` \
--location="global" \
--format="default"
Example:
gcloud iam workload-identity-pools describe "price-finder-pool" --project "{PROJECT-ID}" --location="global" --format="default"
Return format:
`WORKLOAD_IDENTITY_POOL_ID` = projects/`YOUR-PROJECT-NUMBER`/locations/global/workloadIdentityPools/`"WORKLOAD-ID-POOL-NAME"`
Step 06:
Create a Workload Identity Provider in that pool.
Format:
gcloud iam workload-identity-pools providers create-oidc `"WORKLOAD-ID-POOL-PROVIDER-NAME"` \
--project=`"PROJECT_ID"` \
--location="global" \
--workload-identity-pool=`"WORKLOAD-ID-POOL-NAME"` \
--display-name=`"DISPLAY_NAME_FOR_PROVIDER"` \
--attribute-mapping="google.subject=assertion.sub,attribute.actor=assertion.actor,attribute.repository=assertion.repository" \
--issuer-uri="https://token.actions.githubusercontent.com"
Example:
gcloud iam workload-identity-pools providers create-oidc “price-finder-provider” --project "{PROJECT-ID}" --location="global" --workload-identity-pool="price-finder-pool" --display-name="price finder provider" --attribute-mapping="google.subject=assertion.sub,attribute.actor=assertion.actor,attribute.repository=assertion.repository" --issuer-uri="https://token.actions.githubusercontent.com"
Step 07:
Allow authentications from the Workload Identity Provider originating from your repository to impersonate the Service Account created above.
Format:
gcloud iam service-accounts add-iam-policy-binding "my-service-account@${PROJECT_ID}.iam.gserviceaccount.com" \
--project=`"PROJECT_ID"` \
--role="roles/iam.workloadIdentityUser" \
--member="principalSet://iam.googleapis.com/`WORKLOAD_IDENTITY_POOL_ID( RETURN VALUE FROM STEP 5)`/attribute.repository/`yourgithubname/reponame`"
Example:
gcloud iam service-accounts add-iam-policy-binding "test-service-account@{PROJECT-ID}.iam.gserviceaccount.com" --project "{PROJECT-ID}" --role="roles/iam.workloadIdentityUser" --member="principalSet://iam.googleapis.com/`WORKLOAD_IDENTITY_POOL_ID( RETURN VALUE FROM STEP 5)`/attribute.repository/`yourgithubname/reponame`"
Step 08:
Extract the Workload Identity Provider resource name.
Format:
gcloud iam workload-identity-pools providers describe "my-provider" \
--project=`"PROJECT_ID"` \
--location="global" \
--workload-identity-pool=`"WORKLOAD-ID-POOL-NAME"` \
--format="default"
Example:
gcloud iam workload-identity-pools providers describe "price-finder-provider" --project="{PROJECT-ID}" --location="global" --workload-identity-pool=“price-finder-pool” --format="default"
Return format:
projects/`YOUR-PROJECT-NUMBER`/locations/global/workloadIdentityPools/`WORKLOAD-ID-POOL-NAME`/providers/`WORKLOAD-ID-POOL-PROVIDER-NAME`
⭕ Important
YOU NEED TO USE THIS VALUE AS "workload_identity_provider" in main.yml
file, located in .github\workflows
directory under "Authenticate to Google Cloud" section.
📌 Note: When replacing the workload_identity_provider, remove the $
and curly brackets. Just include the value within single quotes.
Step 09:
Making the service public (Allow unauthenticated).
Format:
gcloud run services add-iam-policy-binding [SERVICE_NAME] \
--member="allUsers" \
--role="roles/run.invoker" \
--project=`"PROJECT_ID"`
SERVICE_NAME
= Name of the cloud run service
Example:
gcloud run services add-iam-policy-binding get-prediction --member="allUsers" --role="roles/run.invoker" --project="{PROJECT-ID}"
Step 10:
Update the cloudbuild.yaml
file with current project details.
- Update the
PROJECT-ID
in the section below and replace the content incloudbuild.yaml
file in the project root directory.
steps:
- name: 'gcr.io/cloud-builders/docker'
args: ['build', '-t', 'gcr.io/{PROJECT-ID}/get-prediction', '.']
images: ['gcr.io/{PROJECT-ID}/get-prediction']
Step 11:
Update the main.yml
file in .github\workflows
'directory.
PROJECT_ID
-> Replace with your own project idREGION
-> If necessary, replace it with a region suitable for you or keep as it is.service_account
-> Can be found under IAM & Admin > Service accounts (Example: [email protected])
📌 Note: When replacing the service account, remove the $
and curly brackets. Just include the value within single quotes.
Step 12:
- Switch to the
release
branch - Commit the new changes and push to the repository.
You will see the github actions running the workflow process under Actions
tab in repository.
Here are some related projects