Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/pottava/chatbot-on-googlecloud


https://github.com/pottava/chatbot-on-googlecloud

Last synced: over 1 year ago
JSON representation

Awesome Lists containing this project

README 

          
A chatbot sample on Google Cloud

===



## Basic configurations



### 1.1. Define an environment



Set your project as environment variables.



```bash

gcloud config set project "## Your project ID here ##"

export PROJECT_ID=$( gcloud config get-value project )

export PROJECT_NUMBER=$( gcloud projects list --filter="${GOOGLE_CLOUD_PROJECT}" --format="value(PROJECT_NUMBER)" )

```



Specify Tokyo as a region we use.



```bash

export GOOGLE_CLOUD_REGION=asia-northeast1

```



It is recommended to create resources with different environment names in this tutorial, like development and production. We will assume the development environment and prefix it with `dev`.



```bash

export ENVIRONMENT_NAME=dev

```



If multiple users are using one project, try using the verified email address as the username.



```bash

export ACCOUNT_EMAIL=$( gcloud auth list --filter 'status:ACTIVE' --format 'value(account)' )

export ENVIRONMENT_NAME="dev-$( echo ${ACCOUNT_EMAIL} | cut -d "@" -f 1 )"

```



### 1.2. Enable Google Cloud APIs



In Google Cloud, you can use resources by enabling **Service APIs** for each project. Now please enable the services you need.



```bash

gcloud services enable compute.googleapis.com run.googleapis.com bigquery.googleapis.com \

    discoveryengine.googleapis.com storage.googleapis.com artifactregistry.googleapis.com \

    cloudbuild.googleapis.com cloudresourcemanager.googleapis.com

```



## Basic services



### 2.1. Google Cloud Storage



Create a bucket to upload the files that will be used as evidence for the RAG system.



```bash

export BUCKET_NAME="${ENVIRONMENT_NAME}-rag-storage"

gcloud storage buckets create "gs://${BUCKET_NAME}" --location "${GOOGLE_CLOUD_REGION}" \

    --uniform-bucket-level-access --public-access-prevention --enable-autoclass

```



Download a suitable file as sample.pdf and upload it to your bucket.



```bash

gcloud storage cp sample.pdf gs://${BUCKET_NAME}/rag/jp/ja/car/2024/manual/sample.pdf

```



### 2.2. Vertex AI Search



1. Go to https://console.cloud.google.com/gen-app-builder/start and enable it.

2. Select "Search" under "Select app type"

3. Enter your app name, company or organization name as appropriate, and select "Global" as the app location

4. Proceed to creating a data store, select "Cloud Storage" as the data source, specify the folder to import, and click "Create".

5. Return to the app creation screen, select the data source and click "Create"



This will create a vector database and set up a search system.



### 2.3. IAM service account



Set the permissions required for your Cloud Run application.



```sh

gcloud iam service-accounts create "${ENVIRONMENT_NAME}-chatbot" \

    --description "Service Account for Chatbot applications"

gcloud projects add-iam-policy-binding "${PROJECT_ID}" \

    --member "serviceAccount:${ENVIRONMENT_NAME}-chatbot@${PROJECT_ID}.iam.gserviceaccount.com" \

    --role "roles/discoveryengine.editor"

gcloud projects add-iam-policy-binding "${PROJECT_ID}" \

    --member "serviceAccount:${ENVIRONMENT_NAME}-chatbot@${PROJECT_ID}.iam.gserviceaccount.com" \

    --role "roles/bigquery.dataEditor"

gcloud storage buckets add-iam-policy-binding "gs://${BUCKET_NAME}" \

    --member "serviceAccount:${ENVIRONMENT_NAME}-chatbot@${PROJECT_ID}.iam.gserviceaccount.com" \

    --role "roles/storage.admin"

```



### 2.4. Cloud Run



First, create a Cloud Run service using the sample application.



```bash

gcloud run deploy "${ENVIRONMENT_NAME}-chatbot" --region "${GOOGLE_CLOUD_REGION}" \

    --platform "managed" --cpu 1.0 --memory 512Mi --image gcr.io/cloudrun/hello \

    --service-account "${ENVIRONMENT_NAME}-chatbot@${PROJECT_ID}.iam.gserviceaccount.com" \

    --execution-environment gen2 --no-allow-unauthenticated

```



Grant permissions to allow communication if the request is from the service account created in 2.2.



```bash

gcloud run services add-iam-policy-binding "${ENVIRONMENT_NAME}-chatbot" --region "${GOOGLE_CLOUD_REGION}" \

    --member "serviceAccount:${ENVIRONMENT_NAME}-chatbot@${PROJECT_ID}.iam.gserviceaccount.com" \

    --role "roles/run.invoker"

```



Let's connect to the service. The following command proxies to the service.



```bash

gcloud beta run services proxy --region "${GOOGLE_CLOUD_REGION}" "${ENVIRONMENT_NAME}-chatbot"

```



In this state, access to http://localhost:8080 will be forwarded to the service on Cloud Run.



### 2.5. BigQuery



Create a dataset,



```bash

export BQ_DATASET_ID=${ENVIRONMENT_NAME//-/_}

bq --location "${GOOGLE_CLOUD_REGION}" mk --dataset "${BQ_DATASET_ID}"

```



And a table.



```bash

bq mk --table --description "Q&A table - ${ENVIRONMENT_NAME}" \

    --schema 'ts:TIMESTAMP,env:STRING,ver:STRING,q:STRING,a:STRING' \

    --time_partitioning_field ts --time_partitioning_type DAY \

    "${BQ_DATASET_ID}.qna"

```



## Developing GenAI applications



### 3.1. Start the API server



Check the data store ID on the console of Vertex AI Search, and export as a following environment variable.



```bash

export VAIS_DATASTORE_ID=

```



After passing Google Cloud authentication,



```bash

gcloud auth application-default login

```



Let's try running the application locally.



```bash

cd src

pnpm install

npm run lint

GOOGLE_CLOUD_PROJECT="${PROJECT_ID}" PORT=9000 npm start

```



Try execute the following command from another terminal.



```bash

curl -sXPOST -H 'Content-Type: application/json' -d '{"q":"Hello :)"}' http://localhost:9000/api/v1/chat

```



### 3.2. Docker containers



Let's build it into a docker container.



```bash

docker build -t chatbot .

```



Start it locally, and once you can confirm that it works properly as before,



```bash

docker run --name chatbot -d --rm -p 9000:8080 \

    -v "${HOME}/.config/gcloud:/gcp/config:ro" -e CLOUDSDK_CONFIG=/gcp/config \

    -e GOOGLE_APPLICATION_CREDENTIALS=/gcp/config/application_default_credentials.json \

    -e GOOGLE_CLOUD_PROJECT="${PROJECT_ID}" -e VAIS_DATASTORE_ID="${VAIS_DATASTORE_ID}" \

    -e BQ_DATASET_ID="${BQ_DATASET_ID}" -e CURRENT_VERSION="docker" \

    chatbot

docker logs -f chatbot

```



Remove the container.



```bash

docker rm -f chatbot

```



### 3.3. Artifact Registry



In the Artifact Registry, create a repository to store your application.



```bash

gcloud artifacts repositories create "${ENVIRONMENT_NAME}-chatbot" \

    --repository-format docker --location "${GOOGLE_CLOUD_REGION}" \

    --description "${ENVIRONMENT_NAME}'s chatbot"

```



Build the app and push it to the repository.



```sh

gcloud builds submit \

    --tag "${GOOGLE_CLOUD_REGION}-docker.pkg.dev/${PROJECT_ID}/${ENVIRONMENT_NAME}-chatbot/front:dev" \

    .

```



### 3.4. Deploy to Cloud Run



Let's run it on the cloud.



```bash

gcloud run deploy "${ENVIRONMENT_NAME}-chatbot" --region "${GOOGLE_CLOUD_REGION}" \

    --image "${GOOGLE_CLOUD_REGION}-docker.pkg.dev/${PROJECT_ID}/${ENVIRONMENT_NAME}-chatbot/front:dev" \

    --update-env-vars "GOOGLE_CLOUD_PROJECT=${PROJECT_ID},VAIS_DATASTORE_ID=${VAIS_DATASTORE_ID},BQ_DATASET_ID=${BQ_DATASET_ID}" \

    --timeout 60

```



You can proxy the service,



```bash

gcloud beta run services proxy --region "${GOOGLE_CLOUD_REGION}" "${ENVIRONMENT_NAME}-chatbot"

```



Then execute the following command from another terminal.



```bash

curl -sXPOST -H 'Content-Type: application/json' -d '{"q":"Can I use Bluetooth?"}' http://localhost:8080/api/v1/chat

```