blog-recommendation-ts

Blog recommendation

This is a example application showcasing a recommendation system for blogs.

Please check the repo's README for prerequisites for running this example.

This sample uses the Vector extension to sqlite, which is available with Turso.

Steps to use

• Create a Turso database using personal account and get configs

turso db create --enable-extensions
# Get the url
turso db show <db-name>
# Create access token
turso db tokens create <db-name> --expiration none 

• make a copy of runtime-config-template.toml as runtime-config.toml and fill in the details

• Shell into Turso Db to create virtual table using the vss module

$ turso db shell <db-name>
- CREATE virtual TABLE vss_blog_posts USING vss0(embedding(384));

$ npm install
$ spin  build -u --runtime-config-file runtime-config.toml --sqlite @migrations.sql

Note: If you are using the Cloud GPU component, remember to add the [llm_compute] section to the runtime-config.toml file.

• Create embeddings for all the posts

./update_embeddings.sh http://localhost:3000

• Query for similar articles

curl -X POST -d '{"blogPath": "scale-to-zero-problem"}' http://localhost:3000/getRecommendations

Deploy the application to Fermyon Cloud

$ spin deploy