Mendy Giter

Software Engineer | Architect

[Mendy Giter @mendygiter]

@rahulluthra22 This looks awesome, Please send

[Mendy Giter @mendygiter]

@benln ME! Best time to work without interruption please add me

[Mendy Giter @mendygiter]

Been working on these 2 item enrichment services for a couple weeks now, kept getting bugs and could not get it to work.

At last finally pushed it to prod with no error!

[Mendy Giter @mendygiter]

@jpschroeder @mendywol

[Mendy Giter @mendygiter]

I'm back at least for a bit 😜

Finally got my download image service working, I set up a task queue in gcp that gets called every time an items gets added and downloads all the product images to that product folder.
#GCP

[Mendy Giter @mendygiter]

Small for loop I wrote to batch the processing of downloading each product image for each item.

[Mendy Giter @mendygiter]

It's been a while but I am back,

My first @googlecloud Task to download the images from my DB for each product, and save them to the bucket. ⬇️

[Mendy Giter @mendygiter]

it's been while, only successfully deployed the backend will have to continue tomorrow.

💤💤💤

[Mendy Giter @mendygiter]

Till now I was building & testing locally, I realize now that is not an option going further as I am doing some scraping functions.

Let's see how long this takes!!!

[Mendy Giter @mendygiter]

@pinecone All 🐞 resolved!

Successfully using NL and returning results back to the user

[Mendy Giter @mendygiter]

After 4 days of trying to get my semantic search feature to work, we finally have success only to get this response🫠

[Mendy Giter @mendygiter]

So funky getting responses from @pinecone but it's not returning it to me, rather saying no matches found when I see it returned 5 rows.

[Mendy Giter @mendygiter]

Good news, bug resolved!! Pincone docs are a bit off on how to upsert and create embeddings for records.
Took about 5hr to resolve great learning experience, Shout out to @pinecone with a great tool using NL to get info on docs

Working code ↓

[Mendy Giter @mendygiter]

Ughh keep getting bugs trying to upsert to the vector db, issue is around how I am passing all my params into the chunk_text field to create the embeddings and ultimately the vectors.

[Mendy Giter @mendygiter]

Ughh it's been a long night, made progress from where we were a few hours ago but still not resolved.

This is where we at: Failed to upsert to Pinecone: TypeError: records is not iterable🤦‍♂️

Good night🍻

[Mendy Giter @mendygiter]

Decided against using @OpenAI embeddings, instead using @pinecone managed embeddings which allows me to create new embedding and upsert in one function. Thanks @tejchilli for the tip.

Added Pinecone connection and successfully connected to my index.

[Mendy Giter @mendygiter]

@tejchilli @OpenAI @pinecone @MongoDB Thats awesome, thanks for the heads up

[Mendy Giter @mendygiter]

1/2 Core functionality of the app is built🥳

Now the real fun begins, building functionality to enable users to use NL for search, this requires semantic search functionality and a vector DB.

[Mendy Giter @mendygiter]

It's been a while since I have built an app, I finally got the euphoric feeling again when I saw all my work displayed on a webpage and working.

Note: don't forget this feeling