Resolving inter-service communication issue

This week was heck of a roller coaster of emotions, trying to solve an issue. All I faced is the same error again and again. Precisely, error code 307 and 422 that is Temporary Redirect Unprocessable Entity respectively.

Started with a simple notebook for gateway service just like I did for ingestion service at its starting phase. But here is a catch, this time i am dealing gateway service, meaning I need to use uvicorn command here. In case of ingestion service, when I started, all I did is assign a file path to a variable, pass it to PyPDFLoader, split the returned Document into chunks and store them into a storage, called as vector store.

I used uvicorn when I was ready with all the class definitions and folder structures, just to wrap the core script around fastapi and deal with api testing using postman. The functionality that I assigned for gateway is different. As the name suggests, it shall act as a mediator between the client and the backend. For that I need to connect gateway service to ingestion service.

I installed all the necessary libraries in a virtual environment, namely fastapi, uvicorn and httpx, activated the virtual environment. I had an impression that i need use schema validation for the incoming data, convert it to dictionary using .model_dump() and then pass it to httpx.AsyncClient().post(), since i am using post operation.

The post operation has been used because the file shall be uploaded using postman, shall help to store the embeddings in a database. That uploaded file shall be of type UploadFile imported from fastapi library. At the end i stated uvicorn.run() statement to activate my gateway service..

This is where the story begins...

I encountered my first issue in jupiter notebook running a uvicorn statement from where gateway service shall begin to run. I forgot that even jupiter notebook run on a server and since i am running it locally, of course it shall run on localhost.

What I discovered is that while stating on jupyter it is incorrect to use uvicorn statement as:

The reason the first method did helped because a jupyter notebook is not just a text document, it is an active web application. It requires its own constantly running event loop in the background to handle cell executions, process outputs, and communicate with your browser. When I called uvicorn.run(), uvicorn attempts to start a brand new event loop to listen for incoming web requests. Python's standard asyncio library strictly forbids starting a new event loop while one is already running in the same thread. It throws the RuntimeError to prevent the two managers from fighting over control.

By thread I mean a worker which executes a set of instructions available as a recipe.

instead the following code fragment needs to be used:

A simple screenshot captured from my github repo. The output was previously showcased after running that cell and then commented it out.

Another alternative can by stating the following in a notebook cell :

%%writefile main.py
from fastapi import FastAPI
app = FastAPI()

@app.get("/")
def read_root():
    return {"Hello": "World"}

this shall automatically create a main.py file sibling to the notebook file and the following uvicorn command can be used:

!uvicorn main:app --port 8001 --reload

However my issue has not finished yet. This was just the first part.
The second part is while uploading a pdf in postman, and clicking on Send button, I faced a series on errors one after another.

now even after making some amendments in jupyter notebook, i encountered the error which is shown previously in my codebase. I had to restart the session again and again, may be because of the fact that the previously compiled python script persisted even after making changes. So i shifted to simple .py file which was a better option.

After I was fed up of facing the same error for 2 days, I browsed in google "do i need schema validation using pydantic when a function param of type fastapi.UploadFile, wrapped around the given pydantic model and apply .model_dump() just to be passed in httpx.AsyncClient() ?"

The answer was, only when multiple parameters are passed in the function, in this the function name is forward_to_ingestion, but this time i have only 1 param.

Plus I was mixing things up unncessarily putting .model_dump() after wrapping Around IngestionSchema pydantic class not reallizing that I already put UploadFile at the function header to begin with. I removed the schema defintion then.

At that time i used .model_dump, hoping this would return a json data, but where is the key, i completely forgot that it is simply a variable incoming before which the content within the uploaded file needs to read. At least i did the right thing to use await and .read() for that.

The result is still the same..made me check what mistake did i commit in ingestion service, in an isolated fashion, no inter service communication in this case. Was running fine..

i passed the same snippet to AI, told me to use

tempfile.NamedTemporaryFile

helps to keep memory leakage in check and for security reasons.

meanwhile i also faced 307 error code..specifying /ingest/

I removed the trailing slash in postman, after I crashed into another error

The terminal showed this

One thing i was sure there is something still missing in my gateway service. I passed this snippet:

it simply added follow_redirects=True because of the fact that

httpx does not follow HTTP redirects by default, which is crucial for forwarding files and post data.

I then encountered another problem this time it timeout error, not shown in postman but in terminal.

In postman the same internal server error is shown.

I then specified timeout argument as well.

And I did not believe, it really worked...after seeing this in postman

Yes it is my fault that i should have specified status code as 201 since 200 is the default status code under gateway service, which helps in debugging.

whereas the terminal where ingestion was active shown this as output

But this taught me one thing is to NEVER GIVE UP!!