IFSGuid Backend

Async AI Chat Application Powered by gpt4free on top of FastAPI

Setup

It contains two generic entity models Interaction and Message. We store all of the data in the Postgres database. The interaction with the database is done with the SQLAlchemy library, and the simple GET and POST endpoints are exposed via the API, which is written with the FastAPI framework.

To manage dependencies, we use poetry.

To launch an API instance, you should:

1. Have a running Postgres instance, e.g. in a container. The application will read the .env file to access the database.
2. Create a virtual environment and install the dependencies in it. You can run poetry install for that.
3. Use start_app.sh to run the server. By default, it will bind to http://localhost:8000.
4. An automatically generated documentation can be found at http://localhost:8000/docs. The endpoints are accessible at http://localhost:8000/api/<endpoint_name>.

You can also run the project via docker-compose (i.e. docker compose up -d) on port 80 in which you would need the .docker.env containing the following variable to create the database:

SQLALCHEMY_DATABASE_URI=postgresql+asyncpg://<username>:<password>@ifsguid_db/<db-name>

---

Here is a benchmark of the API using by wrk to demonstrate the performance of the service in different configuration:

Service	Loading Strategy	WRK Configuration	Throughput (reqs/sec)
Async	Joined	4 threads, 10 conns	132
Async	Selectin	4 threads, 10 conns	112
Sync	Lazy	4 threads, 10 conns	36
Sync	Joined	4 threads, 10 conns	132
Sync	Selectin	4 threads, 10 conns	114
Async	Joined	4 threads, 50 conns	159
Sync	Joined	4 threads, 50 conns	1
Async	Joined	4 threads, 15 conns	126
Sync	Joined	4 threads, 15 conns	69