Automating Client Generation

[timonrieger]

I’m wondering if there would be interest in automating the Python client generation using https://openapi-generator.tech/. Since the Lichess server is OpenAPI 3–compatible, the client could be generated directly from the spec.

Some benefits of this approach:

• Ensures the client stays in sync with the API
• Reduces manual maintenance and boilerplate
• Improves consistency across endpoints

Depending on how the current client is structured, this might introduce breaking changes.

[timonrieger]

also relates to #18

[kraktus]

We initially wanted to but at the time berserk there wasn’t any code generator that supported openapi 3.1

[timonrieger]

We initially wanted to but at the time berserk there wasn’t any code generator that supported openapi 3.1

so noop?

[kraktus]

not against it, rather for it, but all would depend on the resulting code quality and API for the users. would it handle timestamp to date time conversion, etc.

[timonrieger]

not against it, rather for it, but all would depend on the resulting code quality and API for the users. would it handle timestamp to date time conversion, etc.

i see. What other "custom" implementations are in the current package next to the timestamp conversions? We run an API for data scientist (which prefer pandas Dataframes 😄 ), so we wrap some responses before returning to the user.

I can show you the implementation for this if you want.

[fitztrev]

What other "custom" implementations

ndjson handling might be one of the non-standard things that the client would need to do

I use a typescript library that uses types generated by the openapi spec and that's one of the modifications that's needed

[timonrieger]

ndjson handling might be one of the non-standard things that the client would need to do

is that used for the streams? can you point me to the current implementation?

[gameroman]

ndjson handling might be one of the non-standard things that the client would need to do

is that used for the streams? can you point me to the current implementation?

it uses https://pypi.org/project/ndjson

but in my project i just inlined it