PlugBear Python SDK

Connect a custom LLM to PlugBear through a web server.

FastAPI

Installation

To install the PlugBear Python SDK, run the following command:

pip install 'plugbear[fastapi]'

Usage

Here's a simple example to get you started:

importcontextlibimportplugbear.fastapifromfastapiimportFastAPIPLUGBEAR_API_KEY=""PLUGBEAR_ENDPOINT="/plugbear"@contextlib.asynccontextmanagerasyncdeflifespan(app: FastAPI):
awaitplugbear.fastapi.register(
app,
llm_func=some_llm,
api_key=PLUGBEAR_API_KEY,
endpoint=PLUGBEAR_ENDPOINT,
)
yieldapp=FastAPI(lifespan=lifespan)
asyncdefsome_llm(context: plugbear.fastapi.Request) ->str:
# template prompt using `context` to your own LLM# and return resultresult: str= ...
returnresult

For versions below 0.93.0

importplugbear.fastapifromfastapiimportFastAPIapp=FastAPI()
PLUGBEAR_API_KEY=""PLUGBEAR_ENDPOINT="/plugbear"@app.on_event("startup")asyncdef_startup():
awaitplugbear.fastapi.register(
app,
llm_func=some_llm,
api_key=PLUGBEAR_API_KEY,
endpoint=PLUGBEAR_ENDPOINT,
)
asyncdefsome_llm(context: plugbear.fastapi.Request) ->str:
# template prompt using `context` to your own LLM# and return resultresult: str= ...
returnresult

Other Web server frameworks

Although it does not yet provide an interface as convenient as FastAPI, it can be directly linked to PlugBear.

Installation

pip install plugbear

Usage

Taking Django as an example, it looks like this:

settings.py

importplugbearfromasgiref.syncimportasync_to_syncPLUGBEAR_API_KEY=""PLUGBEAR_ENDPOINT="/plugbear"pb=async_to_sync(plugbear.PlugBear.init(api_key=PLUGBEAR_API_KEY))

urls.py

fromdjango.confimportsettingsfromdjango.urlsimportpathfrom . importviewsurlpatterns= [
path(settings.PLUGBEAR_ENDPOINT, views.handle_plugbear, name="plugbear"),
]

views.py

importplugbearfromdaciteimportfrom_dictfromdjango.httpimportHttpResponsedefhandle_plugbear(request):
context=from_dict(data_class=plugbear.Request, data=request.data)
result=some_llm(context)
returnHttpResponse(result)
defsome_llm(context: plugbear.Request) ->str:
# template prompt using `context` to your own LLM# and return resultresult: str= ...
returnresult