Max•13mo ago

observationId in Python SDK

are you using langchain or the normal sdk?

6 Replies

Alex•13mo ago

Normal SDK, no langchain I have langfuse.model.CreateGeneration objects handy

Max•13mo ago

from langfuse.model import CreateGeneration, Usage, UpdateGeneration
 
from datetime import datetime
 
generationStartTime = datetime.now()
 
# chat_completion = openai.ChatCompletion.create(model="gpt-3.5-turbo", messages=[{"role": "user", "content": "Hello world"}])
# ...
 
generation = trace.generation(CreateGeneration(
    name="summary-generation",
    startTime=generationStartTime,
    endTime=datetime.now(),
    model="gpt-3.5-turbo",
    modelParameters={"maxTokens": "1000", "temperature": "0.9"},
    prompt=[{"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Please generate a summary of the following documents \nThe engineering department defined the following OKR goals...\nThe marketing department defined the following OKR goals..."}],
    metadata={"interface": "whatsapp"}
))

generation.id # <- id for observations

from langfuse.model import CreateGeneration, Usage, UpdateGeneration
 
from datetime import datetime
 
generationStartTime = datetime.now()
 
# chat_completion = openai.ChatCompletion.create(model="gpt-3.5-turbo", messages=[{"role": "user", "content": "Hello world"}])
# ...
 
generation = trace.generation(CreateGeneration(
    name="summary-generation",
    startTime=generationStartTime,
    endTime=datetime.now(),
    model="gpt-3.5-turbo",
    modelParameters={"maxTokens": "1000", "temperature": "0.9"},
    prompt=[{"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Please generate a summary of the following documents \nThe engineering department defined the following OKR goals...\nThe marketing department defined the following OKR goals..."}],
    metadata={"interface": "whatsapp"}
))

generation.id # <- id for observations

you can do the same with spans

span = trace.span(CreateSpan(
        name="embedding-search",
        startTime=retrievalStartTime,
        endTime=datetime.now(),
        metadata={"database": "pinecone"},
        input = {'query': 'This document entails the OKR goals for ACME'},
    )
)
span.id # <- id for observations

span = trace.span(CreateSpan(
        name="embedding-search",
        startTime=retrievalStartTime,
        endTime=datetime.now(),
        metadata={"database": "pinecone"},
        input = {'query': 'This document entails the OKR goals for ACME'},
    )
)
span.id # <- id for observations

Alex•13mo ago

Great, thanks! Just curious now, so the ids are generated directly by the sdk on the client side? No validation needed on the server? Or, does CreateGeneration communicate with the server under the hood?

Max•13mo ago

yes exactly 🙂 its non blocking, no validation server side

Alex•13mo ago

very cool, thanks!

Max•13mo ago

sure 🙂