Fiddler

Fiddler 是企業生成式和預測系統運營的先驅，提供統一平台，使資料科學、MLOps、風險、合規、分析和其他業務線團隊能夠以企業規模監控、解釋、分析和改進 ML 部署。

1. 安裝與設定

#!pip install langchain langchain-community langchain-openai fiddler-client

2. Fiddler 連線詳細資訊

在您可以使用 Fiddler 新增有關模型的資訊之前

1. 您用來連線到 Fiddler 的 URL
2. 您的組織 ID
3. 您的授權權杖

這些資訊可以在您的 Fiddler 環境的「設定」頁面中找到。

URL = ""  # Your Fiddler instance URL, Make sure to include the full URL (including https://). For example: https://demo.fiddler.ai
ORG_NAME = ""
AUTH_TOKEN = ""  # Your Fiddler instance auth token

# Fiddler project and model names, used for model registration
PROJECT_NAME = ""
MODEL_NAME = ""  # Model name in Fiddler

3. 建立 Fiddler 回呼處理常式執行個體

from langchain_community.callbacks.fiddler_callback import FiddlerCallbackHandler

fiddler_handler = FiddlerCallbackHandler(
    url=URL,
    org=ORG_NAME,
    project=PROJECT_NAME,
    model=MODEL_NAME,
    api_key=AUTH_TOKEN,
)

API 參考：FiddlerCallbackHandler

範例 1：基本鏈

from langchain_core.output_parsers import StrOutputParser
from langchain_openai import OpenAI

# Note : Make sure openai API key is set in the environment variable OPENAI_API_KEY
llm = OpenAI(temperature=0, streaming=True, callbacks=[fiddler_handler])
output_parser = StrOutputParser()

chain = llm | output_parser

# Invoke the chain. Invocation will be logged to Fiddler, and metrics automatically generated
chain.invoke("How far is moon from earth?")

API 參考：StrOutputParser | OpenAI

# Few more invocations
chain.invoke("What is the temperature on Mars?")
chain.invoke("How much is 2 + 200000?")
chain.invoke("Which movie won the oscars this year?")
chain.invoke("Can you write me a poem about insomnia?")
chain.invoke("How are you doing today?")
chain.invoke("What is the meaning of life?")

範例 2：具有提示範本的鏈

from langchain_core.prompts import (
    ChatPromptTemplate,
    FewShotChatMessagePromptTemplate,
)

examples = [
    {"input": "2+2", "output": "4"},
    {"input": "2+3", "output": "5"},
]

example_prompt = ChatPromptTemplate.from_messages(
    [
        ("human", "{input}"),
        ("ai", "{output}"),
    ]
)

few_shot_prompt = FewShotChatMessagePromptTemplate(
    example_prompt=example_prompt,
    examples=examples,
)

final_prompt = ChatPromptTemplate.from_messages(
    [
        ("system", "You are a wondrous wizard of math."),
        few_shot_prompt,
        ("human", "{input}"),
    ]
)

# Note : Make sure openai API key is set in the environment variable OPENAI_API_KEY
llm = OpenAI(temperature=0, streaming=True, callbacks=[fiddler_handler])

chain = final_prompt | llm

# Invoke the chain. Invocation will be logged to Fiddler, and metrics automatically generated
chain.invoke({"input": "What's the square of a triangle?"})

API 參考：ChatPromptTemplate | FewShotChatMessagePromptTemplate