Getting error

[Rudra9214]

from typing import List, Dict
from dotenv import load_dotenv
from langchain_openai import ChatOpenAI
from pydantic import BaseModel, Field
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.messages import AIMessage, HumanMessage, ToolMessage
import json # Import JSON for formatting

load_dotenv()
llm = ChatOpenAI(model="gpt-4o")

Pydantic for structured output

class ResponseItem(BaseModel):
TYPE: str = Field(description="Type of the response")
PAYLOAD: Dict[str, str] = Field(description="Payload containing question and answer")

Pydantic for the full response structure

class ResponseSchema(BaseModel):
responses: List[ResponseItem] = Field(description="List of responses")

System message for chatbot

system_prompt = """You are a highly skilled JSON creator specializing in converting user input into structured JSON formats.
Your task is to process lines of text representing test cases and return their structured JSON equivalents based on a predefined schema.

The output JSON should be formatted as follows:
[
{
"TYPE": "QNA",
"PAYLOAD": {
"QUESTION": "<user's question>",
"ANSWER": ""
}
},
{
"TYPE": "HUMAN_HANDOFF",
"PAYLOAD": {}
}
]

Here are some examples of how to structure the output:

example_user: Convert the following line to JSON: "What is the capital of France?"
example_assistant: [
{
"TYPE": "QNA",
"PAYLOAD": {
"QUESTION": "What is the capital of France?",
"ANSWER": "Paris"
}
}
]

example_user: Convert the following line to JSON: "Say hello in Spanish"
example_assistant: [
{
"TYPE": "GREETING",
"PAYLOAD": {
"QUESTION": "Say hello in Spanish",
"ANSWER": "Hola"
}
}
]

example_user: Convert the following line to JSON: "How to contact customer support?"
example_assistant: [
{
"TYPE": "HUMAN_HANDOFF",
"PAYLOAD": {
"QUESTION": "How to contact customer support?",
"ANSWER": "You can reach customer support at [email protected] or call 123-456-7890."
}
}
]

example_user: Convert the following line to JSON: "What are the features of the iPhone 14?"
example_assistant: [
{
"TYPE": "PRODUCT_INFO",
"PAYLOAD": {
"QUESTION": "What are the features of the iPhone 14?",
"ANSWER": "The iPhone 14 features a 6.1-inch display, A15 Bionic chip, and improved battery life."
}
}
]"""

prompt = ChatPromptTemplate.from_messages([("system", system_prompt), ("human", "{input}")])

LLM with structured output

structured_llm = llm.with_structured_output(ResponseSchema)

Combine prompt and LLM

few_shot_structured_llm = prompt | structured_llm

Invoke the model with a user question

Format the structured JSON response

print(few_shot_structured_llm.invoke("connect me to an agent"))

Getting below error: raise TypeError(
TypeError: Expected mapping type as input to ChatPromptTemplate. Received <class 'str'>.

[dosubot[bot]]

Hey @Rudra9214! How can I assist you today?

The error you're encountering, TypeError: Expected mapping type as input to ChatPromptTemplate. Received <class 'str'>, suggests that the input to the ChatPromptTemplate is not in the expected format. The ChatPromptTemplate expects a dictionary-like input, but it seems you're providing a string directly.

To resolve this, ensure that when you invoke the model, you pass a dictionary with the key that matches the placeholder in your prompt. In your case, the placeholder is {input}, so you should pass a dictionary with "input" as the key. Here's how you can modify your code:

# Invoke the model with a user question
response = few_shot_structured_llm.invoke({"input": "connect me to an agent"})

# Format the structured JSON response
print(response)

This change should align the input with the expected format for the ChatPromptTemplate, resolving the TypeError.

To continue talking to Dosu, mention @dosu.

[Rudra9214]

@dosu

raise KeyError(
KeyError: 'Input to ChatPromptTemplate is missing variables {'\n "TYPE"'}. Expected: ['\n "TYPE"', 'input'] Received: ['input']\nNote: if you intended {\n "TYPE"} to be part of the string and not a variable, please escape it with double curly braces like: '{{\n "TYPE"}}'.\nFor troubleshooting, visit: https://python.langchain.com/docs/troubleshooting/errors/INVALID_PROMPT_INPUT'

[YarickVodila]

@Rudra9214 @dosu
I ran into the same problem when parsing data. You need to escape each dictionary in the list with curly brackets { }

Example:

from langchain_core.prompts import ChatPromptTemplate

example_user = """
Контакты редакции/Contacts
Адрес редакции: 191023, Санкт-Петербург, Невский пр., 60
Контактные телефоны: (812) 571-65-55; (812) 571-12-19 (факс); (812) 494-05-28
E-mail редакции: [email protected]. 
Контактное лицо – Сорока Ольга Владимировна
время работы – пн, пт, с 10:00 до 18:00
Журнал включен в Российский индекс научного цитирования
Address: 191023, Saint-Petersburg, Nevsky PR., 60 
Contact phone numbers: (812) 571-65-55; (812) 571-12-19 (Fax); (812) 494-05-28 
E-mail: [email protected]. 
Contact person Soroka Olga Vladimirovna 
The journal is included in Russian science citation index 
Подписывайтесь на нас:
Контакты
 191023, г. Санкт-Петербург
 Невский пр., д. 60
8 800 100-06-44
 (бесплатно по России)
+7 812 570-55-76
 (приемная комиссия)
+7 812 571-65-55
 (приемная ректора)
+7 812 570-55-50
 Экономический лицей
Отписаться от почтовой рассылки.
191023, г. Санкт-Петербург, Невский пр., д. 60
8 800 100-06-44 (бесплатно по России)
+7 812 570-55-76(приемная комиссия)
+7 812 571-65-55(приемная ректора)
+7 812 570-55-50(Экономический лицей)
"""
example_user = re.sub(r'[\n\t ]+', ' ', example_user)


example_assistant = """
[
    {{
        "name": "Ольга",
        "patronymic": "Владимировна",
        "last_name": "Сорока",
        "position": "",
        "emails": ["[email protected]"],
        "phone_numbers": ["(812) 571-65-55", "(812) 571-12-19", "(812) 494-05-28"],
        "department": ""
    }},
    {{
        "name": "",
        "patronymic": "",
        "last_name": "",
        "position": "",
        "emails": [],
        "phone_numbers": ["+7 812 570-55-76"],
        "department": "приемная комиссия"
    }},
    {{
        "name": "",
        "patronymic": "",
        "last_name": "",
        "position": "",
        "emails": [],
        "phone_numbers": ["+7 812 571-65-55"],
        "department": "приемная ректора"
    }},
    {{
        "name": "",
        "patronymic": "",
        "last_name": "",
        "position": "",
        "emails": [],
        "phone_numbers": ["+7 812 570-55-50"],
        "department": "Экономический лицей"
    }},
    {{
        "name": "",
        "patronymic": "",
        "last_name": "",
        "position": "",
        "emails": [],
        "phone_numbers": ["8 800 100-06-44"],
        "department": "general"
    }}
]
"""
example_assistant = re.sub(r'[\n\t ]+', ' ', example_assistant)


# json_schema = """
# [
#         {
#             "name": "",
#             "patronymic": "",
#             "last name": "",
#             "position": "",
#             "emails": [],
#             "phone numbers": [],
#             "department": ""
#         },
#         {
#             "name": "",
#             "patronymic": "",
#             "last_name": "",
#             "position": "",
#             "emails": [],
#             "phone_numbers": [],
#             "department": ""
#         }
# ]
# """.replace('"', "'")


json_schema = re.sub(r'[\n\t ]+', ' ', json_schema)


system = f"""Извлеките из текста имена, фамилии, отчества, должности, отделы, номера телефонов и электронные письма. Очень важно, чтобы ключи были в словаре

Пример данных:

example_user: {example_user}
example_assistant: {example_assistant}"""

# system = """Извлеките из текста имена, фамилии, отчества, должности, отделы, номера телефонов и электронные письма. Очень важно, чтобы ключи были в словаре

# Пример данных:

# example_user: Контакты редакции/Contacts Адрес редакции: 191023, Санкт-Петербург, Невский пр., 60 Контактные телефоны: (812) 571-65-55; (812) 571-12-19 (факс); (812) 494-05-28 E-mail редакции: [email protected]. Контактное лицо – Сорока Ольга Владимировна время работы – пн, пт, с 10:00 до 18:00 Журнал включен в Российский индекс научного цитирования Address: 191023, Saint-Petersburg, Nevsky PR., 60 Contact phone numbers: (812) 571-65-55; (812) 571-12-19 (Fax);\xa0(812) 494-05-28\xa0 E-mail: [email protected]. Contact person Soroka Olga Vladimirovna The journal is included in Russian science citation index Подписывайтесь на нас: Контакты 191023, г. Санкт-Петербург Невский пр., д. 60 8 800 100-06-44 (бесплатно по России) +7 812 570-55-76 (приемная комиссия) +7 812 571-65-55 (приемная ректора) +7 812 570-55-50 Экономический лицей Отписаться от почтовой рассылки. 191023, г. Санкт-Петербург, Невский пр., д. 60 8 800 100-06-44 (бесплатно по России) +7 812 570-55-76(приемная комиссия) +7 812 571-65-55(приемная ректора) +7 812 570-55-50(Экономический лицей)
# example_assistant: [{{"name": "Ольга", "patronymic": "Владимировна", "last_name": "Сорока", "position": "", "emails": ["[email protected]"], "phone_numbers": ["(812) 571-65-55", "(812) 571-12-19", "(812) 494-05-28"], "department": "" }}, {{ "name": "", "patronymic": "", "last_name": "", "position": "", "emails": [], "phone_numbers": ["+7 812 570-55-76"], "department": "приемная комиссия" }}, {{ "name": "", "patronymic": "", "last_name": "", "position": "", "emails": [], "phone_numbers": ["+7 812 571-65-55"], "department": "приемная ректора" }}, {{ "name": "", "patronymic": "", "last_name": "", "position": "", "emails": [], "phone_numbers": ["+7 812 570-55-50"], "department": "Экономический лицей" }}, {{ "name": "", "patronymic": "", "last_name": "", "position": "", "emails": [], "phone_numbers": ["8 800 100-06-44"], "department": "general"}}]"""

prompt = ChatPromptTemplate([("system", system), ("human", "{text}")])
prompt

[dosubot[bot]]

Could you provide more details about how you're constructing and using the ChatPromptTemplate? Specifically, it would be helpful to see the part of your code where you define the prompt and how you're passing variables to it. This will help in understanding why the TYPE variable is expected and how to address the KeyError.

To continue talking to Dosu, mention @dosu.