├── stroutputparser1.py
├── jsonoutputparser.py
├── stroutputparser.py
├── structuredoutputparser.py
└── pydanticoutputparser.py


/stroutputparser1.py:
--------------------------------------------------------------------------------
 1 | from langchain_openai import ChatOpenAI
 2 | from dotenv import load_dotenv
 3 | from langchain_core.prompts import PromptTemplate
 4 | from langchain_core.output_parsers import StrOutputParser
 5 | 
 6 | load_dotenv()
 7 | 
 8 | 
 9 | model = ChatOpenAI()
10 | 
11 | # 1st prompt -> detailed report
12 | template1 = PromptTemplate(
13 |     template='Write a detailed report on {topic}',
14 |     input_variables=['topic']
15 | )
16 | 
17 | # 2nd prompt -> summary
18 | template2 = PromptTemplate(
19 |     template='Write a 5 line summary on the following text. /n {text}',
20 |     input_variables=['text']
21 | )
22 | 
23 | parser = StrOutputParser()
24 | 
25 | chain = template1 | model | parser | template2 | model | parser
26 | 
27 | result = chain.invoke({'topic':'black hole'})
28 | 
29 | print(result)
30 | 
31 | 


--------------------------------------------------------------------------------
/jsonoutputparser.py:
--------------------------------------------------------------------------------
 1 | from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
 2 | from dotenv import load_dotenv
 3 | from langchain_core.prompts import PromptTemplate
 4 | from langchain_core.output_parsers import JsonOutputParser
 5 | 
 6 | load_dotenv()
 7 | 
 8 | # Define the model
 9 | llm = HuggingFaceEndpoint(
10 |     repo_id="google/gemma-2-2b-it",
11 |     task="text-generation"
12 | )
13 | 
14 | model = ChatHuggingFace(llm=llm)
15 | 
16 | parser = JsonOutputParser()
17 | 
18 | template = PromptTemplate(
19 |     template='Give me 5 facts about {topic} \n {format_instruction}',
20 |     input_variables=['topic'],
21 |     partial_variables={'format_instruction': parser.get_format_instructions()}
22 | )
23 | 
24 | chain = template | model | parser
25 | 
26 | result = chain.invoke({'topic':'black hole'})
27 | 
28 | print(result)
29 | 
30 | 


--------------------------------------------------------------------------------
/stroutputparser.py:
--------------------------------------------------------------------------------
 1 | from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
 2 | from dotenv import load_dotenv
 3 | from langchain_core.prompts import PromptTemplate
 4 | 
 5 | load_dotenv()
 6 | 
 7 | llm = HuggingFaceEndpoint(
 8 |     repo_id="google/gemma-2-2b-it",
 9 |     task="text-generation"
10 | )
11 | 
12 | model = ChatHuggingFace(llm=llm)
13 | 
14 | # 1st prompt -> detailed report
15 | template1 = PromptTemplate(
16 |     template='Write a detailed report on {topic}',
17 |     input_variables=['topic']
18 | )
19 | 
20 | # 2nd prompt -> summary
21 | template2 = PromptTemplate(
22 |     template='Write a 5 line summary on the following text. /n {text}',
23 |     input_variables=['text']
24 | )
25 | 
26 | prompt1 = template1.invoke({'topic':'black hole'})
27 | 
28 | result = model.invoke(prompt1)
29 | 
30 | prompt2 = template2.invoke({'text':result.content})
31 | 
32 | result1 = model.invoke(prompt2)
33 | 
34 | print(result1.content)
35 | 
36 | 


--------------------------------------------------------------------------------
/structuredoutputparser.py:
--------------------------------------------------------------------------------
 1 | from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
 2 | from dotenv import load_dotenv
 3 | from langchain_core.prompts import PromptTemplate
 4 | from langchain.output_parsers import StructuredOutputParser, ResponseSchema
 5 | 
 6 | load_dotenv()
 7 | 
 8 | # Define the model
 9 | llm = HuggingFaceEndpoint(
10 |     repo_id="google/gemma-2-2b-it",
11 |     task="text-generation"
12 | )
13 | 
14 | model = ChatHuggingFace(llm=llm)
15 | 
16 | schema = [
17 |     ResponseSchema(name='fact_1', description='Fact 1 about the topic'),
18 |     ResponseSchema(name='fact_2', description='Fact 2 about the topic'),
19 |     ResponseSchema(name='fact_3', description='Fact 3 about the topic'),
20 | ]
21 | 
22 | parser = StructuredOutputParser.from_response_schemas(schema)
23 | 
24 | template = PromptTemplate(
25 |     template='Give 3 fact about {topic} \n {format_instruction}',
26 |     input_variables=['topic'],
27 |     partial_variables={'format_instruction':parser.get_format_instructions()}
28 | )
29 | 
30 | chain = template | model | parser
31 | 
32 | result = chain.invoke({'topic':'black hole'})
33 | 
34 | print(result)


--------------------------------------------------------------------------------
/pydanticoutputparser.py:
--------------------------------------------------------------------------------
 1 | from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
 2 | from dotenv import load_dotenv
 3 | from langchain_core.prompts import PromptTemplate
 4 | from langchain_core.output_parsers import PydanticOutputParser
 5 | from pydantic import BaseModel, Field
 6 | 
 7 | load_dotenv()
 8 | 
 9 | # Define the model
10 | llm = HuggingFaceEndpoint(
11 |     repo_id="google/gemma-2-2b-it",
12 |     task="text-generation"
13 | )
14 | 
15 | model = ChatHuggingFace(llm=llm)
16 | 
17 | class Person(BaseModel):
18 | 
19 |     name: str = Field(description='Name of the person')
20 |     age: int = Field(gt=18, description='Age of the person')
21 |     city: str = Field(description='Name of the city the person belongs to')
22 | 
23 | parser = PydanticOutputParser(pydantic_object=Person)
24 | 
25 | template = PromptTemplate(
26 |     template='Generate the name, age and city of a fictional {place} person \n {format_instruction}',
27 |     input_variables=['place'],
28 |     partial_variables={'format_instruction':parser.get_format_instructions()}
29 | )
30 | 
31 | chain = template | model | parser
32 | 
33 | final_result = chain.invoke({'place':'sri lankan'})
34 | 
35 | print(final_result)
36 | 


--------------------------------------------------------------------------------