ai_workflows / app /workflows /courses /suggest_expectations.py
theRealNG's picture
workflows(course_learn): Refactor tasks to concepts
b49d827
from langchain import hub, callbacks
from langchain_core.output_parsers import JsonOutputParser
from langchain_openai import ChatOpenAI
from pydantic import BaseModel, Field, UUID4
from typing import List
from pydantic import BaseModel
from typing import List, Optional
import os
class Expectation(BaseModel):
expectation: str = Field(
description="The learning outcome that the course designer has identified for the learner to demonstrate upon successful completion of the module.")
check_question: str = Field(
description="Targeted question that the course designer have developed to assess the learner's understanding of the learning outcomes.")
class Expectations(BaseModel):
expectations: List[Expectation]
class Response(BaseModel):
run_id: UUID4
expectations: List[Expectation]
class Inputs(BaseModel):
course: str
module: str
concepts: List[str]
existing_expectations: Optional[List[Expectation]]
class SuggestExpectations:
def kickoff(self, inputs={}):
self.course = inputs["course"]
self.module = inputs["module"]
self.existing_expectations = inputs["existing_expectations"]
self.concepts = inputs["concepts"]
llm_response = self._get_suggestions()
return {
"run_id": self.run_id,
"expectations": llm_response["expectations"]
}
def _get_suggestions(self):
parser = JsonOutputParser(pydantic_object=Expectations)
chain = self._build_chain()
# Existing Expectations
existing_expectations = []
for expectation in self.existing_expectations:
existing_expectations.append(f"""
Learning Outcome: {expectation.expectation}
Check Question: {expectation.check_question}
""")
existing_expectations_str = ""
if len(existing_expectations) > 0:
existing_expectations_str = "Here are existing Learning Outcomes & Check Questions for the module, don't repeat these learning outcomes:\n ```"
existing_expectations_str += "\n".join(existing_expectations)
existing_expectations_str += "\n```"
with callbacks.collect_runs() as cb:
llm_response = chain.invoke({
"course": self.course, "module": self.module, "concepts": "* " + ("\n* ".join(self.concepts)),
"format_instructions": parser.get_format_instructions(),
"existing_expectations": existing_expectations_str
})
self.run_id = cb.traced_runs[0].id
return llm_response
def _build_chain(self):
parser = JsonOutputParser(pydantic_object=Expectations)
prompt = hub.pull("course_learn_suggest_expectations")
llm = ChatOpenAI(model=os.environ['OPENAI_MODEL'], temperature=0.2)
chain = (prompt | llm | parser).with_config({
"tags": ["course_learn", "suggest_expectations"], "run_name": "Suggest Module Expectations",
"metadata": {
"version": "v1.0.0",
"growth_activity": "course_learn",
"env": os.environ["ENV"],
"model": os.environ["OPENAI_MODEL"],
}
})
return chain
# Example usage
# suggester = SuggestExpectations()
# response = suggester.kickoff(inputs={
# "course": "SQL",
# "module": "Query Optimization Techniques",
# "concepts": [
# "Watch this video https://youtu.be/BHwzDmr6d7s?si=sfFYnd73y9w0EjGB to understand SQL execution order and some optimization techniques.",
# "Watch this video https://youtu.be/FoznjTU929c?si=6M3xUIUwAxE6EbKS to understand SQL explain command usage.",
# "Go over these articles https://intellipaat.com/blog/sql-optimization-techniques/ & https://www.thoughtspot.com/data-trends/data-modeling/optimizing-sql-queries to understand various query optimization techniques."
# ]
# })
# print(response)