owusu/ds-fire-fighter
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge branch 'dev' of http://23.29.118.76:3000/owusu/ds-fire-fighter into dev

Browse Source
This commit is contained in:
OwusuBlessing
2025-02-11 19:21:30 +01:00
parent f430714a3d a760469ecd
commit 7200de4846
3 changed files with 384 additions and 131 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/prev.py
+304
View File
@@ -0,0 +1,304 @@
import os
import requests
from dotenv import load_dotenv
from langchain_openai import ChatOpenAI
from langchain_core.prompts.prompt import PromptTemplate
from langchain_core.output_parsers import StrOutputParser, JsonOutputParser
from langchain_openai import OpenAIEmbeddings
from langchain_core.documents import Document
from uuid import uuid4
import json
import getpass
import numpy as np
from concurrent.futures import ThreadPoolExecutor, as_completed
from sklearn.metrics.pairwise import cosine_similarity
from typing import List
import time
from datetime import datetime
import pytz
import logging
load_dotenv()
os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")
llm_temp = ChatOpenAI(model="gpt-4o-mini", temperature=0.7)
def generate_theme(conversation_data, resume, full_history, form_response=None, feedback=None, previous_result=None) -> dict:
try:
# Define prompt for summarizing and extracting the required fields
theme_prompt = PromptTemplate(
template="""
<|begin_of_text|><|start_header_id|>system<|end_header_id|>
You are a Fire Fighter Interview preparation assistant that generates STARTPOP FORMAT based on user interaction with AI.
Your responsibilities include carefully analyzing user interactions, themes, resumes,Onboarding questions and answers and work history to generate detailed STARTPOP formats for specific themes.
### Context and Guidelines:
1. **Purpose**: Generate a single behavioral question with a detailed STARTPOP format.
2. **Input Sources**:
- Current theme
- User interaction with AI
- User resume
- Full work history
- Onboarding questions and answers for additional context
3. **Output Format**: JSON object with the following fields:
- `theme_title`: Title of the theme provided.
- `question`: Behavioral question aligned with the theme.
- `Situation`: A bulleted list (75-100 words).
- `Task`: A bulleted list (50 words).
- `Action`: A bulleted list (2 negative actions and 2 positive actions).
- `Results and Transitions`: A bulleted list (25-50 words).
- `Personal Lessons`: A bulleted list (25-50 words).
- `Observations of Others`: A bulleted list (25 words).
- `Professional Connection`: A bulleted list (25-50 words). Additionally:
- Connect to the theme of the question.
- Creatively express why you should be part of their team.
### Key Concepts in Firefighting:
Throughout most Probationary Firefighter Interviews, evaluators assess alignment with the **7 Main Concepts of Firefighting**:
- **High Performance Teams**
- **Situational Awareness**
- **Being a Great Problem Solver**
- **Customer Service**
- **Building Construction, Mechanical Aptitude**
- **Emergency Medicine Experience**
- **Mental and Physical Health**
Additionally, they evaluate communication skills, competence, and likability.
### 20 Important Themes:
These themes are used for behavioral questions:
- Customer Service
- Conflict
- Challenge
- Leadership
- Stress
- Successful Team
- Diversity
- Mistake
- Unsuccessful Team
- Disagreement
- Bent a Rule
- Delivered a Difficult Message
- Displayed Integrity
- Took a Shortcut
- Didn’t Follow the Rules
- Emergency Response
- Dealt with Disabilities
- Solved a Big Problem
- Continuous Improvement
- Handled Sensitive Information
### Behavioral Question Starters:
Behavioral questions often begin with phrases like:
- "Tell me a time when..."
- "Can you tell me about a time when you..."
- "Describe a situation where you had to..."
- "Give me an example of how you..."
- "Have you ever been in a position where you needed to..."
- "Walk me through a time when you..."
### STARTPOP Framework:
The STARTPOP framework enhances the traditional STAR method. It includes:
1. **Situation**: Set up the scenario concisely (include dates, ages, places, and circumstances).
2. **Task**: Explain what needed to be done and why.
3. **Actions**: Outline both negative and positive approaches.
4. **Results and Transitions**: Share outcomes and ensure coherence.
5. **Personal Lessons**: Reflect on what you learned.
6. **Observations of Others**: Share insights about others involved.
7. **Professional Connection**: Relate the experience to firefighting and express your desire to join the team.
### Example STARTPOP:
**Question**: Tell me a time when you made a mistake and how you fixed it?
- **Situation**:
- In the Fall, my business, Tiger Building Services, does eavestrough cleaning.
- In 2019, we were working on a job late in the day, tired and running out of sunlight.
- I used handheld blowers without checking the wetness of debris, creating a muddy mess on the customer's deck.
- The customer was upset, and I realized my mistake.
- **Task**:
- Defuse the situation and clean up the mess quickly.
- Protect my company's reputation and ensure good customer experiences.
- **Actions**:
- Negative: Matching the customer's anger or ignoring the problem.
- Positive: Getting off the roof safely, apologizing, and switching strategies.
- Positive: Cleaning the gutters by hand and offering a free soft wash service.
- **Results and Transitions**:
- The job took longer than expected, but we waived fees due to the inconvenience.
- The customer was satisfied after our resolution plan.
- **Personal Lessons**:
- I learned to own up to mistakes, stay empathetic, and de-escalate tense situations.
- **Observations of Others**:
- People are entitled to their emotions, and following SOPs prevents mistakes.
- **Professional Connection**:
- Mistakes happen, but learning from them is crucial.
- I align with Markham Fire's values of transparency and accountability.
### JSON Output Requirements:
Generate a well-structured JSON output with the following fields:
- `theme_title`
- `question`
- `Situation`
- `Task`
- `Action`
- `Results and Transitions`
- `Personal Lessons`
- `Observations of Others`
- `Professional Connection`
### Review Process:
1. Ensure all news items align with the specified theme and meet relevance criteria.
2. Verify the JSON format is flawless, comprehensive, and well-structured.
### Additional Notes:
- You may be provided with feedback and previous results if the user is dissatisfied.
- Use this feedback to refine and regenerate the STARTPOP.
<|eot_id|><|start_header_id|>user<|end_header_id|>
Rules for Generating Each Component:
1. Situation: 75-100 words.
2. Task: 50 words.
3. Actions: 2 negative actions and 2 positive actions.
4. Results: 25-50 words.
5. Personal Lessons: 25-50 words.
6. Observations of Others: 25 words.
7. Professional Connection: 25-50 words + creative connection to the theme and team invitation.
NOTE: MAKE SURE THE OUT IS WELL DETAILED
CONVERSATION DATA: {conversation_data}
FEEDBACK: {feedback}
PREVIOUS RESULT: {previous_result}
USER RESUME: {resume}
FULL WORK HISTORY: {full_history}
Onboarding questions and answers for additional context: {form_response}
<|start_header_id|>assistant<|end_header_id|>
Return just the JSON output without any other explanation or comments.
Thank you for your thorough and precise processing!
""",
input_variables=["resume", "conversation_data", "feedback","form_response" "previous_result", "full_history"],
)
# Pipeline to process the prompt and parse output
theme_router = theme_prompt | llm_temp | JsonOutputParser()
# Call the pipeline and generate the cohesive output
output = theme_router.invoke({
"conversation_data": conversation_data,
"feedback": feedback,
"previous_result": previous_result,
"resume": resume,
"full_history": full_history,
"form_response":form_response
})
print(f"Output: {output}")
return output
except Exception as e:
print(f"Error: {e}")
return {}
from fastapi import Response, HTTPException, Depends
from typing import Optional
import os
import requests
import datetime
import base64 # For encoding the PDF content in Base64
@app.post("/rescue-career/generate-theme")
async def generate_pdf_endpoint(
request: GeneratePDFRequest,
api_key: str = Depends(get_api_key)
):
try:
# Fetch conversation data using the conversation_id
conversation_data = await get_conversation_data(request.conversation_id)
if not conversation_data:
raise HTTPException(
status_code=404,
detail=f"No conversation found with ID {request.conversation_id}"
)
resume_docs = ""
if request.resume_url:
docs = load_document(request.resume_url)
if not docs:
raise HTTPException(
status_code=400,
detail="Invalid resume URL: Unable to fetch document"
)
resume_docs = "\n".join(f"- {doc.page_content}" for doc in docs)
full_history_docs = ""
if request.full_history_url:
docs = load_document(request.full_history_url)
if not docs:
raise HTTPException(
status_code=400,
detail="Invalid full_history URL: Unable to fetch document"
)
full_history_docs = "\n".join(f"- {doc.page_content}" for doc in docs)
form_response_docs = ""
if request.form_id:
try:
x_api_key = os.getenv("BACKEND_XAPI_KEY")
url = f"{os.getenv('BACKEND_BASE_URL')}/v3/api/custom/theme-document/answer/{request.form_id}?x-project={x_api_key}"
result = requests.get(url)
form_response = result.json() # Return response in JSON format
form_response_docs = "\n".join(f"- {form_response}")
except:
raise HTTPException(
status_code=400,
detail="Unable to fetch onboarding data"
)
# Generate theme data using the generate_theme function
theme_data = generate_theme(
conversation_data=conversation_data,
feedback=request.feedback,
previous_result=request.previous_results,
resume=resume_docs,
form_response=form_response_docs,
full_history=full_history_docs
)
if not theme_data:
raise HTTPException(
status_code=500,
detail="Failed to generate theme data"
)
# Generate the PDF using the create_pdf function
pdf_content = create_pdf(theme_data)
# Encode the PDF content in Base64
pdf_base64 = base64.b64encode(pdf_content).decode("utf-8")
# Create filename with timestamp
timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
filename = f"theme_{timestamp}.pdf"
# Return both the PDF (as Base64) and the theme data in a JSON response
return {
"theme_data": theme_data,
"pdf": {
"filename": filename,
"content": pdf_base64
}
}
except Exception as e:
raise HTTPException(
status_code=500,
detail=f"Error generating PDF: {str(e)}"
)
scripts/generate_quiz.py
+76 -127
View File
@@ -21,155 +21,104 @@ load_dotenv()
os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")
llm_temp = ChatOpenAI(model="gpt-4o-mini", temperature=0.7)
llm_temp = ChatOpenAI(model="gpt-4o-mini", temperature=0.7,max_tokens=10000)
def generate_quiz(startpop_pdf, quiz_type=None) -> dict:
try:
# Define prompt for summarizing and extracting the required fields
# Define the prompt template for generating the quiz
quiz_prompt = PromptTemplate(
template="""<|begin_of_text|><|start_header_id|>system<|end_header_id|>
You are a Fire Fighter Interview preparation assistant that generates QUIZ for user based on STARTPOP FORMAT PDF BASED on
template="""
<|begin_of_text|><|start_header_id|>system<|end_header_id|>
You are an assistant designed to generate firefighter interview quizzes based on the STARTPOP framework provided in a PDF document. Your task is to analyze the content of the PDF and create a quiz tailored to the specified quiz type.
IN THE STARTPOP FORMAT PDF, each theme has its own questions with corresponding STARTPOP framework for each question.
### Quiz Types:
1. **Single Line Text Inputs**:
- Output format: `{"question": "Your question", "correct_answer": "Your correct answer"}`
Your responsibility is to carefully analyze the provided PDF data and then generate a quiz for the user.
You will also be provided with the type of quiz.
2. **Multiple Choice Questions**:
- Output format: `{"question": "Your question", "options": ["Option 1", "Option 2"], "correct_answer": "Correct Option"}`
There are three different types of quizzes namely:
3. **True or False Questions**:
- Output format: `{"question": "Your question", "options": ["True", "False"], "correct_answer": "True or False"}`
1- Single line text inputs
2- Multiple Choice questions
3- True or False questions
Each quiz must include a field called `"quiz_type"` with values `1`, `2`, or `3` corresponding to the quiz type.
For each quiz type, return the following JSON format:
### Project Overview:
Firefighter interviews evaluate candidates based on **7 Main Concepts** and **20 Important Themes**. These are used to assess alignment with firefighting principles, communication skills, problem-solving abilities, and overall competence.
1. For Single Line Text Inputs:
- A list of objects, each with {{"question": "Your question", "correct_answer": "Your correct answer"}}
#### 7 Main Concepts:
- High Performance Teams
- Situational Awareness
- Being a Great Problem Solver
- Customer Service
- Building Construction & Mechanical Aptitude
- Emergency Medicine Experience
- Mental & Physical Health
2. For Multiple Choice Questions:
- A list of objects, each with {{"question": "Your question", "options": ["Option 1", "Option 2"], "correct_answer": "Correct Option"}}
#### 20 Important Themes:
- Customer Service
- Conflict
- Challenge
- Leadership
- Stress
- Successful Team
- Diversity
- Mistake
- Unsuccessful Team
- Disagreement
- Bent a Rule
- Delivered a Difficult Message
- Displayed Integrity
- Took a Shortcut
- Didn’t Follow the Rules
- Emergency Response
- Dealt with Disabilities
- Solved a Big Problem
- Continuous Improvement
- Handled Sensitive Information
3. For True or False Questions:
- A list of objects, each with {{"question": "Your question", "options": ["True", "False"], "correct_answer": "True or False"}}
#### Behavioral Question Starters:
Questions often begin with phrases like:
- "Tell me a time when..."
- "Can you tell me about a time when you..."
- "Describe a situation where you had to..."
- "Give me an example of how you..."
- "Have you ever been in a position where you needed to..."
- "Walk me through a time when you..."
Each response should also include a field called "quiz_type" which can be either 1, 2, or 3 respectively.
#### STARTPOP Framework:
The STARTPOP framework enhances the STAR method by adding depth and variety to responses. Its components are:
1. **Situation**: Set up the context (dates, places, circumstances).
2. **Task**: Explain what needed to be done and why.
3. **Actions**: Outline both positive and negative approaches.
4. **Results**: Share outcomes in a time-specific manner.
5. **Transitions**: Ensure professional coherence.
6. **Personal Lessons**: Discuss what you learned.
7. **Other People Observations**: Share insights about others involved.
8. **Professional Connection**: Relate the experience to firefighting.
Return just the JSON output without any other explanation or comments.
### Instructions:
- Analyze the provided STARTPOP PDF to extract relevant themes and concepts.
- Generate a quiz that builds user confidence by focusing on interview-based scenarios.
- Avoid questions directly about the STARTPOP framework itself (e.g., "What is STARTPOP?").
- Use the specified quiz type (`quiz_type`) to determine the output format.
TO KNOW MORE ABOUT THE PROJECT READ BELOW
----START------
Throughout most Probationary Firefighter Interviews, they will be evaluating a ton of things. Typically, they want to see how you align with the **7 Main Concepts of Firefighting**. They are also watching how nervous you are, your communication skills, and your overall general competence for the role. At the end of the day, you want them to like you.
STARTPOP FULL PDF: {startpop_pdf}
QUIZ TYPE: {quiz_type}
### 7 Main Concepts:
- **High Performance Teams**
- **Situational Awareness**
- **Being a Great Problem Solver**
- **Customer Service**
- **Building Construction, Mechanical Aptitude**
- **Emergency Medicine Experience**
- **Mental and Physical Health**
Your crew of four firefighters is usually comprised of a Driver, a Captain, and two firefighters in the back. That is a High-Performance Team.
We are frequently dispatched to calls that require using our understanding of Building Construction Concepts, Mechanical Aptitude, and Emergency Medical Experience. When you respond to an emergency event that is inherently dangerous (like a vehicle fire, a car accident in a slanted ditch, a person trapped under a machine, a house fire, or a chemical suicide), you need to use your Situational Awareness to keep that crew safe.
Sometimes the tools, training, and tactics that you have been taught work perfectly. Sometimes they don’t. Can you be a Good Problem Solver to quickly come up with something to make the situation better for the people, places, and environments that we protect?
Ultimately, your crew will be serving the public, and the chiefs need to know that you can be trained to be above their desired standard so that you give the public great Customer Service.
### 20 Important Themes
Consider the 7 concepts to be the soil. All of your stories grow out of that soil. But not every story works for every question. You need to handpick the right one at the right times to give them. Sort of like how you handpick flowers out of the soil. You NEED to have **20 different flowers** so that you are fully prepared for whatever behavioral question they throw at you. These are the **20 Themes** that you would use for behavioral questions:
- Customer Service
- Conflict
- Challenge
- Leadership
- Stress
- Successful Team
- Diversity
- Mistake
- Unsuccessful Team
- Disagreement
- Bent a Rule
- Delivered a Difficult Message
- Displayed Integrity
- Took a Shortcut
- Didn’t Follow the Rules
- Emergency Response
- Dealt with Disabilities
- Solved a Big Problem
- Continuous Improvement
- Handled Sensitive Information
### Behavioral Question Starters
Behavioral questions usually start with phrases like:
- “Tell me a time when…”
- “Can you tell me about a time when you…”
- "Describe a situation where you had to…"
- "Give me an example of how you…"
- "Have you ever been in a position where you needed to…"
- "Walk me through a time when you…"
### STARTPOP Framework
The STAR Format is what most people tell you to do in order to answer a firefighter interview question. It’s a great framework. I highly recommend it. I just advise that you pump it up even further. I call it **STARTPOP**.
Try and pull from different parts of your life. My Chief Training Officer told me that he enjoys candidates that are able to use different experiences to answer the questions. Listening to someone drone on and on about a singular time or type of event in their life is a massive turn-off to the interview panel. That’s a bad thing. Just like most things, variety is the spice of life.
#### Components of STARTPOP:
1. **Situation**:
- Set up the answer in the mind of the question asker.
- Your storytelling skills matter here. It has to be concise and impactful (no more than 25 seconds long).
- Include dates, ages, places, and circumstances.
2. **Task**:
- Explain what you needed to do and why you needed to do it.
- Recap the situation quickly from a different angle.
3. **Actions**:
- Outline both the negative and the positive way of doing things.
- Show high moral character in every question.
4. **Results**:
- Explain what happened as a result of your actions.
- Share results in a time-specific manner (e.g., “5 months later X happened”).
5. **Transitions**:
- Speak in a way that aligns with professional expectations.
- Ensure coherence in your responses.
6. **Personal Lessons**:
- Discuss what you learned about yourself.
- Address any concerns the interviewers might have about hiring you.
7. **Other People Observations**:
- Share insights about others in the situation.
- Keep it short and to the point.
8. **Professional Connection**:
- Relate your experience directly to the fire service.
- Conclude strongly, avoiding phrases like “and so yeah…”.
----END------
NOTE: THE QUIZ FOCUES ON BULIDNG USER CONFIDENCE BY ANANLYZING THE QUESTIONS AND FRAMEWORK FOR EACH QUESTION IN THE STARTPOP FRAMEWORK PDF,SOLELY USE THIS PDF PROVIDED BY THE USER
BASED ON THIS FRAMEWORK , CREATE INTERVIEW BASED QUIZ FOR FIRE FIGHTING ROLE BY ANALYZING THIS DOCUMENT
NOTE : THE QUIZ SHOULD NOT BE BASED ON STARTPOP FRAMEWORK ITSELF BUT ANALYZE THE STARTPOP FRAMEWORK PRESENTED TO GENERATE INTERVIEW BASED QUIZ
e.g "The STARTPOP framework is specifically designed for firefighter interviews", THIS KIND OF QUESTION SHOULD NOT BE ASKED IN THE QUIZ....
Thank you for your thorough and precise processing!
STARTPOP FULL PDF :{startpop_pdf}
question type : {quiz_type}
P
<|eot_id|><|start_header_id|>user<|end_header_id|>""",
input_variables=["startpop_pdf", "quiz_type", "question"],
<|eot_id|><|start_header_id|>user<|end_header_id|>
""",
input_variables=["startpop_pdf", "quiz_type"],
)
# Pipeline to process the prompt and parse output
# Pipeline to process the prompt and parse the output
quiz_router = quiz_prompt | llm_temp | JsonOutputParser()
# Call the pipeline and generate the cohesive output
output = quiz_router.invoke({"startpop_pdf": startpop_pdf, "quiz_type": quiz_type, "question": "Your question here"})
output = quiz_router.invoke({"startpop_pdf": startpop_pdf, "quiz_type": quiz_type})
return output
except Exception as e:
print(f"Error:{e}")
print(f"Error: {e}")
return {}
scripts/generate_theme.py
+4 -4
View File
@@ -21,7 +21,7 @@ load_dotenv()
os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")
llm_temp = ChatOpenAI(model="gpt-4o-mini", temperature=0.7)
llm_temp = ChatOpenAI(model="gpt-4o-mini", temperature=0.7,max_tokens=5000)
def generate_theme(conversation_data, resume, full_history, form_response=None, feedback=None, previous_result=None) -> dict:
try:
# Define prompt for summarizing and extracting the required fields
@@ -159,9 +159,9 @@ Generate a well-structured JSON output with the following fields:
- Use this feedback to refine and regenerate the STARTPOP.
<|eot_id|><|start_header_id|>user<|end_header_id|>
Rules for Generating Each Component:
1. Situation: 100 - 120 words.
2. Task: 100 words.
Rules for Generating Each Component:FOLLOW THIS STRICTLY
1. Situation: 80 - 120 words.
2. Task: 70 words.
3. Actions: 2 negative actions and 2 positive actions.
4. Results: 50-70 words.
5. Personal Lessons: 50-70 words.