inital code

Makefile

@@ -0,0 +1,2 @@
+deps:
+	pipenv requirements > requirements.txt

app.py

@@ -0,0 +1,101 @@
+import json
+import os
+import gradio as gr
+import requests
+import recruiting_assistant
+
+
+def search_resume(input_text):
+    url = f"https://n970resrb9.execute-api.eu-west-1.amazonaws.com/dev/prediction"  # replace with your API endpoint
+    headers = {
+        "Content-Type": "application/json",
+        "x-api-key": os.environ["API_KEY"],
+    }  # adjust headers as needed
+    response = requests.post(
+        url, headers=headers, data=json.dumps({"text": input_text})
+    )
+    response_data = response.json()
+
+    if "prediction" in response_data:
+        prediction = response_data["prediction"]
+        if isinstance(prediction, list):
+            # Insert a newline after each '.'
+            # Insert a newline after each '.' and add "Candidate <follow up number>:\n" before each item
+            updated_prediction = [
+                f"Candidate {i + 1}:\n=============================\n{s}"
+                for i, s in enumerate(prediction)
+            ]
+            updated_prediction = [s.replace(". ", ".\n") for s in updated_prediction]
+            updated_prediction = [s.replace("•", "\n - ") for s in updated_prediction]
+            return "\n\n".join(updated_prediction)
+    return "No 'prediction' key found in the response or the 'body' is not a list."
+
+demo = gr.Blocks()
+
+with demo:
+    with gr.Group():
+        with gr.Box():
+            with gr.Row(elem_id="prompt-container").style(
+                    mobile_collapse=False, equal_height=True
+            ):
+                with gr.Column():
+                    gr.Markdown(
+                        """
+
+                    ## 1. Provide a vacancy and get back relevant resumes from an entire database of resumes for various roles.
+                    """
+                    )
+                    text_vacancy = gr.Textbox(
+                        hint="Paste here a Vacancy...",
+                        lines=7,
+                        label="Copy/paste here a vacancy",
+                    )
+                    b1 = gr.Button("Search Resume").style(
+                        margin=False,
+                        rounded=(False, True, True, False),
+                        full_width=False,
+                    )
+                    text_search_result = gr.Textbox(
+                        hint="Top resumes will appear here ...",
+                        label="Top resumes found in the database",
+                    )
+                    b1.click(
+                        search_resume, inputs=text_vacancy, outputs=text_search_result
+                    )
+                    gr.Markdown(
+                        """
+
+                    ## 2. Select an appropriate resume for this vacancy, paste it in the textfield and get a relevant introduction email.
+                    """
+                    )
+                    text_resume = gr.Textbox(
+                        hint="Paste here a Resume...",
+                        label="Copy / Paste here your prefered resume from above and click the button to write an intro ",
+                    )
+                    b2 = gr.Button("Write a relevant intro").style(
+                        margin=False,
+                        rounded=(False, True, True, False),
+                        full_width=False,
+                    )
+                    gr.Markdown(
+                        """
+
+                    ## 3. You have a relevant introduction email to send to the customer.
+                    """
+                    )
+                    text_intro = gr.Textbox(label="Intro Email")
+                    evaluation = gr.Textbox(label="Evaluation of the skills")
+                    b2.click(
+                        recruiting_assistant.create_intro,
+                        inputs=[text_vacancy, text_resume],
+                        outputs=[text_intro, evaluation],
+                    )
+
+                    gr.Examples(
+                        fn=search_resume,
+                        inputs=text_vacancy,
+                        outputs=text_search_result,
+                        cache_examples=False,
+                    )
+
+demo.launch()

recruiting_assistant.py

@@ -0,0 +1,170 @@
+import json
+import os
+from langchain.chat_models import ChatOpenAI
+from langchain.prompts import ChatPromptTemplate
+from langchain.chains import LLMChain, SequentialChain
+
+llm = ChatOpenAI(temperature=0.0, openai_api_key=os.environ["OPENAI"])
+
+
+def create_intro(vacancy, resume):
+
+    template_vacancy_get_skills = """
+    Can you generate me a list of the skills that a candidate is supposed to have for the below vacancy delimited by three backticks. 
+    If you do not know if skills are available mention that you do not know and do not make up an answer. 
+    Mention the skills in 1 to maximum three words for each skill. Return the skills as a JSON list.
+
+    ```
+    {vacancy}
+    ```
+    """
+
+    prompt_vacancy_get_skills = ChatPromptTemplate.from_template(
+        template=template_vacancy_get_skills
+    )
+    vacancy_skills = LLMChain(
+        llm=llm, prompt=prompt_vacancy_get_skills, output_key="vacancy_skills"
+    )
+
+    template_resume_check_skills = """
+    ```
+    {vacancy_skills}
+    ```
+
+    Based on the above list of skills required by a vacancy delimited by backticks,
+    Can you create a JSON object based on the below keys each starting with '-', with respect to the resume below delimited by three backticks?
+
+    - "skills_present": <list the skills present. If no skills are present return an empty list, do not make up an answer. >
+    - "skills_not_present": <list the skills not present. If all skills are present return an empty list, do not make up an answer.>
+    - "score": <calculate a percentage of the number of skills present with respect to the total skills requested>
+
+    ```
+    {resume}
+    ```
+    """
+
+    prompt_resume_check_skills = ChatPromptTemplate.from_template(
+        template=template_resume_check_skills
+    )
+    resume_skills = LLMChain(
+        llm=llm, prompt=prompt_resume_check_skills, output_key="resume_skills"
+    )
+
+    template_resume_past_experiences = """
+    Can you generate me a list of the past work experiences that the candidate has based on the resume below enclosed by three backticks.
+    Mention the experiences in one sentence of medium length. Return the experiences as a JSON list.
+
+    ```
+    {resume}
+    ```
+    """
+
+    prompt_resume_past_experiences = ChatPromptTemplate.from_template(
+        template=template_resume_past_experiences
+    )
+    past_experiences = LLMChain(
+        llm=llm, prompt=prompt_resume_past_experiences, output_key="past_experiences"
+    )
+
+    template_vacancy_check_past_experiences = """
+    ```
+    {past_experiences}
+    ```
+
+    Based on the above list of past experiences by a vacancy delimited by backticks,
+    Can you create a JSON object based on the below keys each starting with '-', with respect to the vacancy below delimited by three backticks?
+
+    - "relevant_experiences": <list the relevant experiences. If no experiences are relevant return an empty list, do not make up an answer. >
+    - "irrelevant_experiences": <list the irrelevant experiences. If all experiences are relevant return an empty list, do not make up an answer.>
+    - "score": <calculate a percentage of the number of skills present with respect to the total skills requested>
+
+    ```
+    {resume}
+    ```
+    """
+
+    prompt_vacancy_check_past_experiences = ChatPromptTemplate.from_template(
+        template=template_vacancy_check_past_experiences
+    )
+    check_past_experiences = LLMChain(
+        llm=llm,
+        prompt=prompt_vacancy_check_past_experiences,
+        output_key="check_past_experiences",
+    )
+
+    template_introduction_email = """
+       You are a recruitment specialist that tries to place the right profiles for the right job.
+       I have a vacancy below the delimiter <VACANCY> and ends with </VACANCY> 
+       and I have a candidate its resume below the delimiter <RESUME> and it ends with </RESUME>.
+
+       <VACANCY>
+       {vacancy}
+       </VACANCY>
+
+       <RESUME>
+       {resume}
+       </RESUME>
+
+       Can you fill in the introduction below and only return as answer this introduction?
+
+       - Role: < the role of the vacancy >
+       - Candidate: < name of the candidate >
+       - Education: < name the education of the candidate >
+       - Experience: < name the 2 most relevant experiences from the candidate for this vacancy. Get them from the "relevant_experiences" key of the JSON object {past_experiences}. If there us no relevant experience, leave this empty. Do not make up an answer or get them from the irrelevant experiences. >
+       - Skills: print here a comma seperated list of the "skills_present" key of the JSON object {resume_skills}
+       """
+
+    prompt_introduction_email = ChatPromptTemplate.from_template(
+        template=template_introduction_email
+    )
+    introduction_email = LLMChain(
+        llm=llm, prompt=prompt_introduction_email, output_key="introduction_email"
+    )
+
+    match_resume_vacancy_skills_chain = SequentialChain(
+        chains=[
+            vacancy_skills,
+            resume_skills,
+            past_experiences,
+            check_past_experiences,
+            introduction_email,
+        ],
+        input_variables=["vacancy", "resume"],
+        output_variables=[
+            "vacancy_skills",
+            "resume_skills",
+            "past_experiences",
+            "check_past_experiences",
+            "introduction_email",
+        ],
+        verbose=False,
+    )
+
+    result = match_resume_vacancy_skills_chain({"vacancy": vacancy, "resume": resume})
+    print(result)
+
+    resume_skills = json.loads(result["resume_skills"])
+    relevant_skills = len(resume_skills["skills_present"])
+    total_skills = len(
+        resume_skills["skills_present"] + resume_skills["skills_not_present"]
+    )
+    score_skills = round(100.0 * (relevant_skills / total_skills), 2)
+
+    check_past_experiences = json.loads(result["check_past_experiences"])
+    relevant_experiences = len(check_past_experiences["relevant_experiences"])
+    total_experiences = len(
+        check_past_experiences["relevant_experiences"]
+        + check_past_experiences["irrelevant_experiences"]
+    )
+    score_experiences = round(100.0 * (relevant_experiences / total_experiences), 2)
+
+    new_line = "\n"
+
+    score = f"""
+    Skills (Score: {score_skills}%)
+    Relevant Skills: {",".join(resume_skills["skills_present"])}
+    Not Relevant Skills: {",".join(resume_skills["skills_not_present"])}    
+    """
+    return result["introduction_email"], score
+
+