Instructions for lab 12

Author: siddhi

agent.py

@@ -0,0 +1,151 @@
+import asyncio
+import json
+from langchain_openai import ChatOpenAI
+from langchain_core.prompts import ChatPromptTemplate
+from config import settings
+
+db = {
+    "job_descriptions": {
+        1: "I need an AI Engineer who knows langchain"
+    },
+    "state": {
+        "session123": {
+            "skills": [],
+            "evaluation": []
+        }
+    }
+}
+
+
+def extract_skills(session_id: str, job_id: int) -> list[str]:
+    """Given a job_id, lookup job descriptiona and extract the skills for that job description"""
+    job_id = int(job_id)
+    job_description = db["job_descriptions"][job_id]
+    skills = ["Python", "SQL", "System Design"]
+    db["state"][session_id]["skills"] = skills
+    print(f"\n📋 Extracted skills: {', '.join(skills)}")
+    return skills
+  
+def update_evaluation(session_id: str, skill: str, evaluation_result: bool) -> bool:
+    """This function takes the session_id, skill, and the evaluation result and saves it to the database. Returns success or failure (bool)"""
+    try:
+        print(f"Saving to DB: {skill} - {evaluation_result}")
+        if isinstance(evaluation_result, str):
+            evaluation_result = True if evaluation_result == "True" else False
+        db["state"][session_id]["evaluation"]. append((skill, evaluation_result))
+        return True
+    except KeyError:
+        return False
+
+def transfer_to_skill_evaluator(session_id: str, skill: str) -> bool:
+    """This function takes a skill, evaluates it and returns the evaluation result for the skill as a boolean pass / fail"""
+    result = True
+    print(f"Evaluating skill: {skill}. Result {result}")
+    return result
+
+tools_mapping = {
+    "extract_skills": extract_skills,
+    "update_evaluation": update_evaluation,
+    "transfer_to_skill_evaluator": transfer_to_skill_evaluator
+}
+
+ORCHESTRATOR_SYSTEM_PROMPT = """
+You are an interview orchestrator. Your goal is to evaluate the candidate on the required skills.
+
+# INSTRUCTIONS
+
+Follow the following steps exactly
+
+1. Extract key skills from the job description using extract_skills tool
+2. Then welcome the candidate, explain the screening process and ask the candidate if they are ready 
+3. Then, for EACH skill in the list, use transfer_to_skill_evaluator tool to delegate evaluation
+4. Once you get the response, use the update_evaluation tool to save the evaluation result into the database
+5. Once all skills are evaluated, mention that the screening is complete and thank the candidate for their time
+
+# OUTPUT FORMAT
+
+Output as a JSON following the JSON schema below:
+
+```
+{{
+    "type": "object",
+    "properties": {{
+        "response": {{ "type": "string" }}
+        "tool_name": {{ "type": "string"}},
+        "tool_params": {{
+            "type": "array",
+            "items": {{
+                "type": "object",
+                "properties": {{
+                    "param": {{ "type": "string" }},
+                    "value": {{ "type": "string" }}
+                }}
+            }}
+        }}
+    }},
+    "required": []
+}}
+
+Use the "tool_name" and "tool_params" properties to execute a tool and use the "response" property to reply to the user without a tool call. 
+
+# TOOLS
+
+You have access to the following tools
+
+1. `extract_skills(session_id: str, job_id: int) -> list[str]`
+
+Given a job_id, lookup job descriptiona and extract the skills for that job description
+
+2. `update_evaluation(session_id: str, skill: str, evaluation_result: bool) -> bool`
+
+This function takes the session_id, skill, and the evaluation result and saves it to the database. Returns success or failure (bool)
+
+3. `transfer_to_skill_evaluator(session_id, skill: str) -> bool`
+
+This function takes a skill, evaluates it and returns the evaluation result for the skill as a boolean pass / fail
+"""
+
+ORCHESTRATOR_USER_PROMPT = """
+Start an interview for the following values:
+
+session_id: {session_id}
+job_id: {job_id}
+
+Begin by welcoming the applicant, extracting the key skills, then evaluate each one.
+"""
+
+def run_orchestrator_agent(session_id, job_id):
+    llm = ChatOpenAI(model="gpt-5.1", temperature=0, api_key=settings.OPENAI_API_KEY)
+    messages = [
+        ("system", ORCHESTRATOR_SYSTEM_PROMPT),
+        ("human", ORCHESTRATOR_USER_PROMPT),
+    ]
+    user_reply = ""
+    while user_reply != "bye":
+        orchestrator_prompt = ChatPromptTemplate.from_messages(messages)
+        orchestrator_chain = orchestrator_prompt | llm
+        output = orchestrator_chain.invoke({"job_id": job_id, "session_id": session_id})
+        data = json.loads(output.content)
+        print(f"Output by LLM: {data}")
+        if "response" in data:
+            print(data["response"])
+        if "tool_name" in data and data["tool_name"] != "":
+            tool_name = data["tool_name"]
+            params = {param["param"]: param["value"] for param in data["tool_params"]}
+            tools = tools_mapping[tool_name]
+            tool_output = tools(**params)
+            print(f"TOOL OUTPUT = {tool_output}")
+            messages.append(("assistant", output.content.replace("{", "{{").replace("}", "}}")))
+            messages.append(("ai", str(tool_output)))
+        else:
+            user_reply = input("User: ")
+            messages.append(("human", user_reply))
+
+def main():
+    job_id = 1
+    session_id = "session123"
+    run_orchestrator_agent(session_id, job_id)
+    print(f"FINAL EVALUATION STATUS: {db['state'][session_id]}")
+
+if __name__ == "__main__":
+    main()