Delete app.py

app.py

@@ -1,180 +0,0 @@
-import os
-import gradio as gr
-from pathlib import Path
-
-UNIVERSAL_PROMPT_PATH = "CAT_universal_prompt.txt"
-MODULE_DIR = "modules"  # <-- now using /modules subfolder
-
-from openai import OpenAI
-from dotenv import load_dotenv
-load_dotenv()
-client = OpenAI()
-
-# Type aliases
-from typing import List, cast
-from openai.types.chat import ChatCompletionMessageParam
-
-
-def call_model(system_prompt: str, history: list[dict[str, str]]) -> str:
-    # Build as simple dicts first
-    msgs: list[dict[str, str]] = [{"role": "system", "content": system_prompt}]
-    for m in history:
-        role = m.get("role")
-        content = m.get("content")
-        if role in ("user", "assistant") and isinstance(content, str):
-            msgs.append({"role": role, "content": content})
-
-    # Cast once at the call site to satisfy the SDK types
-    typed_msgs = cast(List[ChatCompletionMessageParam], msgs)
-
-    resp = client.chat.completions.create(
-        model="gpt-4o-mini",
-        messages=typed_msgs,
-        temperature=0.4,
-    )
-    return resp.choices[0].message.content or ""
-
-
-def load_text(path: str) -> str:
-    with open(path, "r", encoding="utf-8") as f:
-        return f.read()
-
-def assemble_prompt(universal_prompt_text: str, module_text: str) -> str:
-    def extract(label: str) -> str:
-        marker = label + ":"
-        start = module_text.find(marker)
-        if start == -1:
-            return ""
-        start += len(marker)
-        next_markers = ["\nLEARNING OBJECTIVES:", "\nRUBRIC:", "\nMODULE NAME:"]
-        end_positions = [module_text.find(m, start) for m in next_markers if module_text.find(m, start) != -1]
-        end = min(end_positions) if end_positions else len(module_text)
-        return module_text[start:end].strip()
-
-    learning_objectives = extract("LEARNING OBJECTIVES")
-    rubric = extract("RUBRIC")
-
-    prompt = universal_prompt_text.replace("{LEARNING_OBJECTIVES}", learning_objectives)
-    prompt = prompt.replace("{RUBRIC}", rubric)
-    return prompt
-
-def init_state():
-    return {
-        "assembled_prompt": "",
-        "history": [],
-        "mode": "roleplay",
-        "student_turns": 0,
-        "invited_wrap": False,
-        "mentor_step": 0,
-        "student_name": ""
-    }
-
-def start_session(module_file, student_name=""):
-    state = init_state()
-    state["student_name"] = student_name
-    universal = load_text(UNIVERSAL_PROMPT_PATH)
-    module_text = load_text(Path(MODULE_DIR) / module_file)
-
-    # Give the model the name so it can personalize the dialogue (without overusing it)
-    name_hint = f"\n\n[Student first name: {student_name}. Use it naturally once in the opening; don’t overuse.]" if student_name else ""
-    state["assembled_prompt"] = assemble_prompt(universal, module_text) + name_hint
-
-    state["history"].append({"role": "system", "content": state["assembled_prompt"]})
-    opening = call_model(state["assembled_prompt"], state["history"])
-    state["history"].append({"role": "assistant", "content": opening})
-    return state, state["history"]
-
-def chat(user_msg, state):
-    if not user_msg.strip():
-        return "", state["history"], state
-
-    if state["mode"] == "roleplay":
-        state["student_turns"] += 1
-    state["history"].append({"role": "user", "content": user_msg})
-
-    invite_now = False
-    force_now = False
-
-    if state["mode"] == "roleplay":
-        if state["student_turns"] >= 10:
-            force_now = True
-        elif (7 <= state["student_turns"] <= 9) and not state["invited_wrap"]:
-            invite_now = True
-
-    if state["mode"] == "roleplay" and (invite_now or force_now):
-        if not state["invited_wrap"]:
-            invite = ('Thanks for walking through this with me. We’ve covered a lot. '
-                      'Before we wrap up, is there anything else you’d like me to consider '
-                      'before I give you a preliminary assessment?')
-            state["history"].append({"role": "assistant", "content": invite})
-            state["invited_wrap"] = True
-            return "", state["history"], state
-        else:
-            ack = "I appreciate you sharing that. I’ll take it into account. Let’s step back and review how you approached this situation."
-            state["history"].append({"role": "assistant", "content": ack})
-            state["mode"] = "mentor"
-
-    if state["mode"] == "roleplay" and not state["invited_wrap"]:
-        reply = call_model(state["assembled_prompt"], state["history"])
-        state["history"].append({"role": "assistant", "content": reply})
-        return "", state["history"], state
-
-    if state["mode"] == "mentor":
-        # Step 1: general intro (no assumption of tools)
-        if state.get("mentor_step", 0) == 0:
-            eval_intro = (
-                "Before we wrap up: name two specific concepts, tools, or frameworks you used in this scenario, "
-                "and in one short sentence each say how you applied them. If you didn’t use any, name two insights "
-                "you learned and how you would apply them next time."
-            )
-            state["history"].append({"role": "assistant", "content": eval_intro})
-            state["mentor_step"] = 1
-            return "", state["history"], state
-
-        # Step 2: concise rubric-based evaluation
-        else:
-            eval_request = (
-                "Evaluate the student's performance using the module rubric. Provide these sections: "
-                "Overall rating (Unsatisfactory, Satisfactory, or Excellent) with a one-sentence justification; "
-                'Career competencies; Uniquely human capacities; Argument analysis; Ethical frameworks; ESG awareness; '
-                "Application; Interaction quality; Strength; Area to improve; Advice for next time; Fictional consequence. "
-                "Quote at least one student phrase. Keep the whole evaluation under 180 words and end with 'END OF SCENE'."
-            )
-            state["history"].append({"role": "user", "content": eval_request})
-            reply = call_model(state["assembled_prompt"], state["history"])
-            if "END OF SCENE" not in reply:
-                reply += "\n\nEND OF SCENE"
-            state["history"].append({"role": "assistant", "content": reply})
-            state["mode"] = "done"
-            return "", state["history"], state
-
-
-with gr.Blocks(title="CAT (MVP)") as demo:
-    gr.Markdown("## 😼Conversational Assessment Tool (CAT) — MVP")
-    with gr.Row():
-        module_file = gr.Dropdown(
-            label="Select Module File",
-            choices=[p.name for p in sorted(Path(MODULE_DIR).glob("module*.txt"))],
-            value="module01.txt",
-            interactive=True
-        )
-        name_tb = gr.Textbox(label="Your first name", placeholder="e.g., Maya", value="", interactive=True)
-        start_btn = gr.Button("Start")   # fine to keep inside the row (optional)
-
-    chatbot = gr.Chatbot(label="CAT Conversation", type="messages")
-    user_in = gr.Textbox(label="Your message", placeholder="Type here and press Enter")
-    state = gr.State(init_state())
-
-    def _start(module_name, student_name):
-        student_name = student_name.strip()
-        if not student_name:
-            # Return a valid state object plus a warning message in the chat
-            return init_state(), [{"role": "assistant", "content": "⚠ Please enter your first name before starting."}]
-        st, hist = start_session(module_name, student_name)
-        return st, hist
-
-    start_btn.click(_start, [module_file, name_tb], [state, chatbot])
-    user_in.submit(chat, [user_in, state], [user_in, chatbot, state])
-
-if __name__ == "__main__":
-    demo.launch()