Updated structure v1.2

app.py

@@ -1,56 +1,37 @@
-# app.py
 import gradio as gr
-from core.utils import generate_first_question
-from core.mbti_analyzer import analyze_mbti
-from core.interviewer import generate_question
-
-def analyze_and_ask(user_text, prev_count):
-    """Пошаговый генератор — стриминг без async и без streaming=True."""
-    if not user_text.strip():
-        yield "⚠️ Please enter your answer.", "", prev_count
-        return
-
-    try:
-        n = int(prev_count.split("/")[0]) + 1
-    except Exception:
-        n = 1
-    counter = f"{n}/30"
-
-    # 1️⃣ Шаг 1 — анализ
-    mbti_gen = analyze_mbti(user_text)
-    mbti_text = ""
-    for chunk in mbti_gen:
-        mbti_text = chunk
-        yield mbti_text, "💭 Interviewer is thinking...", counter
-
-    # 2️⃣ Шаг 2 — вопрос
-    interviewer_gen = generate_question("default_user", user_text)
-    next_q = ""
-    for chunk in interviewer_gen:
-        next_q = chunk
-        yield mbti_text, next_q, counter
-
-# --------------------------------------------------------------
-# Gradio интерфейс
-# --------------------------------------------------------------
-with gr.Blocks(theme=gr.themes.Soft(), title="MBTI Personality Interviewer") as demo:
-    gr.Markdown("## 🧠 MBTI Personality Interviewer\nОпредели личностный тип и получи следующий вопрос от интервьюера.")
-
-    with gr.Row():
-        with gr.Column(scale=1):
-            inp = gr.Textbox(
-                label="Ваш ответ",
-                placeholder="Например: I enjoy working with people and organizing events.",
-                lines=4
-            )
-            btn = gr.Button("Анализировать и задать новый вопрос", variant="primary")
-        with gr.Column(scale=1):
-            mbti_out = gr.Textbox(label="📊 Анализ MBTI", lines=4)
-            interviewer_out = gr.Textbox(label="💬 Следующий вопрос от интервьюера", lines=3)
-            progress = gr.Textbox(label="⏳ Прогресс", value="0/30")
-
-    btn.click(analyze_and_ask, inputs=[inp, progress], outputs=[mbti_out, interviewer_out, progress])
-
-    demo.load(lambda: ("", generate_first_question(), "0/30"), inputs=None, outputs=[mbti_out, interviewer_out, progress])
-
-demo.queue(max_size=20).launch(server_name="0.0.0.0", server_port=7860)
+from core.interviewer_phi3 import generate_question, categories
+from core.analyzer_mbti import classify_answer
+
+state = {"i": 0, "history": []}
+
+def next_step(user_answer):
+    # Сохраняем ответ
+    if user_answer.strip():
+        state["history"].append(user_answer)
+    else:
+        return "⚠️ Please type your answer.", ""
+
+    # Анализ MBTI
+    traits = classify_answer(user_answer)
+    traits_text = "\n".join([f"{t['label']} → {t['score']:.3f}" for t in traits])
+
+    # Следующий вопрос
+    if state["i"] < len(categories):
+        cat = categories[state["i"]]
+        q = generate_question(state["history"], cat)
+        state["i"] += 1
+    else:
+        q = "✅ Interview finished. Personality summary calculated."
+
+    return traits_text, q
+
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("## 🧠 MBTI Interviewer (Phi-3)")
+    inp = gr.Textbox(label="Answer", lines=3)
+    btn = gr.Button("Next Question", variant="primary")
+    out1 = gr.Textbox(label="📊 MBTI Analysis", lines=3)
+    out2 = gr.Textbox(label="💬 Next Question", lines=3)
+    btn.click(fn=next_step, inputs=inp, outputs=[out1, out2])
+    demo.load(lambda: ("", generate_question([], categories[0])), outputs=out2)
+
+demo.launch()

core/analyzer_mbti.py

@@ -0,0 +1,9 @@
+from transformers import pipeline
+
+MBTI_MODEL = "f3nsmart/MBTIclassifier"
+classifier = pipeline("text-classification", model=MBTI_MODEL, return_all_scores=True)
+
+def classify_answer(answer: str):
+    res = classifier(answer)[0]
+    sorted_res = sorted(res, key=lambda x: x["score"], reverse=True)
+    return sorted_res[:3]  # top 3 traits

core/interviewer_phi3.py

@@ -0,0 +1,34 @@
+from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
+import random, json, os
+
+MODEL_NAME = "microsoft/Phi-3-mini-4k-instruct"
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, torch_dtype="auto", device_map="auto")
+
+generator = pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=70, temperature=0.7, top_p=0.9)
+
+DATA_PATH = "data"
+categories = sorted([f.replace(".json", "") for f in os.listdir(DATA_PATH) if f.endswith(".json")])
+
+def load_category_sample(cat_name):
+    path = os.path.join(DATA_PATH, f"{cat_name}.json")
+    with open(path, "r", encoding="utf-8") as f:
+        data = json.load(f)
+    return random.choice(data).get("instruction", "")
+
+def generate_question(history, current_cat):
+    """
+    Сценарий генерации вопроса по текущей категории MBTI.
+    """
+    sample = load_category_sample(current_cat)
+    hist_text = "\n".join([f"Q{i//2+1 if i%2==0 else ''}: {h}" for i, h in enumerate(history)])
+    prompt = (
+        f"You're generating interview questions for MBTI testing.\n"
+        f"Previous dialogue:\n{hist_text}\n"
+        f"Generate one new open-ended question related to {current_cat.replace('_', ' ')} "
+        f"based on this example:\n'{sample}'\n"
+        f"Do not repeat or rephrase previous ones. Output only the question text."
+    )
+    output = generator(prompt)[0]["generated_text"]
+    q = output.split("\n")[-1].strip()
+    return q

core/mbti_analyzer.py

@@ -1,19 +0,0 @@
-# core/mbti_analyzer.py
-from transformers import pipeline
-import asyncio
-
-MBTI_MODEL = "f3nsmart/MBTIclassifier"
-mbti_pipe = pipeline("text-classification", model=MBTI_MODEL, return_all_scores=True)
-
-async def analyze_mbti_async(user_text: str):
-    """Асинхронный MBTI-анализ."""
-    loop = asyncio.get_event_loop()
-    return await loop.run_in_executor(None, lambda: mbti_pipe(user_text)[0])
-
-def analyze_mbti(user_text: str):
-    """Генератор для стриминга результата."""
-    yield "⏳ Analyzing personality traits..."
-    res = asyncio.run(analyze_mbti_async(user_text))
-    res_sorted = sorted(res, key=lambda x: x["score"], reverse=True)
-    mbti_text = "\n".join([f"{r['label']} → {r['score']:.3f}" for r in res_sorted[:3]])
-    yield mbti_text