Deploy interview assistant app

Dockerfile

@@ -0,0 +1,30 @@
+FROM python:3.11-slim
+
+# Set environment variables (may need to change authority)
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1
+
+# ENV to inject secret
+# ARG HF_TOKEN
+# ENV HF_TOKEN=${HF_TOKEN}
+
+# Create working directory (and authorised token)
+WORKDIR /app
+
+# Install SSL root certs and system deps required by pymongo + DNS
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    ca-certificates curl dnsutils gcc openssl && \
+    rm -rf /var/lib/apt/lists/*
+
+# Copy and install dependencies
+COPY requirements.txt .
+RUN pip install --upgrade pip && pip install --no-cache-dir -r requirements.txt
+
+# Copy source code (app files)
+COPY . .
+
+# Expose the port used by Uvicorn
+EXPOSE 7860
+
+# Run the FastAPI application using Uvicorn
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860", "--log-level", "debug"]

app.py

@@ -0,0 +1,123 @@
+# Access site: https://binkhoale1812-interview-ai.hf.space/
+import os
+import tempfile
+import psutil
+from pathlib import Path
+from typing import Dict
+
+from fastapi import FastAPI, File, UploadFile, HTTPException
+from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import JSONResponse, FileResponse
+from fastapi.staticfiles import StaticFiles
+
+import google.generativeai as genai
+from transformers import pipeline, AutoProcessor, AutoModelForSpeechSeq2Seq
+
+############################################
+# ── Configuration ────────────────────────
+############################################
+
+GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
+if not GEMINI_API_KEY:
+    raise RuntimeError("GEMINI_API_KEY environment variable must be set!")
+
+# Tiny Whisper model is light enough for CPU Spaces; change if GPU is available
+ASR_MODEL_ID = "openai/whisper-tiny"  # ~39 MB
+ASR_LANGUAGE = "en"  # Force to English for interview setting
+
+############################################
+# ── FastAPI App ───────────────────────────
+############################################
+
+app = FastAPI(title="Interview Q&A Assistant", docs_url="/docs")
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+# Serve frontend assets
+app.mount("/statics", StaticFiles(directory="statics"), name="statics")
+
+############################################
+# ── Global objects (lazy‑loaded) ──────────
+############################################
+
+asr_pipeline = None  # Speech‑to‑text
+llm = None          # Gemini model
+
+
+@app.on_event("startup")
+async def load_models():
+    """Load Whisper."""
+    global asr_pipeline, llm
+    # Whisper tiny – seq2seq pipeline
+    asr_pipeline = pipeline(
+        "automatic-speech-recognition",
+        model=ASR_MODEL_ID,
+        chunk_length_s=30,
+        torch_dtype="auto",
+        device="cpu",
+    )
+
+
+############################################
+# ── Helpers ───────────────────────────────
+############################################
+
+def build_prompt(question: str) -> str:
+    """Craft a prompt that elicits concise, structured answers."""
+    return (
+        "You are a helpful career‑coach AI. Answer the following interview "
+        "question clearly and concisely, offering practical insights when "
+        "appropriate.\n\n"
+        f"Interview question: \"{question}\""
+    )
+
+def memory_usage_mb() -> float:
+    return psutil.Process().memory_info().rss / 1_048_576  # bytes→MiB
+
+############################################
+# ── Routes ────────────────────────────────
+############################################
+
+@app.get("/")
+async def root() -> FileResponse:
+    """Serve the single‑page app."""
+    return FileResponse(Path("statics/index.html"))
+
+
+@app.post("/voice-transcribe")
+async def voice_transcribe(file: UploadFile = File(...)):  # noqa: B008
+    """Receive audio, transcribe, push to Gemini, return answer."""
+    if file.content_type not in {"audio/wav", "audio/x-wav", "audio/mpeg"}:
+        raise HTTPException(status_code=415, detail="Unsupported audio type")
+    # Save to a temp file (Whisper expects a filename/bytes)
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
+        tmp.write(await file.read())
+        tmp_path = tmp.name
+    try:
+        # ── 1. Transcribe
+        transcript: Dict = asr_pipeline(tmp_path, generate_kwargs={"language": ASR_LANGUAGE})
+        question = transcript["text"].strip()
+        if not question:
+            raise ValueError("Empty transcription")
+        # ── 2. LLM answer
+        prompt = build_prompt(question)
+        # Gemini Flash 2.5 – tuned for short latency
+        client = genai.Client(api_key=GEMINI_API_KEY)
+        response = client.models.generate_content(
+                model="gemini-2.5-flash-preview-04-17",
+                contents=prompt
+            )
+        answer = response.text.strip()
+        return JSONResponse(
+            {
+                "question": question,
+                "answer": answer,
+                "memory_mb": round(memory_usage_mb(), 1),
+            }
+        )
+    finally:
+        os.remove(tmp_path) # Rm audio when done

requirements.txt

@@ -0,0 +1,17 @@
+# Core server
+fastapi
+uvicorn[standard]
+aiofiles               # Static file serving
+python-multipart       # File uploads
+
+# Voice‑to‑text (Whisper via 🤗 Transformers)
+transformers>=4.40
+torch
+huggingface_hub
+
+# Gemini Flash 2.5
+google-genai
+python-dotenv          # Optional – read GOOGLE_API_KEY
+
+# Utilities
+psutil                 # Lightweight health logging

statics/index.html

@@ -0,0 +1,28 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+  <meta charset="UTF-8" />
+  <title>Interview Q&A Assistant</title>
+  <meta name="viewport" content="width=device-width, initial-scale=1" />
+  <link rel="stylesheet" href="/statics/styles.css" />
+  <link rel="icon" type="image/png" href="/statics/icon.png" />
+</head>
+<body>
+  <main class="container">
+    <h1>Interview Q&amp;A Assistant</h1>
+    <p class="subtitle">Hold the button, ask your interview question, release to get an answer.</p>
+
+    <button id="record-button" class="record-btn">🎙 Hold&nbsp;to&nbsp;Ask</button>
+
+    <section class="output-section">
+      <h2>Your Question</h2>
+      <pre id="question-output" class="output"></pre>
+
+      <h2>AI&nbsp;Answer</h2>
+      <pre id="answer-output" class="output"></pre>
+    </section>
+  </main>
+
+  <script src="/statics/script.js"></script>
+</body>
+</html>

statics/script.js

@@ -0,0 +1,75 @@
+/*******************************
+ * Interview Q&A Frontend JS   *
+ *******************************/
+
+// Elements
+const recordBtn   = document.getElementById("record-button");
+const questionEl  = document.getElementById("question-output");
+const answerEl    = document.getElementById("answer-output");
+
+// Typing animation util
+function typeEffect(el, text, speed = 30) {
+  el.textContent = "";
+  let idx = 0;
+  const timer = setInterval(() => {
+    el.textContent += text.charAt(idx);
+    idx++;
+    if (idx >= text.length) clearInterval(timer);
+  }, speed);
+}
+
+// Audio recording setup
+let mediaRecorder = null;
+let chunks = [];
+
+async function initMedia() {
+  try {
+    const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
+    mediaRecorder = new MediaRecorder(stream);
+
+    mediaRecorder.ondataavailable = e => chunks.push(e.data);
+
+    mediaRecorder.onstop = async () => {
+      const audioBlob = new Blob(chunks, { type: "audio/wav" });
+      chunks = [];
+
+      // Build form data
+      const form = new FormData();
+      form.append("file", audioBlob, "record.wav");
+
+      // UX feedback
+      typeEffect(questionEl, "⌛ Transcribing…");
+      answerEl.textContent = "";
+
+      try {
+        const res = await fetch("https://binkhoale1812-interview-ai.hf.space/voice-transcribe", { method: "POST", body: form });
+        if (!res.ok) throw new Error(`HTTP ${res.status}`);
+        const data = await res.json();
+
+        typeEffect(questionEl, data.question || "[no speech detected]");
+        setTimeout(() => typeEffect(answerEl, data.answer || "[no answer]"), 500);
+      } catch (err) {
+        typeEffect(answerEl, "❌ " + err.message);
+      }
+    };
+  } catch (err) {
+    alert("Microphone access denied – please allow permissions.");
+  }
+}
+
+// Hold‑to‑record UX
+function bindRecordBtn() {
+  if (!mediaRecorder) return;
+  recordBtn.addEventListener("mousedown", () => mediaRecorder.start());
+  recordBtn.addEventListener("mouseup",   () => mediaRecorder.stop());
+
+  // Touch devices
+  recordBtn.addEventListener("touchstart", e => { e.preventDefault(); mediaRecorder.start(); });
+  recordBtn.addEventListener("touchend",   e => { e.preventDefault(); mediaRecorder.stop(); });
+}
+
+// Init on page load
+window.addEventListener("DOMContentLoaded", async () => {
+  await initMedia();
+  bindRecordBtn();
+});

statics/styles.css

@@ -0,0 +1,35 @@
+/* Simple, clean aesthetic */
+:root {
+  --primary: #0052cc;
+  --accent:  #ff6666;
+  --bg:      #f8f9fc;
+  --mono:    "Courier New", monospace;
+}
+
+html,body {
+  margin: 0; padding: 0; background: var(--bg); font-family: Arial, sans-serif;
+}
+
+.container {
+  max-width: 720px; margin: 40px auto; padding: 24px;
+  background: #fff; border-radius: 8px; box-shadow: 0 4px 12px rgba(0,0,0,.08);
+}
+
+h1 { margin-top: 0; text-align: center; color: var(--primary); }
+.subtitle { text-align: center; color: #444; margin-bottom: 32px; }
+
+.record-btn {
+  display: block; margin: 0 auto 24px; padding: 14px 28px;
+  background: var(--accent); color: #fff; border: none; border-radius: 50px;
+  font-size: 17px; cursor: pointer; transition: background .25s;
+}
+
+.record-btn:hover { background: #ff4d4d; }
+
+.output-section h2 { margin: 24px 0 8px; color: var(--primary); }
+
+.output {
+  background: #000; color: #0f0; padding: 16px; min-height: 60px;
+  border-radius: 4px; overflow-x: auto; font-family: var(--mono);
+  white-space: pre-wrap; word-wrap: break-word;
+}

temp_hf_space

@@ -0,0 +1 @@
+Subproject commit b818ff92402f2d47358e68c4219e456d1dc9a815