Update app.py

app.py

@@ -1,121 +1,190 @@
-import torch
-from datasets import Dataset
-from transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments, Trainer, BitsAndBytesConfig
-from peft import LoraConfig, get_peft_model, prepare_model_for_kbit_training
-model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
-
-bnb_config = BitsAndBytesConfig(
-    load_in_4bit=True,
-    bnb_4bit_compute_dtype=torch.float16,
-    bnb_4bit_use_double_quant=True,
-    bnb_4bit_quant_type="nf4",
-)
-
-model = AutoModelForCausalLM.from_pretrained(
-    model_name,
-    quantization_config=bnb_config,
-    device_map="auto"
-)
-
-tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=True)
-tokenizer.pad_token = tokenizer.eos_token
-model.config.use_cache = False
-model.gradient_checkpointing_enable()
-model = prepare_model_for_kbit_training(model)
-
-lora_config = LoraConfig(
-    r=8,
-    lora_alpha=32,
-    target_modules=["q_proj", "v_proj"],
-    lora_dropout=0.05,
-    bias="none",
-    task_type="CAUSAL_LM"
-)
-
-model = get_peft_model(model, lora_config)
-import pandas as pd
-from datasets import Dataset
-
-
-# Load data from CSV
-df = pd.read_csv("Customer-Support.csv")
-
-
-# Rename columns to match expected keys
-df = df.rename(columns={"query": "instruction", "response": "output"})
-
-
-# Select required columns
-data = df[["instruction", "output"]].fillna("")
-
-
-# Convert DataFrame to list of dictionaries
-data = data.to_dict(orient="records")
-
-
-# Create Hugging Face Dataset
-dataset = Dataset.from_list(data)
-
-
-# Format each example
-def format_instruction(example):
-    return f"### Instruction:\n{example['instruction']}\n\n### Response:\n{example['output']}"
-
-
-# Map formatted text
-dataset = dataset.map(lambda x: {"text": format_instruction(x)})
-
-def tokenize_function(example):
-    tokenized = tokenizer(example["text"], truncation=True, padding="max_length", max_length=512)
-    tokenized["labels"] = tokenized["input_ids"].copy()
-    return tokenized
-
-tokenized_dataset = dataset.map(tokenize_function, batched=True)
-training_args = TrainingArguments(
-    output_dir="./tinyllama-qlora-support-bot",
-    per_device_train_batch_size=2,
-    gradient_accumulation_steps=4,
-    learning_rate=2e-4,
-    logging_dir="./logs",
-    num_train_epochs=3,
-    logging_steps=10,
-    save_total_limit=2,
-    save_strategy="epoch",
-    bf16=True,
-    optim="paged_adamw_8bit"
-)
-trainer = Trainer(
-    model=model,
-    args=training_args,
-    train_dataset=tokenized_dataset,
-    tokenizer=tokenizer
-)
-
-trainer.train()
-model.save_pretrained("tinyllama-qlora-support-bot")
-tokenizer.save_pretrained("tinyllama-qlora-support-bot")
-from transformers import pipeline
-
-pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
-
-instruction = "How do I update the app?"
-prompt = f"### Instruction:\n{instruction}\n\n### Response:\n"
-
-output = pipe(prompt, max_new_tokens=100)
-print(output[0]['generated_text'])
 import gradio as gr
-
-def generate_response(instruction):
-    prompt = f"### Instruction:\n{instruction}\n\n### Response:\n"
-    output = pipe(prompt, max_new_tokens=100, do_sample=True, temperature=0.7)
-    # Extract only the response part
-    response = output[0]["generated_text"].split("### Response:\n")[-1].strip()
-    return response
-
-gr.Interface(
-    fn=generate_response,
-    inputs=gr.Textbox(lines=3, label="Enter your question"),
-    outputs=gr.Textbox(lines=5, label="Support Bot's Response"),
-    title="📞 Customer Support Chatbot",
-    description="Ask a question and get a response from your fine-tuned TinyLLaMA model.",
-).launch()
+import os
+import re
+from langchain_community.document_loaders import PyPDFLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_community.vectorstores import FAISS
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from groq import Groq
+from dotenv import load_dotenv
+from faster_whisper import WhisperModel
+from elevenlabs.client import ElevenLabs
+from gtts import gTTS
+import tempfile
+
+# Load environment variables from .env
+load_dotenv()
+GROQ_API_KEY = os.getenv("GROQ_API_KEY")
+ELEVENLABS_API_KEY = os.getenv("ELEVENLABS_API_KEY")
+
+# Check API keys
+if not GROQ_API_KEY or not ELEVENLABS_API_KEY:
+    raise EnvironmentError("Missing API keys. Please create a .env file with GROQ_API_KEY and ELEVENLABS_API_KEY.")
+
+# Initialize clients
+groq_client = Groq(api_key=GROQ_API_KEY)
+elevenlabs_client = ElevenLabs(api_key=ELEVENLABS_API_KEY)
+whisper_model = WhisperModel("small", device="cpu", compute_type="int8")
+
+def clean_markdown(text):
+    return re.sub(r'[*_#`]+', '', text)
+
+def summarize_resume(resume_text):
+    prompt = f"""Create a concise summary of this resume highlighting:
+    1. Professional title/role
+    2. Years of experience
+    3. Core skills/competencies
+    4. Education background
+    5. Notable achievements
+
+    Resume:
+    {resume_text[:3000]}... [truncated]"""
+    response = groq_client.chat.completions.create(
+        messages=[{"role": "user", "content": prompt}],
+        model="llama3-70b-8192",
+        temperature=0.3,
+    )
+    return clean_markdown(response.choices[0].message.content)
+
+def calculate_ats_score(resume_text):
+    prompt = f"""Analyze this resume and calculate an ATS score (0-100) considering:
+    1. Keyword optimization (20 pts)
+    2. Section organization (20 pts)
+    3. Experience quality (20 pts)
+    4. Education completeness (20 pts)
+    5. Readability (20 pts)
+
+    Return ONLY the numerical score and nothing else.
+
+    Resume:
+    {resume_text[:3000]}... [truncated]"""
+    response = groq_client.chat.completions.create(
+        messages=[{"role": "user", "content": prompt}],
+        model="llama3-70b-8192",
+        temperature=0,
+    )
+    try:
+        return int(response.choices[0].message.content.strip())
+    except:
+        return 50
+
+def process_resume(file):
+    try:
+        loader = PyPDFLoader(file.name)
+        docs = RecursiveCharacterTextSplitter(
+            chunk_size=1000,
+            chunk_overlap=200,
+            separators=["\n\n", "\n", " ", ""]
+        ).split_documents(loader.load())
+        embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
+        FAISS.from_documents(docs, embeddings).save_local("resume_index")
+        full_text = "\n".join([doc.page_content for doc in docs])
+        gr.Info("✅ Resume processed successfully!")
+        return summarize_resume(full_text), f"ATS Score: {calculate_ats_score(full_text)}/100"
+    except Exception as e:
+        gr.Warning(f"❌ Error: {e}")
+        return f"Error: {e}", "ATS Score: N/A"
+
+def transcribe_audio(audio_path):
+    if not audio_path:
+        return "No audio recorded"
+    segments, _ = whisper_model.transcribe(audio_path)
+    return " ".join([segment.text for segment in segments])
+
+def generate_question(resume_text):
+    prompt = f"""Generate one general interview question focusing on:
+    - Teamwork experiences
+    - Challenges overcome
+    - Learning experiences
+    - Career motivations
+    - Problem-solving examples
+
+    Make it conversational and open-ended.
+
+    Resume Excerpt:
+    {resume_text[:2000]}... [truncated]"""
+    response = groq_client.chat.completions.create(
+        messages=[{"role": "user", "content": prompt}],
+        model="llama3-70b-8192",
+        temperature=0.7,
+    )
+    return clean_markdown(response.choices[0].message.content)
+
+def evaluate_response(question, response_text):
+    prompt = f"""Evaluate this interview response on:
+    1. Clarity (1-5)
+    2. Confidence (1-5)
+    3. Relevance (1-5)
+    4. Suggested improvements
+
+    Question: {question}
+    Response: {response_text}"""
+    evaluation = groq_client.chat.completions.create(
+        messages=[{"role": "user", "content": prompt}],
+        model="llama3-70b-8192",
+        temperature=0.2,
+    )
+    return clean_markdown(evaluation.choices[0].message.content)
+
+def gtts_speak(text):
+    try:
+        if not text.strip():
+            raise ValueError("Empty text")
+        tts = gTTS(text, lang="en", tld="com")
+        with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False) as tmp:
+            tts.save(tmp.name)
+            return tmp.name
+    except Exception as e:
+        gr.Warning(f"gTTS Error: {e}")
+        return None
+
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("<h1 style='font-size: 3em; text-align: center;'>🚀 Ready Set Hire</h1>")
+
+    with gr.Tab("📄 Resume Analysis"):
+        with gr.Row():
+            with gr.Column():
+                resume_upload = gr.File(label="📄 Upload Resume (PDF)", file_types=[".pdf"])
+                process_btn = gr.Button("🔍 Analyze Resume", variant="primary")
+            with gr.Column():
+                resume_summary = gr.Textbox(label="📝 Resume Summary", lines=10)
+                hear_summary_btn = gr.Button("🔊 Hear Summary")
+                summary_audio = gr.Audio(visible=True)
+                ats_score = gr.Textbox(label="📊 ATS Compatibility Score", interactive=False)
+        process_btn.click(fn=process_resume, inputs=resume_upload, outputs=[resume_summary, ats_score])
+        hear_summary_btn.click(fn=gtts_speak, inputs=resume_summary, outputs=summary_audio)
+
+    with gr.Tab("🎤 Mock Interview"):
+        with gr.Row():
+            with gr.Column():
+                audio_input = gr.Audio(
+                    label="🎤 Record Your Response",
+                    sources=["microphone"],
+                    type="filepath",
+                    interactive=True
+                )
+                transcribe_btn = gr.Button("📝 Transcribe Response")
+                question_box = gr.Textbox(label="❓ Current Question")
+                generate_btn = gr.Button("🤖 Generate New Question")
+                gtts_question_btn = gr.Button("🔊 Hear Question")
+                question_audio = gr.Audio(visible=True)
+            with gr.Column():
+                transcription = gr.Textbox(label="💬 Your Response")
+                evaluation = gr.Textbox(label="📝 Feedback", lines=8)
+                gtts_feedback_btn = gr.Button("🔊 Hear Feedback")
+                feedback_audio = gr.Audio(visible=True)
+
+        transcribe_btn.click(fn=transcribe_audio, inputs=audio_input, outputs=transcription)
+        generate_btn.click(fn=generate_question, inputs=resume_summary, outputs=question_box)
+        transcription.change(fn=evaluate_response, inputs=[question_box, transcription], outputs=evaluation)
+        gtts_question_btn.click(fn=gtts_speak, inputs=question_box, outputs=question_audio)
+        gtts_feedback_btn.click(fn=gtts_speak, inputs=evaluation, outputs=feedback_audio)
+
+    gr.Markdown("""
+    <div style='text-align:center; margin-top:2em; color:gray'>
+      🚀 Built by Cognify.AI
+    </div>
+    """)
+
+if __name__ == "__main__":
+    demo.launch()