app.py

import base64
import os
from datetime import datetime
from openai import OpenAI
import gradio as gr
import oci
import io
import re
import tempfile
from PIL import Image as PILImage
from collections import Counter
import matplotlib.pyplot as plt
from wordcloud import WordCloud

# === OpenAI API Setup ===
openai_api_key = os.environ.get("OPENAI_API_KEY")
if not openai_api_key:
    raise ValueError("OPENAI_API_KEY environment variable is not set.")

client = OpenAI(api_key=openai_api_key)

# === OCI Object Storage Setup ===
oci_config = {
    "user": os.environ.get("OCI_USER"),
    "tenancy": os.environ.get("OCI_TENANCY"),
    "fingerprint": os.environ.get("OCI_FINGERPRINT"),
    "region": os.environ.get("OCI_REGION"),
    "key_content": os.environ.get("OCI_PRIVATE_KEY")
}

namespace = os.environ.get("OCI_NAMESPACE")
bucket_name = os.environ.get("OCI_BUCKET_NAME")

try:
    object_storage = oci.object_storage.ObjectStorageClient(oci_config)
except Exception as e:
    print("Failed to initialize OCI Object Storage client:", e)

# === Prompts ===
system_prompt = (
    "You are a detail-oriented assistant that specializes in transcribing and polishing "
    "handwritten notes from images. Your goal is to turn rough, casual, or handwritten "
    "content into clean, structured, and professional-looking text that sounds like it "
    "was written by a human—not an AI. You do not include icons, emojis, or suggest next "
    "steps unless explicitly instructed."
)

user_prompt_template = (
    "You will receive an image of handwritten notes. Transcribe the content accurately, "
    "correcting any spelling or grammar issues. Then, organize it clearly with headings, "
    "bullet points, and proper formatting. Maintain the original intent and voice of the "
    "author, but enhance readability and flow. Do not add embellishments or AI-style phrasing."
)

# === Encode uploaded bytes ===
def encode_image_to_base64(file_bytes):
    return base64.b64encode(file_bytes).decode("utf-8")

# === Upload transcription result to OCI ===
def upload_to_object_storage(user_name, text):
    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
    filename = f"{user_name.replace(' ', '_')}_{timestamp}.txt"
    object_storage.put_object(
        namespace_name=namespace,
        bucket_name=bucket_name,
        object_name=filename,
        put_object_body=text.encode("utf-8")
    )
    return filename

# === List object storage ===
def list_object_store():
    try:
        objects = object_storage.list_objects(namespace, bucket_name)
        return [obj.name for obj in objects.data.objects if obj.name.endswith(".txt")]
    except Exception as e:
        return [f"Failed to list objects: {str(e)}"]

# === View file contents ===
def view_transcription(file_name):
    try:
        response = object_storage.get_object(namespace, bucket_name, file_name)
        return response.data.text
    except Exception as e:
        return f"Failed to load file: {str(e)}"

# === Analyze content with OpenAI ===
def summarize_selected_files(file_list):
    combined_text = ""
    for name in file_list:
        combined_text += view_transcription(name) + "\n"
    if not combined_text.strip():
        return "No content found."
    response = client.chat.completions.create(
        model="gpt-4-turbo",
        messages=[
            {"role": "system", "content": "You are a summarization expert."},
            {"role": "user", "content": "Please summarize the following transcriptions in detail:\n" + combined_text}
        ],
        max_tokens=1500
    )
    return response.choices[0].message.content

def recommend_from_selected_files(file_list):
    combined_text = ""
    for name in file_list:
        combined_text += view_transcription(name) + "\n"
    if not combined_text.strip():
        return "No content found."
    response = client.chat.completions.create(
        model="gpt-4-turbo",
        messages=[
            {"role": "system", "content": "You are an operations consultant."},
            {"role": "user", "content": "Please recommend next steps based on these transcriptions:\n" + combined_text}
        ],
        max_tokens=1500
    )
    return response.choices[0].message.content

# === Generate word cloud from selected files ===
def generate_word_map_from_files(file_list):
    combined_text = ""
    for name in file_list:
        combined_text += view_transcription(name) + "\n"
    if not combined_text.strip():
        return None

    wordcloud = WordCloud(width=800, height=400, background_color='white').generate(combined_text)
    plt.figure(figsize=(10, 5))
    plt.imshow(wordcloud, interpolation='bilinear')
    plt.axis("off")
    buf = io.BytesIO()
    plt.savefig(buf, format="png")
    buf.seek(0)

    # Upload image to object storage
    filename = f"wordcloud_{datetime.now().strftime('%Y%m%d_%H%M%S')}.png"
    object_storage.put_object(
        namespace_name=namespace,
        bucket_name=bucket_name,
        object_name=filename,
        put_object_body=buf.getvalue()
    )

    buf.seek(0)
    return PILImage.open(buf)

# === Transcription logic ===
def transcribe_image(file_bytes, user_name):
    if not file_bytes:
        return "No image uploaded."
    encoded = encode_image_to_base64(file_bytes)
    image_url = f"data:image/jpeg;base64,{encoded}"
    response = client.chat.completions.create(
        model="gpt-4-turbo",
        messages=[
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": [
                {"type": "text", "text": user_prompt_template},
                {"type": "image_url", "image_url": {"url": image_url}}
            ]}
        ],
        max_tokens=1500
    )
    timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
    result = f"🗓️ Transcribed on: {timestamp}\n\n{response.choices[0].message.content}"
    upload_to_object_storage(user_name, result)
    return result

# === Gradio Interface ===
with gr.Blocks() as app:
    gr.Markdown("## Handwritten Note Transcriber & Analyzer")

    with gr.Row():
        user_dropdown = gr.Dropdown(
            choices=["Jim Goodwin", "Zahabiya Ali rampurawala", "Keith Gauvin"],
            label="Who is uploading this?"
        )

    input_file = gr.File(label="Upload image", type="binary", file_types=[".jpg", ".jpeg", ".png"])
    output_text = gr.Textbox(label="Transcription Output", lines=30)
    input_file.change(fn=transcribe_image, inputs=[input_file, user_dropdown], outputs=output_text)

    gr.Markdown("### List Object Store Contents")
    gr.Button("List Object Store").click(fn=lambda: "\n".join(list_object_store()), outputs=gr.Textbox(label="Object Store Contents"))

    gr.Markdown("### View Transcription")
    file_selector = gr.Dropdown(choices=list_object_store(), label="Select transcription file")
    view_output = gr.Textbox(label="File Content")
    file_selector.change(fn=view_transcription, inputs=file_selector, outputs=view_output)

    gr.Markdown("### Summarize or Recommend")
    file_multiselect = gr.Dropdown(choices=list_object_store(), label="Select files to analyze", multiselect=True)
    summary_output = gr.Textbox(label="Summary of Selected Transcriptions")
    rec_output = gr.Textbox(label="Recommended Next Steps")
    gr.Button("Summarize Files").click(fn=summarize_selected_files, inputs=file_multiselect, outputs=summary_output)
    gr.Button("Recommend from Files").click(fn=recommend_from_selected_files, inputs=file_multiselect, outputs=rec_output)

    gr.Markdown("### Word Cloud from Files")
    wordcloud_image = gr.Image(label="Word Cloud")
    gr.Button("Generate Word Map from Files").click(
        fn=generate_word_map_from_files,
        inputs=file_multiselect,
        outputs=wordcloud_image
    )

# === Launch App ===
if __name__ == "__main__":
    app.launch(share=True)