app.py

import base64
import io
import os
import pandas as pd
from docx import Document
from io import BytesIO
import dash
import dash_bootstrap_components as dbc
from dash import html, dcc, Input, Output, State, callback_context
import google.generativeai as genai
from docx.shared import Pt
from docx.enum.style import WD_STYLE_TYPE
from PyPDF2 import PdfReader
from io import StringIO

# Initialize Dash app
app = dash.Dash(__name__, external_stylesheets=[dbc.themes.BOOTSTRAP])

# Configure Gemini AI
genai.configure(api_key=os.environ["GEMINI_API_KEY"])
model = genai.GenerativeModel('gemini-2.5-pro-preview-03-25')

# Global variables
uploaded_files = {}
current_document = None
document_type = None
shredded_document = None
pink_review_document = None

# Document types and their descriptions
document_types = {
    "Shred": "Generate a requirements spreadsheet of the Project Work Statement (PWS) identified by action words like shall, will, perform etc. by pws section, requirement.  Do not write as if you're responding to the proposal.  Its a spreadsheet to distill the requirements, not microhealth's approach",
    "Pink": "Create a Pink Team document based on the PWS outline.  Your goal is to be compliant and compelling.",
    "Pink Review": "Evaluate compliance of the Pink Team document against the requirements and output a spreadsheet of non compliant findings by pws number, the goal of that pws section, what made it non compliant and your recommendations for recovery",
    "Red": "Produce a Red Team document based on the Pink Review by pws sections. Your goal is to be compliant and compelling by recovering all the findings in Pink Review",
    "Red Review": "Evaluate compliance of the Red Team document against the requirements and output a spreadsheet of non compliant findings by pws number, the goal of that pws section, what made it non compliant and your recommendations for recovery",
    "Gold": "Create a Pink Team document based on the PWS response by pws sections. Your goal is to be compliant and compelling by recovering all the findings in Red Review",
    "Gold Review": "Perform a final compliance review against the requirements and output a spreadsheet of non compliant findings by pws number, the goal of that pws section, what made it non compliant and your recommendations for recovery",
    "Virtual Board": "Based on the requirements and in particular the evaulation criteria, you will evaluate the proposal as if you were a contracting office and provide section by section evaluation as unsatisfactory, satisfactory, good, very good, excellent and why in a spreadsheet",
    "LOE": "Generate a Level of Effort (LOE) breakdown as a spreadsheet"
}

app.layout = dbc.Container([
    dbc.Row([
        dbc.Col([
            html.H4("Proposal Documents", className="mt-3 mb-4"),
            dcc.Upload(
                id='upload-document',
                children=html.Div([
                    'Drag and Drop or ',
                    html.A('Select Files')
                ]),
                style={
                    'width': '100%',
                    'height': '60px',
                    'lineHeight': '60px',
                    'borderWidth': '1px',
                    'borderStyle': 'dashed',
                    'borderRadius': '5px',
                    'textAlign': 'center',
                    'margin': '10px 0'
                },
                multiple=True
            ),
            html.Div(id='file-list'),
            html.Hr(),
            html.Div([
                dbc.Button(
                    doc_type,
                    id=f'btn-{doc_type.lower().replace("_", "-")}',
                    color="link",
                    className="mb-2 w-100 text-left custom-button",
                    style={'overflow': 'hidden', 'text-overflow': 'ellipsis', 'white-space': 'nowrap'}
                ) for doc_type in document_types.keys()
            ])
        ], width=3),
        dbc.Col([
            html.Div(id='status-bar', className="alert alert-info", style={'marginBottom': '20px'}),
            dcc.Loading(
                id="loading-indicator",
                type="dot",
                children=[html.Div(id="loading-output")]
            ),
            html.Div(id='document-preview', className="border p-3 mb-3"),
            dbc.Button("Download Document", id="btn-download", color="success", className="mt-3"),
            dcc.Download(id="download-document"),
            html.Hr(),
            html.Div(id='pink-review-upload', style={'display': 'none'}, children=[
                dcc.Upload(
                    id='upload-pink-review',
                    children=html.Div(['Drag and Drop or ', html.A('Select Pink Review File')]),
                    style={
                        'width': '100%',
                        'height': '60px',
                        'lineHeight': '60px',
                        'borderWidth': '1px',
                        'borderStyle': 'dashed',
                        'borderRadius': '5px',
                        'textAlign': 'center',
                        'margin': '10px 0'
                    },
                    multiple=False
                ),
                html.Div(id='pink-review-file-name')
            ]),
            dcc.Loading(
                id="chat-loading",
                type="dot",
                children=[
                    dbc.Input(id="chat-input", type="text", placeholder="Chat with AI to update document...", className="mb-2"),
                    dbc.Button("Send", id="btn-send-chat", color="primary", className="mb-3"),
                    html.Div(id="chat-output")
                ]
            )
        ], width=9)
    ])
], fluid=True)

def process_document(contents, filename):
    content_type, content_string = contents.split(',')
    decoded = base64.b64decode(content_string)
    try:
        if filename.lower().endswith('.docx'):
            doc = Document(BytesIO(decoded))
            text = "\n".join([para.text for para in doc.paragraphs])
            return text
        elif filename.lower().endswith('.pdf'):
            pdf = PdfReader(BytesIO(decoded))
            text = ""
            for page in pdf.pages:
                text += page.extract_text()
            return text
        else:
            return f"Unsupported file format: {filename}. Please upload a PDF or DOCX file."
    except Exception as e:
        return f"Error processing document: {str(e)}"

@app.callback(
    Output('file-list', 'children'),
    Output('status-bar', 'children'),
    Input('upload-document', 'contents'),
    State('upload-document', 'filename'),
    State('file-list', 'children')
)
def update_output(list_of_contents, list_of_names, existing_files):
    global uploaded_files, shredded_document
    if list_of_contents is not None:
        new_files = []
        for i, (content, name) in enumerate(zip(list_of_contents, list_of_names)):
            file_content = process_document(content, name)
            uploaded_files[name] = file_content
            new_files.append(html.Div([
                html.Button('×', id={'type': 'remove-file', 'index': name}, style={'marginRight': '5px', 'fontSize': '10px'}),
                html.Span(name)
            ]))
        if existing_files is None:
            existing_files = []
        shredded_document = None  # Reset shredded document when new files are uploaded
        return existing_files + new_files, "Document uploaded. Please click 'Shred' to proceed."
    return existing_files, "Please upload a document and click 'Shred' to begin."

@app.callback(
    Output('file-list', 'children', allow_duplicate=True),
    Output('status-bar', 'children', allow_duplicate=True),
    Input({'type': 'remove-file', 'index': dash.ALL}, 'n_clicks'),
    State('file-list', 'children'),
    prevent_initial_call=True
)
def remove_file(n_clicks, existing_files):
    global uploaded_files, shredded_document
    ctx = dash.callback_context
    if not ctx.triggered:
        raise dash.exceptions.PreventUpdate
    removed_file = ctx.triggered[0]['prop_id'].split(',')[0].split(':')[-1].strip('}')
    uploaded_files.pop(removed_file, None)
    shredded_document = None  # Reset shredded document when a file is removed
    return [file for file in existing_files if file['props']['children'][1]['props']['children'] != removed_file], "Document removed. Please upload a document and click 'Shred' to begin."

def generate_document(document_type, file_contents):
    prompt = f"""Generate a {document_type} based on the following project artifacts:
{' '.join(file_contents)}
Instructions:
1. Create the {document_type} as a detailed document.
2. Use proper formatting and structure.
3. Include all necessary sections and details.
4. Start the output immediately with the document content.
5. IMPORTANT: If the document type is Pink, Red, Gold and not review type, loe or board
then your goal is to be compliant and compelling based on the
requrements, write in paragraph in active voice as
MicroHealth, limit bullets, answer the
requirement with what MicroHealth will do 
to satisfy the requirement, the technical 
approach with innovation for efficiency, 
productivity, quality and measurable 
outcomes, the industry standard that 
methodology is based on if applicable, 
detail the workflow or steps to accomplish 
the requirement with labor categories that 
will do those tasks in that workflow, 
reference reputable research like gartner, 
forrester, IDC, Deloitte, Accenture etc 
with measures of success and substantiation 
of MicroHealth's approach. Never use soft words
like maybe, could be, should, possible be definitive in your language and confident.
6. you must also take into account section L&M of the document which is the evaluation criteria
to be sure we address them.
Now, generate the {document_type}:
"""

    response = model.generate_content(prompt)
    return response.text

@app.callback(
    Output('document-preview', 'children'),
    Output('loading-output', 'children'),
    Output('status-bar', 'children', allow_duplicate=True),
    Output('pink-review-upload', 'style'),
    [Input(f'btn-{doc_type.lower().replace("_", "-")}', 'n_clicks') for doc_type in document_types.keys()],
    State('pink-review-file-name', 'children'),
    prevent_initial_call=True
)
def generate_document_preview(*args):
    global current_document, document_type, shredded_document, pink_review_document
    ctx = dash.callback_context
    if not ctx.triggered:
        raise dash.exceptions.PreventUpdate
    button_id = ctx.triggered[0]['prop_id'].split('.')[0]
    document_type = button_id.replace('btn-', '').replace('-', '_').title()
    pink_review_file = args[-1]

    if not uploaded_files and document_type != "Shred":
        return html.Div("Please upload and shred a document first."), "", "Please upload and shred a document first.", {'display': 'none'}

    if document_type == "Shred":
        if not uploaded_files:
            return html.Div("Please upload a document before shredding."), "", "Please upload a document before shredding.", {'display': 'none'}
        file_contents = list(uploaded_files.values())
        try:
            shredded_document = generate_document(document_type, file_contents)
            return dcc.Markdown(shredded_document), f"{document_type} generated", "Document shredded. You can now proceed with other operations.", {'display': 'none'}
        except Exception as e:
            print(f"Error generating document: {str(e)}")
            return html.Div(f"Error generating document: {str(e)}"), "Error", "An error occurred while shredding the document.", {'display': 'none'}

    if shredded_document is None:
        return html.Div("Please shred a document first."), "", "Please shred a document first.", {'display': 'none'}

    if document_type == "Pink Review":
        return html.Div("Please upload a Pink Team document or use the generated one."), "", "Please upload a Pink Team document or use the generated one.", {'display': 'block'}

    if document_type in ["Red", "Red Review"] and pink_review_document is None:
        return html.Div("Please complete Pink Review first."), "", "Please complete Pink Review first.", {'display': 'none'}

    try:
        if document_type == "Pink Review" and pink_review_file:
            current_document = generate_document(document_type, [pink_review_file, shredded_document])
        elif document_type in ["Red", "Red Review"]:
            current_document = generate_document(document_type, [pink_review_document, shredded_document])
        else:
            current_document = generate_document(document_type, [shredded_document])

        if document_type == "Pink Review":
            pink_review_document = current_document

        return dcc.Markdown(current_document), f"{document_type} generated", f"{document_type} document generated successfully.", {'display': 'none'}
    except Exception as e:
        print(f"Error generating document: {str(e)}")
        return html.Div(f"Error generating document: {str(e)}"), "Error", "An error occurred while generating the document.", {'display': 'none'}

@app.callback(
    Output('pink-review-file-name', 'children'),
    Input('upload-pink-review', 'contents'),
    State('upload-pink-review', 'filename')
)
def update_pink_review_filename(contents, filename):
    if contents is not None:
        return filename
    return ""

@app.callback(
    Output('chat-output', 'children'),
    Output('document-preview', 'children', allow_duplicate=True),
    Input('btn-send-chat', 'n_clicks'),
    State('chat-input', 'value'),
    prevent_initial_call=True
)
def update_document_via_chat(n_clicks, chat_input):
    global current_document, document_type
    if not chat_input or current_document is None:
        raise dash.exceptions.PreventUpdate
    
    prompt = f"""Update the following {document_type} based on this instruction: {chat_input}
Current document:
{current_document}
Instructions:
1. Provide the updated document content.
2. Maintain proper formatting and structure.
3. Incorporate the requested changes seamlessly.
Now, provide the updated {document_type}:
"""
    
    response = model.generate_content(prompt)
    current_document = response.text
    
    return f"Document updated based on: {chat_input}", dcc.Markdown(current_document)

@app.callback(
    Output("download-document", "data"),
    Input("btn-download", "n_clicks"),
    prevent_initial_call=True
)
def download_document(n_clicks):
    global current_document, document_type
    if current_document is None:
        raise dash.exceptions.PreventUpdate
    
    if document_type == "LOE":
        # Create a pandas DataFrame for LOE
        df = pd.read_csv(StringIO(current_document))
        
        # Save the DataFrame to an Excel file
        output = BytesIO()
        with pd.ExcelWriter(output, engine='xlsxwriter') as writer:
            df.to_excel(writer, sheet_name='LOE', index=False)
        
        return dcc.send_bytes(output.getvalue(), f"{document_type}.xlsx")
    else:
        # Create an in-memory Word document
        doc = Document()
        doc.add_paragraph(current_document)
        
        # Save the document to a BytesIO object
        output = BytesIO()
        doc.save(output)
        
        return dcc.send_bytes(output.getvalue(), f"{document_type}.docx")

if __name__ == '__main__':
    print("Starting the Dash application...")
    app.run(debug=False, host='0.0.0.0', port=7860)
    print("Dash application has finished running.")