app.py

import base64
import io
import os
import pandas as pd
from docx import Document
from io import BytesIO, StringIO
import dash
import dash_bootstrap_components as dbc
from dash import html, dcc, Input, Output, State, callback_context, MATCH, ALL
from docx.shared import Pt
from docx.enum.style import WD_STYLE_TYPE
from PyPDF2 import PdfReader
import openai
import logging
import threading

logging.basicConfig(level=logging.INFO, format='%(asctime)s %(levelname)s %(message)s')

app = dash.Dash(__name__, external_stylesheets=[dbc.themes.BOOTSTRAP], suppress_callback_exceptions=True)

openai.api_key = os.environ.get("OPENAI_API_KEY", "")

uploaded_files = {}
current_document = None
document_type = None
shredded_document = None
pink_review_document = None
uploaded_doc_contents = {}

document_types = {
    "Shred": "Ignore all other instructions and generate only 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": "Ignore all other instructions and generate and evaluate compliance of the Pink Team document against the requirements and output only 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": "Ignore all other instructions and generate and evaluate compliance of the Red Team document against the requirements and output a only 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": "Ignore all other instructions and generate and perform a final compliance review against the requirements and output only 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": "Ignore all other instructions and generate and 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 and produce only spreadsheet",
    "LOE": "Ignore all other instructions and generate and generate a Level of Effort (LOE) breakdown and produce only spreadsheet"
}

def get_right_col_content(selected_type):
    if selected_type == "Shred":
        return dbc.Card(
            dbc.CardBody([
                html.Div([
                    html.Div(className="blinking-dot", style={'margin':'0 auto','width':'16px','height':'16px'}),
                ], style={'textAlign':'center', 'marginBottom':'10px'}),
                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(),
                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", style={'whiteSpace':'pre-wrap'}),
                        dbc.Button("Send", id="btn-send-chat", color="primary", className="mb-3"),
                        html.Div(id="chat-output")
                    ]
                )
            ])
        )
    else:
        return dbc.Card(
            dbc.CardBody([
                html.Div([
                    html.Div(className="blinking-dot", style={'margin':'0 auto','width':'16px','height':'16px'}),
                ], style={'textAlign':'center', 'marginBottom':'10px'}),
                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([
                    html.Label(f"Upload {selected_type} Document"),
                    dcc.Upload(
                        id={'type': 'upload-doc-type', 'index': selected_type},
                        children=html.Div(['Drag and Drop or ', html.A('Select File')]),
                        style={
                            'width': '100%',
                            'height': '60px',
                            'lineHeight': '60px',
                            'borderWidth': '1px',
                            'borderStyle': 'dashed',
                            'borderRadius': '5px',
                            'textAlign': 'center',
                            'margin': '10px 0'
                        },
                        multiple=False
                    ),
                    html.Div(id={'type': 'uploaded-doc-name', 'index': selected_type}),
                    dbc.RadioItems(
                        id={'type': 'radio-doc-source', 'index': selected_type},
                        options=[
                            {'label': 'Loaded Document', 'value': 'loaded'},
                            {'label': 'Uploaded Document', 'value': 'uploaded'}
                        ],
                        value='loaded',
                        inline=True,
                        className="mb-2"
                    ),
                    dbc.Button("Generate Document", id={'type': 'btn-generate-doc', 'index': selected_type}, color="primary", className="mb-3"),
                ], id={'type': 'doc-type-controls', 'index': selected_type}),
                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", style={'whiteSpace':'pre-wrap'}),
                        dbc.Button("Send", id="btn-send-chat", color="primary", className="mb-3"),
                        html.Div(id="chat-output")
                    ]
                )
            ])
        )

app.layout = dbc.Container([
    dcc.Store(id='selected-doc-type', data="Shred"),
    dbc.Row([
        dbc.Col(
            dbc.Card(
                dbc.CardBody([
                    html.H4("Proposal Documents", className="mt-3 mb-4"),
                    html.Div([
                        html.Div(className="blinking-dot", style={'margin':'0 auto','width':'16px','height':'16px'}),
                    ], style={'textAlign':'center', 'marginBottom':'10px'}),
                    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(
                        id='doc-type-buttons'
                    )
                ])
            ),
            width=3
        ),
        dbc.Col(
            html.Div(id='right-col-content'),
            width=9
        )
    ])
], fluid=True)

@app.callback(
    Output('doc-type-buttons', 'children'),
    Input('selected-doc-type', 'data')
)
def render_doc_type_buttons(selected_type):
    buttons = []
    for doc_type in document_types.keys():
        btn_style = {'overflow': 'hidden', 'text-overflow': 'ellipsis', 'white-space': 'nowrap'}
        btn_class = "mb-2 w-100 text-left custom-button"
        if doc_type == selected_type:
            btn_class += " active-doc-type"
        buttons.append(
            dbc.Button(
                doc_type,
                id={'type': 'btn-doc-type', 'index': doc_type},
                color="link",
                className=btn_class,
                style=btn_style
            )
        )
    return buttons

@app.callback(
    Output('selected-doc-type', 'data'),
    [Input({'type': 'btn-doc-type', 'index': ALL}, 'n_clicks')],
    [State({'type': 'btn-doc-type', 'index': ALL}, 'id')],
    prevent_initial_call=True
)
def update_selected_doc_type(n_clicks_list, btn_ids):
    triggered = callback_context.triggered
    if not triggered or all(x is None for x in n_clicks_list):
        raise dash.exceptions.PreventUpdate
    idx = [i for i, x in enumerate(n_clicks_list) if x]
    if idx:
        selected_type = btn_ids[idx[-1]]['index']
    else:
        selected_type = "Shred"
    return selected_type

@app.callback(
    Output('right-col-content', 'children'),
    Input('selected-doc-type', 'data')
)
def update_right_col(selected_type):
    return get_right_col_content(selected_type)

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:
                page_text = page.extract_text()
                if page_text:
                    text += page_text
            return text
        else:
            return f"Unsupported file format: {filename}. Please upload a PDF or DOCX file."
    except Exception as e:
        logging.error(f"Error processing document: {str(e)}")
        return f"Error processing document: {str(e)}"

@app.callback(
    Output('file-list', '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
        logging.info("Documents uploaded and file list updated.")
        return existing_files + new_files
    return existing_files

@app.callback(
    Output('file-list', 'children', allow_duplicate=True),
    Input({'type': 'remove-file', 'index': 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
    logging.info(f"Removed file: {removed_file}")
    return [file for file in existing_files if file['props']['children'][1]['props']['children'] != removed_file]

@app.callback(
    Output('document-preview', 'children'),
    Output('loading-output', 'children'),
    Input({'type': 'btn-doc-type', 'index': 'Shred'}, 'n_clicks'),
    prevent_initial_call=True
)
def generate_shred_doc(n_clicks):
    global current_document, document_type, shredded_document
    if not uploaded_files:
        return html.Div("Please upload a document before shredding."), ""
    file_contents = list(uploaded_files.values())
    try:
        shredded_document = generate_document("Shred", file_contents)
        current_document = shredded_document
        return dcc.Markdown(shredded_document), "Shred generated"
    except Exception as e:
        logging.error(f"Error generating document: {str(e)}")
        return html.Div(f"Error generating document: {str(e)}"), "Error"

@app.callback(
    Output({'type': 'uploaded-doc-name', 'index': MATCH}, 'children'),
    Output({'type': 'upload-doc-type', 'index': MATCH}, 'contents'),
    Output({'type': 'radio-doc-source', 'index': MATCH}, 'value'),
    Input({'type': 'upload-doc-type', 'index': MATCH}, 'contents'),
    State({'type': 'upload-doc-type', 'index': MATCH}, 'filename'),
    State({'type': 'upload-doc-type', 'index': MATCH}, 'id')
)
def update_uploaded_doc_name(contents, filename, id_dict):
    if contents is not None:
        uploaded_doc_contents[id_dict['index']] = (contents, filename)
        logging.info(f"{id_dict['index']} file uploaded: {filename}")
        return filename, contents, "uploaded"
    return "", None, "loaded"

@app.callback(
    Output('document-preview', 'children', allow_duplicate=True),
    Output('loading-output', 'children', allow_duplicate=True),
    Input({'type': 'btn-generate-doc', 'index': ALL}, 'n_clicks'),
    State({'type': 'btn-generate-doc', 'index': ALL}, 'id'),
    State({'type': 'radio-doc-source', 'index': ALL}, 'value'),
    State({'type': 'upload-doc-type', 'index': ALL}, 'contents'),
    State({'type': 'upload-doc-type', 'index': ALL}, 'filename'),
    prevent_initial_call=True
)
def generate_other_doc(n_clicks_list, btn_ids, radio_values, upload_contents, upload_filenames):
    global current_document, document_type, shredded_document, pink_review_document
    ctx = callback_context
    if not ctx.triggered:
        raise dash.exceptions.PreventUpdate
    idx = [i for i, x in enumerate(n_clicks_list) if x]
    if not idx:
        raise dash.exceptions.PreventUpdate
    idx = idx[-1]
    doc_type = btn_ids[idx]['index']
    document_type = doc_type

    if doc_type == "Shred":
        raise dash.exceptions.PreventUpdate

    if shredded_document is None:
        return html.Div("Please shred a document first."), ""

    source = radio_values[idx] if radio_values and len(radio_values) > idx else 'loaded'
    doc_content = None

    if source == 'uploaded':
        if upload_contents and len(upload_contents) > idx and upload_contents[idx] and upload_filenames and len(upload_filenames) > idx and upload_filenames[idx]:
            doc_content = process_document(upload_contents[idx], upload_filenames[idx])
        else:
            return html.Div("Please upload a document to use as source."), ""
    else:
        if doc_type == "Pink":
            doc_content = shredded_document
        elif doc_type == "Pink Review":
            doc_content = pink_review_document if pink_review_document else ""
        elif doc_type == "Red":
            doc_content = pink_review_document if pink_review_document else ""
        elif doc_type == "Red Review":
            doc_content = pink_review_document if pink_review_document else ""
        elif doc_type == "Gold":
            doc_content = shredded_document
        elif doc_type == "Gold Review":
            doc_content = shredded_document
        elif doc_type == "Virtual Board":
            doc_content = shredded_document
        elif doc_type == "LOE":
            doc_content = shredded_document
        else:
            doc_content = shredded_document

    try:
        if doc_type == "Pink Review":
            current_document = generate_document(doc_type, [doc_content, shredded_document])
            pink_review_document = current_document
        elif doc_type in ["Red", "Red Review"]:
            current_document = generate_document(doc_type, [doc_content, shredded_document])
        else:
            current_document = generate_document(doc_type, [doc_content])
        logging.info(f"{doc_type} document generated successfully.")
        return dcc.Markdown(current_document), f"{doc_type} generated"
    except Exception as e:
        logging.error(f"Error generating document: {str(e)}")
        return html.Div(f"Error generating document: {str(e)}"), "Error"

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 
Pink Review, Red Review, Gold Review, or spreadsheet, loe or virtual 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}:
"""

    logging.info(f"Generating document for type: {document_type}")
    try:
        response = openai.ChatCompletion.create(
            model="gpt-4-1106-preview",
            messages=[
                {"role": "system", "content": "You are a helpful, expert government proposal writer."},
                {"role": "user", "content": prompt}
            ],
            max_tokens=4096,
            temperature=0.25,
        )
        generated_text = response['choices'][0]['message']['content']
        logging.info("Document generated successfully.")
        return generated_text
    except Exception as e:
        logging.error(f"Error generating document: {str(e)}")
        raise

@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}:
"""

    logging.info(f"Updating document via chat for {document_type} instruction: {chat_input}")
    try:
        response = openai.ChatCompletion.create(
            model="gpt-4-1106-preview",
            messages=[
                {"role": "system", "content": "You are a helpful, expert government proposal writer."},
                {"role": "user", "content": prompt}
            ],
            max_tokens=4096,
            temperature=0.2,
        )
        current_document = response['choices'][0]['message']['content']
        logging.info("Document updated via chat successfully.")
        return f"Document updated based on: {chat_input}", dcc.Markdown(current_document)
    except Exception as e:
        logging.error(f"Error updating document via chat: {str(e)}")
        return f"Error updating document: {str(e)}", html.Div(f"Error updating document: {str(e)}")

@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":
        try:
            df = pd.read_csv(StringIO(current_document))
            output = BytesIO()
            with pd.ExcelWriter(output, engine='xlsxwriter') as writer:
                df.to_excel(writer, sheet_name='LOE', index=False)
            logging.info("LOE document downloaded as Excel.")
            return dcc.send_bytes(output.getvalue(), f"{document_type}.xlsx")
        except Exception as e:
            logging.error(f"Error downloading LOE document: {str(e)}")
            return dcc.send_string(f"Error downloading LOE: {str(e)}", f"{document_type}_error.txt")
    else:
        try:
            doc = Document()
            doc.add_paragraph(current_document)
            output = BytesIO()
            doc.save(output)
            logging.info("Document downloaded as Word.")
            return dcc.send_bytes(output.getvalue(), f"{document_type}.docx")
        except Exception as e:
            logging.error(f"Error downloading document: {str(e)}")
            return dcc.send_string(f"Error downloading document: {str(e)}", f"{document_type}_error.txt")

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