import asyncio
from datetime import datetime
import traceback
from typing import TypedDict
import gradio as gr
import uuid
import os
import pandas as pd
from tqdm.auto import tqdm
from common import get_db
from config import SheamiConfig
from graph import create_graph
from modules.models import HealthReport, SheamiMilestone, SheamiState, SheamiUser
from pdf_reader import read_pdf
from gradio_modal import Modal
from plot_helper import render_vitals_plot_layout
from report_formatter import render_patient_state
from dotenv import load_dotenv
load_dotenv(override=True)
MAX_FILES = int(os.getenv("MAX_FILES", 3))
async def process_reports(user_email: str, patient_id: str, files: list):
if not files:
yield construct_process_message(message="Please upload at least one PDF file.")
return
yield construct_process_message(
message=f"Initiating processing of {len(files)} files ..."
)
thread_id = str(uuid.uuid4())
workflow = create_graph(
user_email=user_email, patient_id=patient_id, thread_id=thread_id
)
uploaded_reports = []
for file in files:
file_path = file.name
contents = read_pdf(file_path)
# print("contents = [", contents,"]")
uploaded_reports.append(
HealthReport(
report_file_name_with_path=file_path,
report_file_name=os.path.basename(file_path),
report_contents=contents,
)
)
state = SheamiState(
uploaded_reports=uploaded_reports,
thread_id=thread_id,
user_email=user_email,
patient_id=patient_id,
)
config = {"configurable": {"thread_id": thread_id}, "recursion_limit": 50}
# Streaming node progress
buffer = ""
final_state = state
try:
async for msg_packet in workflow.astream(
state, config=config, stream_mode="values"
):
final_state = msg_packet
# try:
# print("units_processed = ", msg_packet["units_processed"])
# except Exception as e:
# print("units_processed not available in: ", msg_packet.keys() , e)
units_processed = (
msg_packet["units_processed"] if "units_processed" in msg_packet else 0
)
units_total = (
msg_packet["units_total"] if "units_total" in msg_packet else 6
)
process_desc = (
msg_packet["process_desc"]
if "process_desc" in msg_packet
else "Working on it ..."
)
overall_units_processed = (
msg_packet["overall_units_processed"]
if "overall_units_processed" in msg_packet
else 0
)
overall_units_total = (
msg_packet["overall_units_total"]
if "overall_units_total" in msg_packet
else 0
)
if "messages" in msg_packet and msg_packet["messages"]:
# Either show all so far, or just latest message
all_but_last = msg_packet["messages"][:-1]
last_message = msg_packet["messages"][-1]
buffer = "\n".join(all_but_last)
yield construct_process_message(
message=buffer,
current_step=process_desc,
units_processed=units_processed,
units_total=units_total,
overall_units_processed=overall_units_processed,
overall_units_total=overall_units_total,
milestones=msg_packet["milestones"],
)
buffer += "\n"
for c in last_message:
buffer += c
yield construct_process_message(
message=buffer,
current_step=process_desc,
units_processed=units_processed,
units_total=units_total,
overall_units_processed=overall_units_processed,
overall_units_total=overall_units_total,
milestones=msg_packet["milestones"],
)
await asyncio.sleep(0.005)
await asyncio.sleep(0.1)
buffer += (
"\n\n"
f"✅ Processed {len(files)} reports.\n"
"Please download the output file from below within 5 min."
)
except Exception as e:
print("Error processing stream", e)
traceback.print_exc()
buffer += f"\n\n❌ Error processing reports. {e}"
final_state["milestones"][-1].status = "failed"
final_state["milestones"][-1].end_time = datetime.now()
# update latest milestone status as failed
await get_db().add_or_update_milestone(
run_id=final_state["run_id"],
milestone=final_state["milestones"][-1].step_name,
status="failed",
end=True,
)
# update run status with failed
await get_db().update_run_stats(
run_id=final_state["run_id"], status="failed", message=f"{e}"
)
finally:
print("In finally ...", final_state["pdf_path"])
if final_state["pdf_path"]:
yield construct_process_message(
message=buffer,
final_output=gr.update(value=final_state["pdf_path"], visible=True),
milestones=final_state["milestones"],
reports_output=msg_packet["standardized_reports"],
trends_output=msg_packet["trends_json"],
)
else:
print("Yielding error message")
yield construct_process_message(
message=buffer,
final_output=gr.update(visible=False),
milestones=final_state["milestones"],
reports_output=msg_packet["standardized_reports"],
trends_output=msg_packet["trends_json"],
error=True,
)
def generate_milestones_data(
num_rows=5,
headers=[
"Step",
"Status",
"Start Time",
"End Time",
"Duration (s)",
],
):
steps = [
"Consume & Standardize Reports",
"Standardize Test Names",
"Standardize Measurement Units",
"Aggregate Trends",
"Interpret & Plot Trends",
]
data = []
for i in range(num_rows):
if i < len(steps):
# First column = step name, rest empty
row = [steps[i]] + ["" for _ in headers[1:]]
else:
# All empty row
row = ["" for _ in headers]
data.append(row)
return headers, data
def disable_component():
return gr.update(interactive=False)
def enable_component():
return gr.update(interactive=True)
def construct_process_message(
message: str,
final_output: str = None,
current_step: str = None,
units_processed: int = 0,
units_total: int = 0,
overall_units_processed: int = 0,
overall_units_total: int = 0,
milestones: list[SheamiMilestone] = [],
reports_output=None,
trends_output=None,
error=False,
):
try:
if units_total > 0:
overall_pct_complete = (
(overall_units_processed + (units_processed / units_total))
/ overall_units_total
) * 100
else:
overall_pct_complete = (overall_units_processed / overall_units_total) * 100
except ZeroDivisionError:
overall_pct_complete = 0
# overall_pct_complete = (
# max(0, int(100 * overall_units_processed / overall_units_total))
# if overall_units_total != 0
# else 0
# )
message = message.replace("\n", "
")
formatted_message = (
f"{message}
"
if not final_output
else f"{message}
"
)
final_message = ""
if final_output:
if error:
final_message = "❌ There was an error processing your request. Please try after sometime."
else:
final_message = "✅ Your health trends report is ready for download!"
else:
final_message = ""
return (
formatted_message, # logs_textbox
disable_component() if not final_output else enable_component(), # run_btn
hide_component() if not final_output else show_component(), # file_input
final_output, # pdf_download
(
hide_component() if not final_output else show_component()
), # final_report_container
milestones_to_rows(milestones), # milestones_df
*render_patient_state(
reports_output, trends_output
), # reports_output, trends_output
final_message, # final_message
)
def render_logo():
return gr.Image(
type="filepath",
value=SheamiConfig.logo_path,
label="My Logo",
show_label=False,
show_download_button=False,
show_fullscreen_button=False,
show_share_button=False,
interactive=False, # <- disables the toolbar
container=False, # <- removes the border/frame if you want it cleaner
)
def render_logo_small():
return gr.Image(
type="filepath",
value=SheamiConfig.logo_small_path,
label="My Logo",
show_label=False,
show_share_button=False,
show_download_button=False,
show_fullscreen_button=False,
interactive=False, # <- disables the toolbar
container=False, # <- removes the border/frame if you want it cleaner
)
def render_banner():
return gr.Image(
type="filepath",
value=SheamiConfig.banner_path,
label="Banner",
show_label=False,
show_share_button=False,
show_download_button=False,
show_fullscreen_button=False,
interactive=False, # <- disables the toolbar
container=False, # <- removes the border/frame if you want it cleaner
)
def toggle_logo_small(logo):
print(logo)
new_logo = (
SheamiConfig.logo_path if "logo-small" in logo else SheamiConfig.logo_small_path
)
return gr.update(value=new_logo)
def clear_component():
return gr.update(value=None)
def hide_component():
return gr.update(visible=False)
def show_component():
return gr.update(visible=True)
def close_side_bar():
return gr.update(open=False)
def open_side_bar():
return gr.update(open=True)
def make_status_tab_active():
return gr.update(selected="my_status_container")
def make_final_report_tab_active():
return gr.update(selected="my_final_report_container")
def milestones_to_rows(milestones: list[SheamiMilestone]) -> list[list]:
num_rows = 10
headers, data = generate_milestones_data(num_rows=num_rows)
for i, m in enumerate(milestones):
if i >= len(data):
break # avoid overflow if more milestones than rows
data[i] = [
m.step_name,
m.status_icon,
m.start_time.strftime("%H:%M:%S") if m.start_time else "",
m.end_time.strftime("%H:%M:%S") if m.end_time else "",
f"{m.time_taken:.2f}" if m.time_taken else "",
]
return data
def handle_file_input_change(files):
if files:
if len(files) > MAX_FILES:
return (
hide_component(),
f"❌ Maximum of {MAX_FILES} files can be uploaded at a time.",
)
else:
return show_component(), f"✅ {len(files)} selected."
else:
return hide_component(), "❌ No files selected."
def get_css():
return """
/* Container spacing */
.pro-radio .wrap {
display: flex;
flex-direction: column;
gap: 8px;
}
/* Hide the default radio dot */
.pro-radio input[type="radio"] {
display: none !important;
}
/* Base card look */
.pro-radio label {
display: block;
background: #fafafa;
color: #222;
font-family: "Inter", sans-serif;
font-size: 15px;
font-weight: 500;
padding: 12px 16px;
border-radius: 8px;
border: 1px solid #ddd;
cursor: pointer;
transition: all 0.2s ease;
outline: none !important;
box-shadow: none !important;
}
/* Hover state */
.pro-radio label:hover {
background: #f0f0f0;
border-color: #bbb;
}
/* Selected card */
.pro-radio input[type="radio"]:checked + span {
background: #e6f0ff; /* light blue background */
border: 1px solid #0066cc; /* blue border */
border-radius: 8px;
font-weight: 600;
color: #0066cc;
display: block;
padding: 12px 16px;
outline: none !important;
box-shadow: none !important;
}
/* Kill any weird inner focus box */
.pro-radio span {
outline: none !important;
box-shadow: none !important;
border: none !important;
}
/* Remove Gradio's green selected background on the LABEL itself */
.pro-radio label:has(input[type="radio"]:checked),
.pro-radio label[aria-checked="true"],
.pro-radio label[data-selected="true"],
.pro-radio .selected,
.pro-radio [data-selected="true"] {
background: #fafafa !important; /* or transparent */
box-shadow: none !important;
border-color: #ddd !important;
}
/* Keep your selected look on the SPAN only (no inner blue box) */
.pro-radio input[type="radio"]:checked + span {
background: #e6f0ff;
border: 1px solid #0066cc;
border-radius: 8px;
display: block;
padding: 12px 16px;
color: #0066cc;
font-weight: 600;
outline: none !important;
box-shadow: none !important;
}
/* Hide native dot + any focus rings */
.pro-radio input[type="radio"] { display: none !important; }
.pro-radio label,
.pro-radio label:focus,
.pro-radio label:focus-within,
.pro-radio input[type="radio"]:focus + span {
outline: none !important;
box-shadow: none !important;
}
.highlighted-text {
color: #FFD700; /* bright gold to stand out */
font-weight: bold; /* makes it pop */
font-family: "Courier New", monospace; /* subtle variation */
background-color: #222; /* faint background contrast */
padding: 0 3px; /* like a tag highlight */
border-radius: 3px; /* smooth corners */
}
#patient-card{
border: 1px solid rgba(0,0,0,0.06);
background: #fafafa;
border-radius: 10px;
padding: 10px;
box-sizing: border-box;
gap: 12px;
}
#logged_in_user {
text-align : center
}
#logged_in_user input textarea {
font-weight : bold;
color : #00FF00;
text-align : center !important;
}
#add_patient_modal {
width : 400px;
}
.dots {
display: inline-block;
min-width: 1.5em; /* enough space for 3 dots */
text-align: left;
color: #00FF00;
}
.dots::after {
content: " .";
min-width : 100px;
animation: dots 1.5s steps(3, end) infinite;
}
@keyframes dots {
0% { content: " "; }
33% { content: " ."; }
66% { content: " .."; }
100% { content: " ..."; }
}
div.transparent_div {
color: #00FF00; /* classic terminal green */
background-color: #111111; /* softer black background */
font-family: monospace; /* console-like font */
font-size: 14px;
line-height: 1.4;
border: none; /* clean console feel */
outline: none;
resize: none;
padding: 8px;
min-height: 300px;
}
#transparent_textbox input,
#transparent_textbox textarea {
color: #00FF00; /* classic terminal green */
background-color: #111111; /* softer black background */
font-family: monospace; /* console-like font */
font-size: 14px;
line-height: 1.4;
border: none; /* clean console feel */
outline: none;
resize: none;
padding: 8px;
}
#transparent_textbox textarea {
overflow-y: auto; /* keep scroll if logs overflow */
}
#centered_col {
display: flex;
justify-content: center; /* center horizontally */
align-items: center; /* center vertically */
height: 100px; /* or depending on your desired height */
}
.text-center {
text-align : center
}
"""
def get_app_title():
return "SHEAMI"
def get_app_theme():
return gr.themes.Ocean()
def get_gradio_block(
container,
user_email_state,
patient_id_state,
fn_callback,
fn_callback_inputs=[],
fn_callback_outputs=[],
):
# gr.Markdown("## 🩺 SHEAMI - Smart Healthcare Excellence Through Artificial Medical Intelligence")
with container:
my_logo = render_logo()
with gr.Row(equal_height=False):
with gr.Column() as inputs_container:
file_input = gr.File(
file_types=[".pdf"],
type="filepath",
file_count="multiple",
label="Upload your Lab Reports (PDF)",
)
file_upload_status = gr.Markdown()
with gr.Row():
gr.Column()
run_btn = gr.Button(
"Process Reports", variant="primary", visible=False, scale=0
)
gr.Column()
# Add CSS to vertically center button inside its column
with gr.Tabs(
visible=False, selected="my_status_container"
) as output_container:
with gr.Tab(
"Report Download", id="my_final_report_container"
) as final_report_container:
final_message = gr.Markdown()
with gr.Row(equal_height=False):
pdf_download = gr.DownloadButton(
label="Download 📊",
scale=0,
)
# Populate file when button clicked
upload_more_reports_btn = gr.Button(
"Upload more", variant="primary", scale=0
)
gr.Column()
with gr.Accordion(
"Standardized Reports", open=False, visible=False
):
reports_output = gr.HTML()
with gr.Accordion("Trends", open=False, visible=False):
trends_output = gr.Code(language="json")
with gr.Tab(
"Run Statistics", id="my_status_container"
) as status_container:
with gr.Row(equal_height=True):
(headers, empty_data) = generate_milestones_data()
milestone_df = gr.DataFrame(
value=empty_data,
headers=headers,
datatype=["str", "str", "str", "str", "str"],
interactive=False,
row_count=5,
)
with gr.Column():
logs_textbox = gr.HTML(
value="Processing request
",
label="Logs",
container=False,
elem_id="transparent_textbox",
)
file_input.change(
handle_file_input_change,
inputs=[file_input],
outputs=[run_btn, file_upload_status],
)
run_btn.click(toggle_logo_small, inputs=[my_logo], outputs=[my_logo]).then(
hide_component, outputs=[inputs_container]
).then(show_component, outputs=[output_container]).then(
show_component, outputs=[logs_textbox]
).then(
show_component, outputs=[status_container]
).then(
make_status_tab_active, outputs=[output_container]
).then(
process_reports,
inputs=[user_email_state, patient_id_state, file_input],
outputs=[
logs_textbox,
run_btn,
file_input,
pdf_download,
final_report_container,
milestone_df,
reports_output,
trends_output,
final_message,
],
queue=True,
).then(
make_final_report_tab_active, outputs=[output_container]
).then(
clear_component, outputs=[file_input]
).then(
fn_callback, outputs=fn_callback_outputs, inputs=fn_callback_inputs
)
upload_more_reports_btn.click(hide_component, outputs=[output_container]).then(
show_component, outputs=[inputs_container]
).then(toggle_logo_small, inputs=[my_logo], outputs=[my_logo]).then(
clear_component, outputs=[logs_textbox]
)
def build(user_email, patient_id):
# Build Gradio UI
with gr.Blocks(
theme=get_app_theme(),
title=get_app_title(),
css=get_css(),
) as sheami_app:
user_email_state = gr.State(user_email)
patient_id_state = gr.State(patient_id)
get_gradio_block(
gr.Column(),
user_email_state=user_email_state,
patient_id_state=patient_id_state,
fn_callback=lambda: None,
)
return sheami_app
def render_selected_patient_actions():
with gr.Column(scale=4):
selected_patient_info = gr.Markdown("⚠ No patient selected")
with gr.Row():
delete_patient_btn = gr.Button(
"❌ Delete",
size="sm",
scale=0,
variant="stop",
interactive=False,
)
edit_patient_btn = gr.Button(
"✏️ Edit",
size="sm",
scale=0,
variant="huggingface",
interactive=False,
)
upload_reports_btn = gr.Button(
"⬆️ Upload",
size="sm",
scale=0,
variant="huggingface",
interactive=False,
)
add_vitals_btn = gr.Button(
"🩺 Add Vitals",
scale=0,
variant="huggingface",
size="sm",
interactive=False,
)
return (
selected_patient_info,
delete_patient_btn,
edit_patient_btn,
upload_reports_btn,
add_vitals_btn,
)
def render_top_menu_bar(logged_in_user: SheamiUser = None):
with gr.Row(elem_classes="menu-bar") as menu_bar:
with gr.Column(scale=2, visible=False) as sheami_logo_container:
sheami_logo = render_logo()
# (selected_patient_info,delete_patient_btn,edit_patient_btn,upload_reports_btn, add_vitals_btn) = render_selected_patient_actions()
gr.Column() # spacer
with gr.Row(scale=2):
if logged_in_user:
gr.Image(
value=logged_in_user.picture_url,
scale=0,
container=False,
show_download_button=False,
show_fullscreen_button=False,
show_share_button=False,
height=30,
visible=False,
)
with gr.Column(
scale=4,
):
gr.Markdown(
value=logged_in_user.name,
elem_classes="text-center",
visible=False,
)
email_in = gr.Text(
label="👤 You are logged in as",
placeholder="doctor1@sheami.com",
value=logged_in_user.email,
interactive=False,
elem_id="logged_in_user",
text_align="left",
show_label=False,
container=False,
elem_classes="text-center",
visible=False,
)
# gr.Button(
# "Logout",
# link="/logout",
# variant="huggingface",
# size="sm",
# visible=True,
# )
else:
email_in = gr.Text()
return (
sheami_logo_container,
email_in,
# selected_patient_info,
# delete_patient_btn,
# edit_patient_btn,
# upload_reports_btn,
# add_vitals_btn,
)
# ------------------------------------------------------------
# Dummy function to simulate saving and showing AI ranges
# ------------------------------------------------------------
async def save_vitals_readings(
patient_id,
reading_date: datetime,
height,
weight,
bp_sys,
bp_dia,
glucose,
pbs,
spo2,
custom_name,
custom_value,
custom_unit,
created_by_user="some_user",
):
if not patient_id:
return "⚠️ Please select a patient from the sidebar.", []
readings = []
if height:
readings.append({"name": "Height", "value": height, "unit": "cm"})
if weight:
readings.append({"name": "Weight", "value": weight, "unit": "kg"})
if bp_sys and bp_dia:
readings.append({"name": "BP", "value": f"{bp_sys}/{bp_dia}", "unit": "mmHg"})
if glucose:
readings.append({"name": "Fasting Glucose", "value": glucose, "unit": "mg/dL"})
if pbs:
readings.append({"name": "PBS", "value": pbs, "unit": "mg/dL"})
if spo2:
readings.append({"name": "SpO₂", "value": spo2, "unit": "%"})
if custom_name and custom_value:
readings.append(
{"name": custom_name, "value": custom_value, "unit": custom_unit}
)
await get_db().save_readings_to_db(
patient_id, reading_date, readings, created_by_user
)
# TODO: call AI here to attach personalized ranges + status
# For now, just return the raw readings
# table = [
# [
# reading_date.strftime("%Y-%m-%d"),
# r["name"],
# r["value"],
# r["unit"],
# "pending AI analysis",
# ]
# for r in readings
# ]
vitals_history = await get_db().get_vitals_by_patient(patient_id)
latest_vitals = await render_latest_vitals_card_layout(patient_id)
vitals_plots = await render_vitals_plot_layout(patient_id)
# print("ui: vitals = ", vitals)
return (
f"✅ Saved for {patient_id} on {reading_date}",
flatten_vitals(vitals_history),
*latest_vitals,
*vitals_plots,
)
def flatten_vitals(docs):
rows = []
for doc in docs:
reading_date = (
doc["date"].strftime("%Y-%m-%d")
if isinstance(doc["date"], datetime)
else doc["date"]
)
for r in doc.get("readings", []):
rows.append(
{
"date": reading_date,
"name": r.get("name", ""),
"value": r.get("value", ""),
"unit": r.get("unit", ""),
"status": r.get("status", "pending AI analysis"),
}
)
df = pd.DataFrame(rows)
if df.empty:
return pd.DataFrame(columns=["date", "name", "value", "unit", "status"])
df = df.fillna("-")
return df
async def render_latest_vitals_card_layout(patient_id: str):
"""
Retrieve the latest vital readings for a patient and generate exactly 20 Gradio Label cards.
If more than 20 readings are present, only the first 20 are used (truncated).
If fewer than 20 readings are available, the output is padded with empty Label cards to reach a total of 20.
Args:
patient_id (str): Unique identifier of the patient whose vital readings are to be fetched.
Returns:
list[gr.Label]: A list of 20 Gradio Label components, each displaying a vital reading or padding as needed.
"""
vitals = await get_db().get_latest_vitals_by_patient(patient_id)
readings = vitals.get("readings", [])
cards = []
# Truncate readings if more than 20
readings = readings[:20]
for reading in sorted(readings, key=lambda x: x["name"]):
cards.append(
gr.Label(
value=f"{reading['value']}{reading['unit']}",
label=reading["name"],
visible=True,
)
)
# Pad with empty cards if less than 20
while len(cards) < 20:
cards.append(gr.Label(value="-", label="", visible=True))
return cards
def empty_state_component(
message: str,
title: str = "No data available",
icon: str = "ℹ️",
):
return gr.HTML(
f"""
""",
label=None,
)
def hide_tabs_if_no_patient_selected(patient_id):
if patient_id:
return show_component(), hide_component()
else:
return hide_component(), show_component()
def show_no_data_found_if_none(data):
if not data:
return show_component(), hide_component() # no data found # actual component
else:
return hide_component(), show_component() # no data found # actual component
def show_no_data_found_if_dataframe_empty(data: pd.DataFrame):
print("rows:", len(data), "cols:", len(data.columns), "empty:", data.empty)
# Case 1: Truly empty (no rows, no cols)
if data.empty:
return show_component(), hide_component()
# Case 2: All rows are completely NaN/blank
if data.dropna(how="all").shape[0] == 0:
return show_component(), hide_component()
# Case 3: All columns are completely NaN/blank
if data.dropna(axis=1, how="all").shape[1] == 0:
return show_component(), hide_component()
# Otherwise → show actual data
return hide_component(), show_component()
def render_about_markdowns():
with gr.Column() as group:
gr.Markdown("# 🧪 How It Works")
gr.Markdown(
"""
- 🤖 Upload your lab reports (**PDF only**) to unlock **AI-powered insights** on test results and **personalized vitals**.
- 🫀 Vitals such as **height, weight, BMI, and other demographics** are factored in to give you **contextualized, patient-specific insights**.
- 📊 When multiple reports are available, **Sheami™** highlights **trends over time** in both lab tests and vitals.
- 💾 All uploaded reports and generated insights stay securely in your workspace for you to **review, download, or remove** at any time.
""",
show_copy_button=False,
)
gr.Markdown("---")
gr.Markdown(
"""
> ⚠️ **Disclaimer**
> This application is intended solely for informational and educational purposes.
> It is **not** a substitute for professional medical advice, diagnosis, or treatment.
> Always seek the guidance of a qualified healthcare provider with any questions
> you may have regarding a medical condition. Never disregard professional advice
> or delay seeking it because of information provided by this app.
""",
show_copy_button=False,
)
gr.Markdown("---")
gr.Markdown("By clicking **Proceed**, you agree to these terms.")
return group