Spaces:
Sleeping
Sleeping
File size: 3,400 Bytes
d685a3e dbff29a d685a3e a09e831 d685a3e 9ba6365 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 |
# File: logic/nlp_report.py
import pandas as pd
from huggingface_hub import hf_hub_download
from llama_cpp import Llama
import logging
MODEL_REPO = "bartowski/deepcogito_cogito-v1-preview-llama-8B-GGUF"
MODEL_FILENAME = "deepcogito_cogito-v1-preview-llama-8B-Q8_0.gguf"
model_path = hf_hub_download(
repo_id=MODEL_REPO,
filename=MODEL_FILENAME,
cache_dir="models"
)
llm = Llama(
model_path=model_path,
n_ctx=32048,
chat_format="chatml",
verbose=False
)
def normalize_bool(value):
if isinstance(value, str):
value = value.strip().lower()
return value in ["yes", "true", "1"]
return bool(value)
def generate_patient_summary(patient_row, tone="executive", override_prompt=None):
try:
gaps = patient_row.get('care_gaps', [])
gain = patient_row.get('projected_gain', 0.0)
score = patient_row.get('risk_score', 0.0)
social_flags = []
if normalize_bool(patient_row.get("Lives_Alone")):
social_flags.append("lives alone")
if normalize_bool(patient_row.get("Housing_Instability")):
social_flags.append("experiencing housing instability")
if normalize_bool(patient_row.get("Transportation_Access")):
social_flags.append("has limited transportation access")
if normalize_bool(patient_row.get("Food_Insecurity")):
social_flags.append("faces food insecurity")
if not normalize_bool(patient_row.get("Primary_Care_Established")):
social_flags.append("does not have a primary care provider")
if not normalize_bool(patient_row.get("FollowUp_Scheduled")):
social_flags.append("has no follow-up scheduled")
social_context = "; ".join(social_flags) if social_flags else "no major social risk factors identified"
base_prompt = override_prompt or f"""
You are a healthcare analyst creating summaries for value-based care programs.
Tone: {tone}
Patient ID: {patient_row.get('patient_id')}
Risk Score: {score}
Projected Revenue Gain: ${gain}
Care Gaps Identified: {', '.join(gaps)}
Social Risk Factors: {social_context}
Write a concise and insightful clinical summary, including key gaps and social considerations. Offer targeted recommendations.
"""
response = llm.create_chat_completion(
messages=[{"role": "user", "content": base_prompt}]
)
if "choices" in response and response["choices"]:
return response["choices"][0]["message"]["content"]
else:
return "[Error] No summary returned."
except Exception as e:
logging.error(f"LLM error: {e}")
return "[Error generating summary]"
# Additional: Quality validation helper for UI feedback
def summarize_data_quality(df):
issues = []
required_cols = ['patient_id', 'age', 'gender', 'hcc_codes']
for col in required_cols:
if col not in df.columns:
issues.append(f"Missing column: {col}")
elif df[col].isnull().any():
issues.append(f"Null values in column: {col}")
percent_missing = df.isnull().mean().round(2) * 100
high_missing = percent_missing[percent_missing > 30].to_dict()
if high_missing:
for col, pct in high_missing.items():
issues.append(f"Over 30% missing in: {col} ({pct:.0f}%)")
summary_df = pd.DataFrame({"Issues": issues})
return summary_df
|