70 lines
3.0 KiB
Markdown
70 lines
3.0 KiB
Markdown
# Triage Prompt
|
|
|
|
Your most important task is to first determine if the user input is a valid health entry or an invalid query.
|
|
|
|
**DO NOT categorize requests for medical advice as 'symptom' or 'question'. They are invalid queries.**
|
|
|
|
- If the input is not health-related, respond with: `{"error": "not_health_related"}`
|
|
- If the input asks for medical advice (e.g., contains "what should I do?", "how do I treat?"), respond with: `{"error": "no_medical_advice"}`
|
|
|
|
---
|
|
|
|
User said: "{{INPUT}}"
|
|
|
|
If, and only if, the input is a valid health entry, classify it into a category and extract basic metadata.
|
|
|
|
IMPORTANT: Respond with ONLY the JSON object below. No explanations, no markdown, no text before or after.
|
|
|
|
```json
|
|
{
|
|
"category": "one of the categories below",
|
|
"language": "en|nl|de|hi|...",
|
|
"has_data": true/false,
|
|
"summary": "brief normalized summary"
|
|
}
|
|
```
|
|
|
|
## Invalid Query Examples
|
|
|
|
Input: "What is the capital of Mexico?"
|
|
```json
|
|
{"error": "not_health_related"}
|
|
```
|
|
|
|
Input: "I have a headache, what should I do?"
|
|
```json
|
|
{"error": "no_medical_advice"}
|
|
```
|
|
|
|
## Categories
|
|
|
|
| Category | Use when | Examples |
|
|
|----------|----------|----------|
|
|
| vital | Measurements: weight, blood pressure, temperature, heart rate | "I weigh 80kg", "BP 120/80" |
|
|
| exercise | Physical activity, workouts, steps | "walked 30 min", "did yoga" |
|
|
| medication | Taking/tracking medicines | "took ibuprofen", "started metformin" |
|
|
| supplement | Vitamins, supplements | "taking omega-3", "vitamin D" |
|
|
| nutrition | Food, drink, dietary habits, abstinence | "stopped coffee", "no alcohol", "ate breakfast" |
|
|
| fertility | Menstrual cycle, period, ovulation, pregnancy | "period started", "ovulating", "I'm pregnant" |
|
|
| symptom | Current symptoms, pain, discomfort | "headache today", "knee hurts" |
|
|
| note | General health notes | "feeling good", "stressed" |
|
|
| history | YOUR past medical events (no documentation) | "appendectomy at 12", "had chickenpox", "broke arm 2015" |
|
|
| family_history | Relatives' health conditions | "mother: breast cancer", "father: diabetes" |
|
|
| surgery | Documented surgical procedures | "scheduled for knee surgery" |
|
|
| hospitalization | Hospital stays | "admitted for pneumonia" |
|
|
| consultation | Doctor visits, appointments | "saw cardiologist today" |
|
|
| diagnosis | Active conditions being managed | "diagnosed with epilepsy" |
|
|
| device | Medical devices, implants | "VP shunt placed", "pacemaker settings" |
|
|
| therapy | Therapeutic treatments | "started physical therapy" |
|
|
| assessment | Evaluations, tests | "developmental assessment scheduled" |
|
|
| birth | Birth-related data | "born at 32 weeks" |
|
|
| imaging_finding | Imaging results | "MRI showed..." |
|
|
| eeg_finding | EEG results | "EEG normal" |
|
|
| provider | Doctors, hospitals, clinics | "Dr. Smith is my neurologist" |
|
|
| question | Open questions | "should I worry about..." |
|
|
|
|
## Rules
|
|
|
|
- `has_data`: true if user provided actual values/facts, false if just describing intent
|
|
- `summary`: normalize to third person, e.g. "walked 30 minutes" → "30 min walk"
|