Skip to main content

Diagnosis Assist

Category: AI & ML
Node id: diagnosis-assist

Standards: HIPAA

AI-powered medical differential diagnosis assistance

What it does​

  • AI-powered differential diagnosis generation
  • Multi-provider support (OpenAI, Anthropic)
  • Specialty-focused analysis (cardiology, neurology, etc.)
  • Confidence scoring for each differential
  • Rare disease inclusion option
  • Urgency classification (emergency, urgent, soon, routine)

Use cases​

  • Emergency department differential diagnosis support
  • Primary care diagnostic decision assistance
  • Specialist consultation preparation
  • Medical education and training scenarios

Runtime​

  • Performance tier: variable

Input ports​

NameTypeRequiredDescription
symptomsjsonYesArray of symptoms (strings or objects with name/duration/severity)
patient_historyjsonNoPatient medical history (medical_history, medications, allergies)
vitalsjsonNoVital signs
lab_resultsjsonNoLaboratory results
demographicsjsonNoPatient demographics (age, gender)

Output ports​

NameTypeRequiredDescription
resultjsonNoMain result containing diagnosis, differential, confidence, reasoning, and recommendations
metadatajsonNoExecution metadata including tokens and diagnosis_count
successbooleanNoWhether the operation succeeded

Settings​

Basic​

SettingNameTypeRequiredDefaultAllowed valuesDescription
ProviderproviderenumNoanthropicopenai, anthropicLLM provider
ModelmodelstringNo""Model name (auto-selected if empty)
Focus Areafocus_areaenumNogeneralgeneral, cardiology, neurology, gastroenterology, respiratory, endocrine, infectious, oncologyMedical specialty focus
Max Differentialsmax_differentialsnumberNo5min 1; max 20Maximum number of differential diagnoses
Include Rare Diseasesinclude_rare_diseasesbooleanNofalseInclude rare diseases in differentials

Auth​

SettingNameTypeRequiredDefaultAllowed valuesDescription
API Keyapi_keysecretNoAPI key for the LLM provider

Advanced​

SettingNameTypeRequiredDefaultAllowed valuesDescription
TemperaturetemperaturenumberNo0.2min 0; max 1Sampling temperature
Max Tokensmax_tokensnumberNo2000min 1Maximum tokens for response

Errors​

CodeDescription
SYMPTOMS_REQUIREDNo symptoms array provided in inputs
INVALID_PROVIDERProvider must be openai or anthropic
LLM_REQUEST_FAILEDLLM API call failed or returned error
PARSE_ERRORFailed to parse LLM response as structured diagnosis

Runtime notes​

  • result.diagnosis is an array of Diagnosis objects with condition, confidence, severity, supporting_findings, reasoning, suggested_tests, red_flags
  • result.differential is the highest severity across all diagnoses; result.confidence is an urgency string (immediate/urgent/soon/routine)
  • LLM response is parsed as JSON array — if parsing fails, a single fallback diagnosis is returned with the raw response in reasoning
  • Always includes a disclaimer in result.recommendations that this is AI-assisted, not a clinical diagnosis

See Build any workflow for how to put nodes together and Ports, types and data flow for how port types connect.

Something unclear? Send documentation feedback by email.