coding_diagnosis extracts the ICD-10-CM diagnosis codes that belong on the claim from the encounter documentation. It is part of the Nanonets Coding Engine tool group.
The full senior-auditor reasoning prompt (ported from Fleming ProcessVisitForDiagnosisCoding) — including the two-condition test, uncertainty/historical exclusions, same-family deduplication, active-medication rescue, and ordering tiers — is baked into the tool.
Inputs
encounter_context(required): the documentation to code from. May wrap the target encounter in<current_encounter>tags and include nearby encounters as context.practice_specific_icd_rules(optional): practice/organization rules. Only ICD-10-relevant rules are applied; CPT/modifier/billing rules are ignored by the prompt.
Output
code_list is a flat, claim-ordered list of the returned codes for convenience.
Human-readable text
In the task feed the tool renders a## Diagnosis ICD-10 table (code, description, confidence, reasoning), plus SUD codes and notes when present. The full result struct stays in structuredContent.
SUD codes (follow-up)
sud_diagnosis_codes is always an empty array for now. The baked Fleming diagnosis prompt returns only icd10_codes and notes; dedicated substance-use-disorder (SUD) extraction is a documented follow-up and would require expanding the prompt.
Notes
- LLM tool. Billed per call.
default_enabledis false.