Wave 2 · Phase A · Knowledge library

AKHANDA · KB Manifest working name

84 knowledge bases · v1.0 draft · pending clinical review

Knowledge bases

Layers

6 · L0–L5

Grade distribution

64 established · 16 emerging · 4 contested · 0 frontier-hypothesis

01 · Conventions

How to read this manifest

KB: Every knowledge base is identified as AKHANDA_KB<nn>, zero-padded. An id is never reused after retirement.
Grade: The KB's claim_grade — established, emerging, contested, or frontier-hypothesis. The grade governs the hedging language every downstream product is licensed to use when it cites the KB.
Deps: The KBs that must be sealed before this one is authored. A KB is authored only on top of finished dependencies.
evidence_basis: Roughly half the library is definitional — ontology, workflow, governance — rather than empirical: there is no evidence base behind an edge-semantics dictionary, only a design decision. Definitional KBs therefore carry established with evidence_basis: definitional in frontmatter, so a design decision is never mistaken for a clinically evidenced claim.

02 · The library

The 84 KBs, layer by layer

Ontology

22 KBs

KB	Title	Grade	Deps	Scope
KB01	Patient Identity & Linkage Rules	established	—	ABHA/MRN/national ID resolution; deterministic and probabilistic matching thresholds; what constitutes one person
KB02	Problem Ontology & SNOMED CT Value Sets	established	—	What a clinical problem is as an entity; the bounded SNOMED subset AKHANDA recognises
KB03	Observation Ontology & LOINC Mapping	established	KB06	What a measured fact is; LOINC binding for every observation type accepted
KB04	Medication Ontology	established	KB06	Ingredient/form/strength/route model; RxNorm, Indian brand register, AYUSH formulations
KB05	Procedure Ontology	established	—	SNOMED/CPT/ICD-10-PCS crosswalk; what counts as a performed procedure
KB06	Unit Normalisation & Conversion Canon	established	—	Canonical unit per measurement class; conversion factors; precision-loss rules
KB07	Reference Range Library	established	KB06, KB29	Normal ranges banded by age, sex, pregnancy, ethnicity — what is abnormal
KB08	Edge Semantics Dictionary	established	KB02, KB03	Meaning and directionality of `caused_by`, `treats`, `contradicts`, `supersedes`, `evidences`, `refutes`, `monitors`
KB09	Belief Statement Type Schemas	established	KB08	Typed payload schema per `statement_type`: risk_score, differential, trajectory, contraindication
KB10	Provenance & Attribution Model	established	KB09	What `basis` must contain for a write to be accepted; agent identity as product@version
KB11	Bitemporal Semantics & Query Canon	established	—	Valid time vs transaction time; the "what did we know at T" contract
KB12	Confidence Calibration Scales	emerging	—	What a confidence value means per statement type; cross-product comparability of the 0–1 scale
KB13	Claim Grade Definitions & Hedging Rules	established	—	The four grades and the downstream language each licenses
KB14	Problem Lifecycle States	established	KB02	suspected → active → controlled → resolved → recurrent; legal transitions
KB15	Severity & Acuity Scales Canon	established	—	Which severity scale is canonical per problem class; cross-scale mapping
KB16	Anatomical & Laterality Model	established	—	Body site representation; laterality as first-class, never free text
KB17	Allergy & Intolerance Ontology	established	KB04	Allergy vs intolerance vs adverse reaction; substance vs class; reaction severity
KB18	Social & Environmental Determinant Vocabulary	emerging	—	Income, housing, caste, occupation, water access, indoor air — controlled vocabulary, India/Oman applicable
KB19	Genomic Variant Representation	established	—	HGVS notation, star alleles, diplotype representation, phenotype call
KB20	Device & Implant Registry Model	established	KB05	Implanted and worn devices as persistent state, incl. MRI safety attributes
KB21	Encounter Context Taxonomy	established	—	OPD, IPD, ICU, ED, home, tele — context as a first-class modifier of interpretation
KB22	Identity Merge & Unmerge Safety Rules	established	KB01	When a merge may be executed, how it is reversed, what survives an unmerge

Domain Knowledge

14 KBs

KB	Title	Grade	Deps	Scope
KB23	Physiological Plausibility Bounds	established	KB06, KB07	Values that are impossible, not merely abnormal — hard rejection at write
KB24	Observation Volatility Profiles	emerging	KB03	How fast each measure can legitimately change; implausible-delta detection
KB25	Problem Co-occurrence Priors	emerging	KB02	Base rates of comorbidity clustering, India/Oman weighted
KB26	Causal Adjacency Canon	contested	KB02, KB08	Which problems can cause which — the permitted `caused_by` edge set
KB27	Medication–Problem Indication Map	established	KB02, KB04	Which drug treats which problem; supports `treats` edge validation
KB28	Lab Result Interpretation Context Rules	established	KB03, KB07	How encounter context, timing and concurrent therapy change interpretation
KB29	Life-Stage Physiological Variance	established	—	Neonate → geriatric normal physiology; the substrate KB07 bands against
KB30	Pregnancy State Model & Trimester Adjustments	established	KB29	Pregnancy as a state that rewrites reference ranges and safety gates
KB31	Renal & Hepatic Function Staging Canon	established	KB07	eGFR equations, CKD staging, Child-Pugh, MELD — the organ-function truth other products consume
KB32	Frailty & Functional Status Models	emerging	KB29	Clinical Frailty Scale, ADL/IADL; frailty as a modifier of every threshold
KB33	Symptom–Sign Semantic Equivalence	emerging	KB02	Patient-reported symptom ↔ clinician-recorded sign mapping
KB34	Indian Epidemiological Priors	emerging	KB25	Prevalence weighting by region, urban/rural, season
KB35	Oman & GCC Epidemiological Priors	emerging	KB25	As above for the Oman deployment
KB36	Traditional Medicine Concept Bridging	contested	KB04	Ayurveda/Siddha/Unani concepts → biomedical entities where a defensible mapping exists, and explicit refusal where it does not

Reasoning Templates

16 KBs

KB	Title	Grade	Deps	Scope
KB37	Belief Conflict Detection Rules	established	KB09, KB08	How two beliefs are determined to conflict — the detection predicate
KB38	Belief Supersession Precedence	contested	KB37	Recency, specificity, agent authority — which belief wins, and when nothing wins
KB39	Observation Deduplication & Reconciliation	established	KB03, KB24	Same measurement arriving twice; near-duplicate resolution
KB40	Problem List Curation Heuristics	emerging	KB02, KB14	Keeping the list clinically useful rather than exhaustive
KB41	Active vs Historical Problem Discrimination	emerging	KB14, KB40	When a problem stops being current without being resolved
KB42	Medication Reconciliation Source Precedence	established	KB04, KB49	EMR vs pharmacy vs patient report vs dispenser — whose list wins
KB43	Trajectory Interpolation Rules	emerging	KB24	Legitimate inference between observations; where interpolation is forbidden
KB44	Missing Data Semantics	established	—	Absent ≠ normal. The doctrine and its enforcement predicate
KB45	Stale Data Decay Functions	emerging	KB24, KB44	How confidence in a fact degrades with age, per measure class
KB46	Cross-Product Score Comparability Rules	contested	KB12	Whether two products' risk scores may be compared, and when they may not
KB47	Contradiction Surfacing Thresholds	established	KB37	When a detected conflict is significant enough to raise to a human
KB48	Graph Query Patterns Library	established	KB08, KB11	Canonical traversals — causal chain, medication history, belief lineage
KB49	State Snapshot Composition Rules	established	KB11, KB45	How a point-in-time snapshot is assembled and what it may omit
KB50	Derived Attribute Computation Canon	established	KB31, KB06	Deterministic derived values (eGFR, BMI, BSA, corrected calcium) — code, never model
KB51	Confidence Propagation Through Inference Chains	emerging	KB12, KB50	How confidence degrades across a chain of derived beliefs
KB52	Anomaly & Implausible-State Detection	emerging	KB23, KB25	Individually plausible facts that are jointly impossible

Workflow

10 KBs

KB	Title	Grade	Deps	Scope
KB53	Write Path State Machine	established	KB10, KB75	Validate → authorise → plausibility → persist → project → emit
KB54	Clinician Arbitration Workflow	established	KB47	How a surfaced contradiction reaches a human and is resolved with recorded reason
KB55	Problem List Review & Attestation Cycle	established	KB40	Periodic human attestation; what an unattested list means downstream
KB56	Identity Merge Review Workflow	established	KB22	Human review gate before any merge executes
KB57	Retraction & Correction Workflow	established	KB53	Correcting the record without deleting history
KB58	Bulk Import Reconciliation Workflow	established	KB39, KB42	Legacy/EMR bulk ingest without corrupting curated state
KB59	Cross-Facility Transfer of State	established	KB49, KB64	Handing a patient's state to another facility
KB60	Subscription & Notification Rules	established	KB74	Which state changes notify which subscribers
KB61	Snapshot Request & Delivery Workflow	established	KB49	Synchronous and async snapshot delivery, incl. lens selection
KB62	Degraded Mode Operating Procedure	established	KB53	Behaviour per blueprint §9; Postgres-down is a hard stop

Integration

12 KBs

KB	Title	Grade	Deps	Scope
KB63	FHIR R4 Resource Mapping	established	KB02–KB05	Bidirectional AKHANDA ↔ FHIR; what has no FHIR equivalent (Belief) and how it is carried
KB64	ABDM / ABHA Health Record Contract	established	KB01, KB63	National health stack compliance
KB65	AAROGYA EMR Sync Contract	established	KB63	Facts in, interpretation out; AAROGYA is the record, AKHANDA is the reasoning
KB66	OMNIOME Read/Write Contract	established	KB09, KB49	The flagship's belief-write surface
KB67	Cardiometabolic Twin State Exchange	established	KB09, KB43	INDRA H-OS trajectory beliefs in and out
KB68	Specialty Clinician Write Contracts	established	KB66	ARJUN, HRIDAYA, NEEL, DRISHTI family, SOBHA, KSHEMA
KB69	GUARDIAN+ Medication State Contract	established	KB04, KB42	The authoritative medication list handoff
KB70	PRAMANA Evidence Emission Contract	established	KB74	Every belief write that constitutes a recommendation
KB71	JEEVANA Goal & Plan Contract	established	KB09	Goal nodes owned by JEEVANA, held in AKHANDA
KB72	Wearable Stream Ingestion	emerging	KB24, KB45	Aegis Bio-ID / ARES Sentinel high-frequency signal; downsampling and retention
KB73	SANGAMA Federated Export Schema	established	KB79, KB93*	Aggregate-only export shape; never record-shaped
KB74	Event Catalogue & Versioning Policy	established	—	Every event AKHANDA emits; deprecation policy

* KB93 is a forward reference to SANGAMA's own KB set — a cross-product dependency, resolved at Phase D.

Safety & Governance

10 KBs

KB	Title	Grade	Deps	Scope
KB75	Write Authorisation Matrix	established	KB10	Which agent may assert which statement type about which node family
KB76	Consent Model & Purpose Binding	established	—	Four purposes: care, QI, research, commercial RWE — separate and independent
KB77	Consent Revocation Cascade	established	KB76, KB60	Propagation to every downstream subscriber within SLA
KB78	Audit Log Specification & Immutability	established	KB53	Hash-chaining, daily anchor, tamper evidence
KB79	PHI Minimisation & Field-Level Access Control	established	KB75	Least-privilege projection; lens-level redaction
KB80	Break-Glass Emergency Access Protocol	established	KB79, KB78	Override with immutable logging and 5-minute notification
KB81	Data Residency & Cross-Border Rules	established	KB76	India DPDP, Oman; what may never leave jurisdiction
KB82	Retention & Deletion Policy	established	KB81	Reconciling append-only architecture with erasure rights
KB83	Belief Poisoning & Adversarial Write Defence	emerging	KB75, KB37	A compromised product writing plausible-but-false beliefs
KB84	Medico-Legal Reconstruction Requirements	established	KB11, KB78	Proving what was known and shown at the moment of a decision

03 · Contested

The four contested KBs

Four KBs carry the grade contested. Each encodes a judgement the field does not agree on, and each therefore licenses only the most hedged downstream language until clinical review upgrades or retires it.

KB26

Causal Adjacency Canon

A permitted-causation set — which problems can cause which — is a strong clinical claim. As a design requirement it is reviewed by two independent clinical reviewers rather than one.

dual clinical review

KB36

Traditional Medicine Concept Bridging

Ayurveda/Siddha/Unani concepts map to biomedical entities only where a defensible mapping exists. As a design requirement the KB carries a register of concepts that will not be mapped.

explicit refusal register

KB38

Belief Supersession Precedence

Recency, specificity, agent authority — which belief wins, and when nothing wins. The field has no agreed precedence order; the KB encodes one and says so.

KB46

Cross-Product Score Comparability Rules

Whether two products' risk scores may be compared at all, and when they may not. Treating scales as interchangeable is a judgement, not a fact.

04 · Boundaries

Boundary analysis

Adjacent KBs are checked for scope overlap before authoring begins, so that two authors never encode the same decision twice — or half a decision each. Five findings.

KB37 ↔ KB47 under review

Conflict detection and surfacing thresholds are one decision

Belief Conflict Detection Rules and Contradiction Surfacing Thresholds are one decision split across two files. Detection without a threshold produces nothing actionable; a threshold without detection has no input. Authoring them separately guarantees they drift. A merge into a single KB is proposed; the canonical count remains 84 until that review concludes, and a retired id is never reused.

KB07 ↔ KB29

Substrate vs application — keep separate, boundary explicit

Both encode age/sex-banded physiology. The boundary: KB29 is what is normal for this life stage (substrate); KB07 is the assay-specific reference interval a lab reports against (application). KB07 cites KB29 rather than restating it — without that statement in both files, an author will duplicate.

KB23 ↔ KB07

Impossible vs abnormal — keep separate, high conflation risk

Plausibility bounds (impossible — reject the write) vs reference ranges (abnormal — accept and flag). Conflating these causes a genuinely dangerous failure: rejecting a real critical value as implausible. Both files open with the distinction stated.

KB39 ↔ KB42

Same mechanism, different domains — extract the shared primitive

Observation dedup and medication source precedence are different clinical domains running the same mechanism: the same fact from multiple sources, ranked by trustworthiness. The precedence primitive is authored once, inside KB49 (State Snapshot Composition Rules), and both KBs cite it — rather than merging two clinically distinct KBs.

KB10 ↔ KB78

Provenance vs audit — keep separate, boundary explicit

Provenance answers what supports this claim; audit answers who did what, when. They share fields and will be conflated unless the split is stated in both files.

Also noted, below threshold

KB44 ↔ KB45 Missing vs stale. Distinct — absent data and old data fail differently. Keep separate.
KB01 ↔ KB22 Identity model vs merge safety. Correctly split across L0/L5.
KB12 ↔ KB51 Scale definition vs propagation. Distinct.

This manifest is the knowledge layer of AKHANDA. The blueprint covers the graph model, the data architecture, the engine, and how it fails safely.

← AKHANDA blueprint