Precision Autoimmune Intelligence Agent -- Architecture Design Document¶

Author: Adam Jones Date: March 2026 Version: 1.0.0 License: Apache 2.0

1. Executive Summary¶

The Precision Autoimmune Intelligence Agent extends the HCLS AI Factory platform to deliver comprehensive autoimmune disease intelligence. Unlike point-of-care tools that address a single disease or scoring system, this agent integrates autoantibody interpretation, HLA-disease associations, disease activity scoring, flare prediction, biologic therapy selection with pharmacogenomic considerations, and diagnostic odyssey analysis across 13 autoimmune diseases -- all grounded in a 14-collection RAG pipeline.

The agent combines 6 deterministic clinical analysis engines with a 14-collection RAG pipeline to answer questions like "Interpret ANA 1:640 homogeneous with positive anti-dsDNA and falling complement" -- simultaneously searching autoantibody reference data, HLA associations, disease activity thresholds, flare patterns, biologic therapy databases, and clinical evidence, then synthesizing a grounded response through Claude.

Key Results¶

Metric	Value
Unit tests passing	455
Milvus collections	14 (13 owned + 1 read-only)
Autoimmune diseases	13 in AutoimmuneDisease enum
Biologic therapies	22 with PGx considerations
Autoantibodies mapped	24 to disease associations with sensitivity/specificity
HLA alleles	22 with disease odds ratios and PMIDs
Disease activity scores	20 scoring systems across 13 diseases
Demo patients	9 with full clinical PDF records
Flare patterns	13 disease-specific biomarker patterns
Classification criteria	10 ACR/EULAR criteria sets
Overlap syndromes	9 cross-disease patterns
Knowledge version	v2.0.0

2. Architecture Overview¶

2.1 Mapping to VAST AI OS¶

VAST AI OS Component	Autoimmune Agent Role
DataStore	Clinical PDFs, reference JSON files, knowledge base (HLA, antibodies, biologics, flare patterns)
DataEngine	Document processor: PDF -> chunks -> BGE-small embedding -> Milvus insert
DataBase	14 Milvus collections (13 owned + 1 read-only) + 9 demo patients
InsightEngine	6 clinical engines + BGE-small embedding + multi-collection RAG
AgentEngine	AutoimmuneAgent orchestrator + Streamlit UI (10 tabs) + FastAPI REST

2.2 System Diagram¶

                        ┌─────────────────────────────────┐
                        │    Streamlit UI (8531)            │
                        │    10 tabs: Query | Analysis |    │
                        │    Ingest | Odyssey | Antibody |  │
                        │    HLA | Activity | Flare |       │
                        │    Therapy | Knowledge             │
                        └──────────────┬──────────────────┘
                                       │
                        ┌──────────────▼──────────────────┐
                        │  AutoimmuneAgent                  │
                        │  Orchestrates 6 analysis engines  │
                        │  + RAG pipeline + export           │
                        └──────────────┬──────────────────┘
                                       │
            ┌──────────────────────────┼───────────────────────────┐
            │                          │                           │
  ┌─────────▼──────────┐   ┌──────────▼──────────┐   ┌──────────▼──────────┐
  │ Deterministic       │   │ RAG Pipeline         │   │ Export               │
  │ Analysis Engines    │   │                      │   │                      │
  │                     │   │ BGE-small-en-v1.5    │   │ FHIR R4 Bundle       │
  │ AutoantibodyInterp  │   │ (384-dim embedding)  │   │ PDF (reportlab)      │
  │ HLAAssociation      │   │         │            │   │ Markdown             │
  │ DiseaseActivity     │   │         ▼            │   │                      │
  │ FlarePredictor      │   │ Parallel Search      │   │ + FHIR Validation    │
  │ BiologicTherapy     │   │ 14 Milvus Collections│   │                      │
  │ DiagnosticOdyssey   │   │ (ThreadPoolExecutor) │   │                      │
  │                     │   │         │            │   │                      │
  │ + DocumentProcessor │   │         ▼            │   │                      │
  │ + TimelineBuilder   │   │ Claude Sonnet 4      │   │                      │
  └─────────────────────┘   └──────────────────────┘   └──────────────────────┘
            │                          │
  ┌─────────▼──────────────────────────▼──────────────────────────────┐
  │                  Milvus 2.4 -- 14 Collections                      │
  │                                                                    │
  │  autoimmune_clinical_documents       autoimmune_patient_labs       │
  │  autoimmune_autoantibody_panels      autoimmune_hla_associations   │
  │  autoimmune_disease_criteria         autoimmune_disease_activity   │
  │  autoimmune_flare_patterns           autoimmune_biologic_therapies │
  │  autoimmune_pgx_rules               autoimmune_clinical_trials     │
  │  autoimmune_literature               autoimmune_patient_timelines  │
  │  autoimmune_cross_disease            genomic_evidence [read-only]  │
  └───────────────────────────────────────────────────────────────────┘

3. Data Collections -- Actual State¶

3.1 `autoimmune_clinical_documents`¶

Ingested clinical documents (PDFs) chunked for vector search.

Field	Type	Description
id	VARCHAR(128)	Primary key
embedding	FLOAT_VECTOR(384)	BGE-small-en-v1.5
text_chunk	VARCHAR(3000)	Document text chunk
patient_id	VARCHAR(64)	Patient identifier
doc_type	VARCHAR(128)	progress_note, lab_report, imaging, pathology, etc.
specialty	VARCHAR(128)	rheumatology, neurology, nephrology, etc.
provider	VARCHAR(256)	Provider name
visit_date	VARCHAR(32)	ISO date
source_file	VARCHAR(512)	Original PDF filename
page_number	INT64	Page number in source
chunk_index	INT64	Chunk index within document

3.2 `autoimmune_patient_labs`¶

Laboratory results with reference range analysis.

Field	Type	Description
id	VARCHAR(128)	Primary key
embedding	FLOAT_VECTOR(384)	BGE-small-en-v1.5
text_chunk	VARCHAR(3000)	Lab result description
patient_id	VARCHAR(64)	Patient identifier
test_name	VARCHAR(256)	Lab test name
value	FLOAT	Measured value
unit	VARCHAR(64)	Unit of measurement
reference_range	VARCHAR(128)	Normal reference range
flag	VARCHAR(32)	normal, high, low, critical
collection_date	VARCHAR(32)	Date of collection
panel_name	VARCHAR(256)	Lab panel name

3.3 `autoimmune_autoantibody_panels` -- 24 records¶

Autoantibody reference data with disease associations.

Field	Type	Description
id	VARCHAR(128)	Primary key
embedding	FLOAT_VECTOR(384)	BGE-small-en-v1.5
text_chunk	VARCHAR(3000)	Autoantibody description
antibody_name	VARCHAR(128)	Antibody name (ANA, anti-dsDNA, RF, etc.)
associated_diseases	VARCHAR(1024)	Associated autoimmune diseases
sensitivity	FLOAT	Diagnostic sensitivity (0-1)
specificity	FLOAT	Diagnostic specificity (0-1)
pattern	VARCHAR(128)	Staining pattern (homogeneous, speckled, etc.)
clinical_significance	VARCHAR(2000)	Clinical interpretation
interpretation_guide	VARCHAR(2000)	Interpretation guidelines

3.4 `autoimmune_hla_associations` -- 22 records¶

HLA allele to disease risk mapping.

Field	Type	Description
id	VARCHAR(128)	Primary key
embedding	FLOAT_VECTOR(384)	BGE-small-en-v1.5
text_chunk	VARCHAR(3000)	Association description
allele	VARCHAR(64)	HLA allele (e.g., HLA-B*27:05)
disease	VARCHAR(256)	Associated disease
odds_ratio	FLOAT	Disease odds ratio
population	VARCHAR(128)	Population studied
pmid	VARCHAR(32)	PubMed ID
mechanism	VARCHAR(1024)	Pathogenic mechanism
clinical_implication	VARCHAR(2000)	Clinical implications

3.5 `autoimmune_disease_criteria` -- 10 records¶

ACR/EULAR classification criteria for autoimmune diseases.

Field	Type	Description
id	VARCHAR(128)	Primary key
embedding	FLOAT_VECTOR(384)	BGE-small-en-v1.5
text_chunk	VARCHAR(3000)	Criteria description
disease	VARCHAR(256)	Disease name
criteria_set	VARCHAR(256)	Criteria set name (e.g., 2010 ACR/EULAR RA)
criteria_type	VARCHAR(64)	classification or diagnostic
year	INT64	Publication year
required_score	VARCHAR(128)	Score threshold
criteria_items	VARCHAR(3000)	Individual criteria items
sensitivity_specificity	VARCHAR(256)	Validation metrics

3.6 `autoimmune_disease_activity` -- 20 records¶

Disease activity scoring systems with thresholds and components.

Field	Type	Description
id	VARCHAR(128)	Primary key
embedding	FLOAT_VECTOR(384)	BGE-small-en-v1.5
text_chunk	VARCHAR(3000)	Scoring system description
score_name	VARCHAR(128)	Score name (DAS28-CRP, SLEDAI-2K, etc.)
disease	VARCHAR(256)	Applicable disease
components	VARCHAR(2000)	Component list
thresholds	VARCHAR(1024)	JSON thresholds for remission/low/moderate/high
interpretation	VARCHAR(2000)	Interpretation guide
monitoring_frequency	VARCHAR(512)	Recommended monitoring schedule

3.7 `autoimmune_flare_patterns` -- 13 records¶

Flare prediction biomarker patterns for each disease.

Field	Type	Description
id	VARCHAR(128)	Primary key
embedding	FLOAT_VECTOR(384)	BGE-small-en-v1.5
text_chunk	VARCHAR(3000)	Pattern description
disease	VARCHAR(256)	Disease name
biomarker_pattern	VARCHAR(2000)	Early warning biomarker list
early_warning_signs	VARCHAR(2000)	Early warning descriptions
typical_timeline	VARCHAR(512)	Typical flare timeline
protective_factors	VARCHAR(1024)	Factors reducing flare risk
intervention_triggers	VARCHAR(1024)	When to escalate treatment

3.8 `autoimmune_biologic_therapies` -- 22 records¶

Biologic therapy database with PGx considerations.

Field	Type	Description
id	VARCHAR(128)	Primary key
embedding	FLOAT_VECTOR(384)	BGE-small-en-v1.5
text_chunk	VARCHAR(3000)	Therapy description
drug_name	VARCHAR(128)	Drug name
drug_class	VARCHAR(128)	Drug class (TNF inhibitor, IL-6R inhibitor, etc.)
mechanism	VARCHAR(512)	Mechanism of action
indicated_diseases	VARCHAR(1024)	Approved indications
pgx_considerations	VARCHAR(2000)	Pharmacogenomic factors
contraindications	VARCHAR(1024)	Contraindications
monitoring	VARCHAR(2000)	Monitoring requirements
dosing	VARCHAR(512)	Dosing information
evidence_level	VARCHAR(64)	Evidence level

3.9 `autoimmune_pgx_rules`¶

Pharmacogenomic dosing rules for autoimmune therapies.

3.10 `autoimmune_clinical_trials`¶

Autoimmune disease clinical trials with biomarker-based eligibility criteria.

3.11 `autoimmune_literature`¶

Published autoimmune research literature with abstracts and disease focus.

3.12 `autoimmune_patient_timelines`¶

Patient diagnostic timeline events for odyssey analysis.

Field	Type	Description
id	VARCHAR(128)	Primary key
embedding	FLOAT_VECTOR(384)	BGE-small-en-v1.5
text_chunk	VARCHAR(3000)	Event description
patient_id	VARCHAR(64)	Patient identifier
event_type	VARCHAR(128)	symptom_onset, diagnosis, treatment_start, flare, etc.
event_date	VARCHAR(32)	ISO date
description	VARCHAR(2000)	Event description
provider	VARCHAR(256)	Provider name
specialty	VARCHAR(128)	Medical specialty
days_from_first_symptom	INT64	Days since first symptom

3.13 `autoimmune_cross_disease` -- 9 records¶

Cross-disease overlap syndromes and shared pathogenic mechanisms.

Field	Type	Description
id	VARCHAR(128)	Primary key
embedding	FLOAT_VECTOR(384)	BGE-small-en-v1.5
text_chunk	VARCHAR(3000)	Overlap description
primary_disease	VARCHAR(256)	Primary disease
associated_conditions	VARCHAR(1024)	Associated conditions
shared_pathways	VARCHAR(1024)	Shared pathogenic mechanisms
shared_biomarkers	VARCHAR(1024)	Shared biomarkers
overlap_criteria	VARCHAR(2000)	Overlap diagnostic criteria
co_occurrence_rate	FLOAT	Co-occurrence rate

3.14 Index Configuration (all collections)¶

Algorithm:  IVF_FLAT
Metric:     COSINE
nlist:      1024
nprobe:     16
Dimension:  384 (BGE-small-en-v1.5)

4. Clinical Analysis Engines¶

4.1 AutoantibodyInterpreter¶

Maps 24 autoantibodies to disease associations with sensitivity and specificity data.

24 Supported Autoantibodies:

Autoantibody	Primary Disease Association	Sensitivity	Specificity
ANA	SLE, Sjogren's, SSc	0.95 (SLE)	0.65
anti-dsDNA	SLE	0.70	0.95
anti-Smith	SLE	0.25	0.99
RF	RA, Sjogren's	0.70 (RA)	0.85
anti-CCP	RA	0.67	0.95
anti-Scl-70	SSc (diffuse)	0.35	0.99
anti-centromere	SSc (limited/CREST)	0.40	0.98
anti-SSA/Ro	Sjogren's, SLE	0.70 (SS)	0.90
anti-SSB/La	Sjogren's	0.40	0.95
anti-Jo-1	Antisynthetase syndrome	0.30	0.99
AChR antibody	Myasthenia Gravis	0.85	0.99
anti-tTG IgA	Celiac Disease	0.93	0.97
TSI	Graves' Disease	0.90	0.95
anti-TPO	Hashimoto's, Graves'	0.90 (Hashi)	0.85
anti-RNP	MCTD, SLE	0.95 (MCTD)	0.85
anti-histone	Drug-induced lupus, SLE	0.95 (DIL)	0.50
ANCA (c-ANCA/PR3)	GPA (Wegener's)	0.90	0.95
ANCA (p-ANCA/MPO)	MPA, EGPA	0.70	0.90
anti-Pm-Scl	SSc-myositis overlap	0.10	0.98
anti-RNA Pol III	SSc (diffuse, renal crisis)	0.20	0.99
anti-cardiolipin	APS	0.80	0.80
lupus anticoagulant	APS	0.55	0.95
anti-beta2-GP I	APS	0.70	0.90
anti-MuSK	MG (AChR-negative)	0.40	0.99

Interpretation logic: For each positive antibody in the patient's panel, the engine returns all disease associations with their sensitivity, specificity, titer, and staining pattern.

4.2 HLAAssociationAnalyzer¶

Matches 22 HLA alleles against a curated disease association database with odds ratios and PubMed references.

22 HLA Alleles:

Allele	Disease	Odds Ratio	PMID
HLA-B*27:05	Ankylosing Spondylitis	87.4	25603694
HLA-B*27:02	Ankylosing Spondylitis	50.0	25603694
HLA-C*06:02	Psoriasis	10.0	23143594
HLA-DQB1*02:01	Celiac Disease	7.0	17554300
HLA-DQB1*03:02	Type 1 Diabetes	6.5	17554300
HLA-B*51:01	Behcet's Disease	5.9	22704706
HLA-DRB1*04:01	Rheumatoid Arthritis	4.2	20301572
HLA-DRB1*04:04	Rheumatoid Arthritis	3.8	20301572
HLA-DRB1*03:01	T1D, SLE, Sjogren's, Graves', Celiac	2.2-7.0	17554300
HLA-DRB1*15:01	Multiple Sclerosis	3.1	21833088
HLA-B*08:01	Myasthenia Gravis, SLE	3.4 (MG)	16710306
HLA-DRB1*04:05	Rheumatoid Arthritis	3.5	20301572
HLA-DRB1*01:01	RA, SSc	2.0-2.1	20301572
HLA-DRB1*08:01	SLE	2.1	19864127
HLA-DQA1*05:01	Celiac Disease	7.0	17554300
HLA-DRB1*15:03	MS (African-descent)	2.8	21833088
HLA-A*02:01	Type 1 Diabetes	1.5	17554300
HLA-DRB1*07:01	T1D (PROTECTIVE)	0.3	17554300
HLA-DPB1*05:01	Systemic Sclerosis	2.3	24098041
HLA-B*44:03	Type 1 Diabetes	1.4	17554300
HLA-DRB1*11:01	Sjogren's Syndrome	2.5	19864127
HLA-DRB1*13:01	T1D (PROTECTIVE)	0.2	17554300

Matching logic: Supports exact allele match and broad allele group matching (e.g., B27:05 matches B27). Results sorted by odds ratio (highest risk first). Protective alleles (OR < 1.0) are identified.

4.3 DiseaseActivityScorer¶

Calculates disease activity levels using simplified CRP/ESR-based scoring against 20 validated scoring systems.

20 Scoring Systems:

Score	Disease	Range	Remission	Low	Moderate	High	Reference
DAS28-CRP	RA	0-10	<2.6	3.2	5.1	>5.1	PMID:15593215
DAS28-ESR	RA	0-10	<2.6	3.2	5.1	>5.1	PMID:15593215
SLEDAI-2K	SLE	0-105	0	4	8	>=12	PMID:12115176
CDAI	RA	0-76	<2.8	10	22	>22	PMID:15641075
BASDAI	AS	0-10	<2	3	4	>4	PMID:8003055
SDAI	RA	0-86	<3.3	11	26	>26	PMID:14872836
PASI	Psoriasis	0-72	<1	5	10	>10	PMID:15888150
Mayo Score	IBD	0-12	<2	5	8	>8	PMID:3317057
Harvey-Bradshaw	IBD	0-30	<4	7	16	>16	PMID:7014041
ESSDAI	Sjogren's	0-123	<1	5	14	>14	PMID:20032223
mRSS	SSc	0-51	<5	14	29	>29	PMID:8546527
EDSS	MS	0-10	<1.5	3.5	6.0	>6.0	PMID:6685237
QMGS	MG	0-39	<3	10	20	>20	PMID:10668691
Marsh Score	Celiac	0-4	0	1	2	>=3	PMID:1437871
Burch-Wartofsky	Graves'	0-140	<10	25	45	>45	PMID:8432869
ASDAS	AS	0-6	<1.3	2.1	3.5	>3.5	PMID:19139421
MG-ADL	MG	0-24	<1	5	10	>10	PMID:10025780
DAPSA	Psoriasis	0-164	<4	14	28	>28	PMID:22328740
HbA1c-T1D	T1D	4-14%	<6.5	7.0	8.5	>8.5	PMID:9742976
TSH-Hashimoto	Hashimoto's	0-100	<2.5	5.0	10.0	>10.0	PMID:12487769

Scoring logic: Retrieves CRP and/or ESR from patient biomarkers, maps to the applicable scoring system for diagnosed conditions, and classifies activity level against thresholds. Returns level (REMISSION/LOW/MODERATE/HIGH/VERY_HIGH), components, and threshold context.

4.4 FlarePredictor¶

Predicts flare risk from biomarker patterns using 13 disease-specific configurations.

Algorithm: 1. Start with base risk score of 0.3 2. For each early warning biomarker present in patient data: - Elevated inflammatory markers (CRP, ESR, IL-6, calprotectin > 5) add +0.15 - Low complement (C3, C4 < 80) adds +0.15 - Low albumin (< 3.5) adds +0.10 - Normal values become protective factors 3. Clamp score to [0.0, 1.0] 4. Classify: IMMINENT (>=0.8), HIGH (>=0.6), MODERATE (>=0.4), LOW (<0.4) 5. Generate recommended monitoring actions

13 Disease Patterns: RA, SLE, IBD, AS, Psoriasis, Sjogren's, SSc, MS, T1D, MG, Celiac, Graves', Hashimoto's. Each with disease-specific early warning biomarkers and protective signals.

4.5 BiologicTherapyAdvisor¶

Matches 22 biologic therapies to patient diagnoses with PGx filtering.

22 Therapies by Class:

Class	Therapies	Key PGx Considerations
TNF inhibitors (5)	Adalimumab, Etanercept, Infliximab, Golimumab, Certolizumab	FCGR3A V158F, HLA-DRB1*03:01 (anti-drug antibodies), TNFA -308
IL-6R inhibitors (2)	Tocilizumab, Sarilumab	IL6R Asp358Ala, CRP masked by IL-6R blockade
Anti-CD20 (2)	Rituximab, Ocrelizumab	FCGR3A V158F, TNFSF13B levels, hepatitis B risk
IL-17A inhibitors (2)	Secukinumab, Ixekizumab	HLA-C*06:02 response prediction, IBD worsening risk
IL-12/23 inhibitors (2)	Ustekinumab, Risankizumab	IL12B/IL23R variants
IL-23 p19 inhibitor (1)	Guselkumab	IL23R rs11209026, HLA-C*06:02
BLyS inhibitor (1)	Belimumab	TNFSF13B genotype
T-cell modulator (1)	Abatacept	CTLA4 +49 A/G, shared epitope response
JAK inhibitors (3)	Tofacitinib, Baricitinib, Upadacitinib	CYP3A4/CYP2C19 metabolism, VTE risk
Integrin inhibitors (2)	Vedolizumab, Natalizumab	JC virus antibody index (PML risk), anti-drug antibodies
TYK2 inhibitor (1)	Deucravacitinib	TYK2 P1104A variant, minimal CYP interaction

Selection logic: Filters therapies by indicated_diseases matching patient diagnoses, returns all matching therapies with PGx considerations, contraindications, and monitoring requirements.

4.6 DiagnosticOdysseyAnalyzer¶

Multi-function diagnostic reasoning engine implementing:

Classification Criteria Evaluation: - 10 ACR/EULAR criteria sets (RA, SLE, AS, SSc, Sjogren's, MS, MG, Celiac, IBD, Psoriasis) - Point-based scoring against disease-specific thresholds - Returns met/unmet criteria, total points, and threshold comparison

Overlap Syndrome Detection: - 9 overlap syndromes (MCTD, SLE-RA overlap, Sjogren's-SLE, POTS/hEDS/MCAS, SSc-myositis, T1D-celiac, thyroid-T1D, RA-Sjogren's, lupus-APS) - Detects based on antibody profile and diagnosed conditions - Reports matched markers, involved diseases, and confidence level

Differential Diagnosis: - Scores diseases from positive antibodies (weighted by specificity) and HLA alleles (log2-scaled odds ratio) - Returns ranked differential with evidence for each disease

Diagnostic Odyssey Analysis: - Reconstructs timeline from event records - Calculates diagnostic delay (days/months/years from first symptom to diagnosis) - Counts specialists seen, misdiagnoses, and key diagnostic tests - Identifies turning points and missed opportunities

5. Multi-Collection RAG Engine¶

5.1 Search Flow¶

Query Text
    │
    ▼
BGE-small-en-v1.5 Embedding (384-dim, asymmetric with query prefix)
    │
    ▼
ThreadPoolExecutor: Parallel search across 14 collections (max_workers=6)
    │
    ▼
Weighted merge + deduplication (content hash + ID dedup)
    │
    ▼
Knowledge base augmentation (HLA, autoantibody, therapy, flare pattern context)
    │
    ▼
Disease area detection (keyword-based routing)
    │
    ▼
Conversation history injection (3-turn memory)
    │
    ▼
Claude Sonnet 4 prompt with patient context + evidence block
    │
    ▼
Grounded response with citations ([AutoAb:name], [HLA:allele], [Therapy:drug], [Literature:PMID])

5.2 Collection Weights¶

Collection	Weight	Rationale
autoimmune_clinical_documents	0.18	Primary patient records
autoimmune_patient_labs	0.14	Lab results with flags
autoimmune_autoantibody_panels	0.12	Autoantibody reference
autoimmune_hla_associations	0.08	HLA-disease mapping
autoimmune_disease_criteria	0.08	Classification criteria
autoimmune_disease_activity	0.07	Activity scoring
autoimmune_flare_patterns	0.06	Flare patterns
autoimmune_biologic_therapies	0.06	Therapy database
autoimmune_clinical_trials	0.05	Clinical trials
autoimmune_literature	0.05	Published research
autoimmune_pgx_rules	0.04	PGx rules
autoimmune_patient_timelines	0.03	Diagnostic timelines
autoimmune_cross_disease	0.02	Overlap syndromes
genomic_evidence	0.02	Shared genomic context
Total	1.00

5.3 Citation Scoring¶

Level	Threshold	Display
High confidence	>= 0.80	Full citation with source link
Medium confidence	>= 0.60	Citation with caveat
Below threshold	< 0.40	Filtered out

6. Export Pipeline¶

6.1 FHIR R4 Bundle¶

Produces a FHIR R4 Bundle containing:

Patient resource with identifier
DiagnosticReport resource (main report)
Observation resources for disease activity scores
Observation resources for flare risk predictions
Conclusion field with critical alerts

6.2 PDF Export¶

Uses reportlab for clinical-grade PDF reports with NVIDIA green branding. Includes critical alerts, disease activity score tables, biologic therapy recommendations with PGx, and clinical query responses.

6.3 Markdown¶

Structured text format with tables for disease activity, flare predictions, HLA associations, and biologic recommendations.

7. Infrastructure¶

7.1 Technology Stack¶

Component	Technology
Language	Python 3.10+
Vector DB	Milvus 2.4
Embeddings	BGE-small-en-v1.5 (BAAI) -- 384-dim
LLM	Claude Sonnet 4 (Anthropic API)
Web UI	Streamlit (10 tabs)
REST API	FastAPI + Uvicorn
Configuration	Pydantic BaseSettings (AUTO_ prefix)
Testing	pytest
PDF Processing	PyPDF2 (ingestion), reportlab (export)
Export	FHIR R4, PDF, Markdown
Containerization	Docker + Docker Compose
Monitoring	Prometheus metrics endpoint

7.2 Service Ports¶

Service	Port
Streamlit UI	8531
FastAPI REST API	8532
Milvus (shared)	19530

7.3 Dependencies on HCLS AI Factory¶

Dependency	Type
Milvus 2.4	Shared vector database (port 19530)
`genomic_evidence` collection	Read-only shared collection from Stage 2 RAG pipeline
BGE-small-en-v1.5	Shared embedding model
Claude API key	Shared Anthropic API key

8. Demo Patients¶

9 Demo Patients¶

#	Name	Age/Sex	Primary Disease	Key Features
1	Sarah Mitchell	34F	SLE	ANA 1:640, anti-dsDNA+, anti-Smith+, lupus nephritis, 27+ PDFs
2	Maya Rodriguez	28F	POTS/hEDS/MCAS	Dysautonomia diagnostic odyssey
3	Linda Chen	45F	Sjogren's	anti-SSA/Ro+, ESSDAI scoring
4	David Park	45M	AS	HLA-B*27+, BASDAI scoring
5	Rachel Thompson	38F	MCTD	Anti-RNP+, overlap syndrome
6	Emma Williams	34F	MS (RRMS)	EDSS scoring, relapse monitoring
7	James Cooper	19M	T1D + Celiac	Overlap syndrome, HLA-DQ2/DQ8+, GAD65+
8	Karen Foster	48F	SSc (dcSSc)	anti-Scl-70+, mRSS scoring
9	Michael Torres	41M	Graves' Disease	TSI+, Burch-Wartofsky scoring

9. File Structure (Actual)¶

precision_autoimmune_agent/
├── src/                            # 9 core modules (~4,500 lines)
│   ├── models.py                   # Pydantic models (AutoimmunePatientProfile, etc.)
│   ├── collections.py              # 14 Milvus collection schemas + manager
│   ├── rag_engine.py               # AutoimmuneRAGEngine (parallel search + Claude)
│   ├── agent.py                    # AutoimmuneAgent orchestrator
│   ├── knowledge.py                # Knowledge v2.0.0 (HLA, antibodies, biologics, flare)
│   ├── diagnostic_engine.py        # Classification criteria, overlap, differential
│   ├── document_processor.py       # PDF ingestion and chunking
│   ├── timeline_builder.py         # Diagnostic odyssey timeline
│   └── export.py                   # FHIR R4 + PDF + Markdown
├── app/
│   └── autoimmune_ui.py            # Streamlit (10 tabs, ~1,100 lines)
├── api/
│   └── main.py                     # FastAPI REST server (14 endpoints, ~580 lines)
├── config/
│   └── settings.py                 # AutoimmuneSettings (AUTO_ prefix)
├── data/                           # Reference data + cache + events
├── demo_data/                      # 9 patient directories with clinical PDFs
├── scripts/                        # setup_collections, generate_demo_patients, pdf_engine
├── tests/                          # 8 test files (455 tests)
├── docker-compose.yml
├── Dockerfile
└── requirements.txt

44 Python files | ~20,000 lines | Apache 2.0

10. Credits¶

Adam Jones
Apache 2.0 License