Clinical Documentation Structuring

from mixpeek import Mixpeek

client = Mixpeek(api_key="YOUR_API_KEY")

# 1. Create namespace for the facility
namespace = client.namespaces.create(name="facility-clinical-docs")

# 2. Build collection with clinical extractors
collection = client.collections.create(
    namespace_id=namespace.id,
    name="patient-charts",
    extractors=[
        "pdf-extraction",       # OCR for scanned charts
        "text-embedding-v2",    # Semantic embeddings
        "image-captioning",     # Wound photos, imaging
    ],
)

# 3. Upload clinical documents
client.buckets.upload(
    collection_id=collection.id,
    url="s3://facility-ehr-export/patient-charts/"
)

# 4. Create MDS-aligned retriever
retriever = client.retrievers.create(
    namespace_id=namespace.id,
    name="mds-documentation",
    stages=[
        {"type": "hybrid_search", "vector_weight": 0.5, "bm25_weight": 0.5, "top_k": 50},
        {"type": "attribute_filter", "conditions": [
            {"field": "mds_section", "operator": "in", "value": ["G", "J", "K"]}
        ]},
        {"type": "rerank", "model": "colbert-v2", "top_k": 10}
    ]
)

# 5. Retrieve MDS-relevant documentation
results = client.retrievers.execute(
    retriever_id=retriever.id,
    query="functional mobility and ADL performance for Section G"
)

for r in results:
    print(f"[{r.metadata.get('mds_section')}] {r.content[:120]}")