Docbuilder & Interface Extraction
Upload a vendor's PDF implementation guide and Vorian Connect extracts the structured interface metadata — field mappings, Z-segment definitions, and vendor-specific quirks. The docbuilder produces a ready-to-use Go dialect snippet and a full human-readable report in minutes, replacing hours of manual PDF annotation.
How It Works
- Upload a vendor PDF through the dashboard docbuilder page — the system extracts text using poppler-utils or pdfminer
- The docbuilder detects the protocol (HL7, ASTM, flat-file, or API-based) from the document text automatically
- For HL7 and ASTM guides, an AI-assisted extraction identifies field mappings, Z-segments, field remaps, and implementation notes
- The extraction produces three outputs: a structured JSON file, a Go dialect code snippet, and a full markdown report
- If HL7 trace files are provided alongside the PDF, the docbuilder cross-references them against the extracted specification and flags discrepancies
- For non-HL7/ASTM documents (flat-file, API), the docbuilder writes a protocol detection report and exits — no wasted processing
What's Included
- PDF text extraction: Automatic fallback from poppler-utils (pdftotext) to pdfminer.six — works on any Linux system with either tool installed
- Protocol detection: Identifies HL7 v2.x, ASTM LIS2-A2, flat-file (CSV/delimited), and API-based formats from document text
- HL7 extraction: Vendor name, model, HL7 version, field mappings (OBX-3 → internal code), Z-segment definitions, field remaps, and implementation notes
- ASTM extraction: Vendor name, model, ASTM version, record type inventory, and field remapping for vendor-specific layouts
- Trace analysis: Cross-reference HL7 trace files against extracted specifications to find discrepancies before interface deployment
- Generative output: Go dialect snippet (ready to compile), markdown report (human-readable), and structured JSON (machine-consumable)
- Admin-gated: Docbuilder access requires tenant admin authentication
Where It Fits
The docbuilder is the first step in the Vorian interface engineering workflow. A lab with a new analyzer sends the vendor's HL7 implementation guide to Vorian — the docbuilder extracts the interface metadata, generates a dialect, and surfaces any discrepancies against real trace data. The generated dialect is deployed to the edge agent's dialect registry, and the instrument is ready for message ingestion. What previously took a week of manual PDF annotation and field mapping is completed in one session.
Technical Details (for evaluators)
The docbuilder is a Go CLI that runs as a child process spawned by the dashboard API. It accepts PDF files, plain text files, and optional HL7 trace files. Extraction uses LLM-based structured data extraction with strict JSON schema validation — the output is verified against the expected schema before being returned to the dashboard. Protocol detection uses keyword density analysis (MSH/OBX for HL7, H|\\^&| for ASTM, CSV/delimited patterns for flat-file).
The docbuilder is designed for lab environments: it runs on the same infrastructure as the edge agent and gateway, requires no external services beyond the configured LLM provider, and processes documents up to 80,000 characters per chunk. Extraction reports are written to a configurable output directory with timestamped filenames for audit trail purposes.