Full Extraction
POST /v1/extractSend a PDF, image, URL, or upload through Processor. PolDex handles document reading, extraction, FastScript adjudication, exports, and webhooks.
Parser Compatibility
Play Well With Others
PolDex is the first dedicated insurance truth layer.
It turns raw documents or parser output into evidence-backed, schema-controlled insurance data.
PolDex is not here to replace your parser. Keep Hyperscience, Docugami, Reducto, Unstructured, LlamaParse, cloud OCR, or your internal parser where it already works.
Send the parsed output to PolDex for insurance-native adjudication, evidence grounding, conflict resolution, truth states, abstention, and schema enforcement through FastScript.
Two Paths
Use PolDex as the full extractor or the adjudication layer.
Customers can choose the entry point that matches their stack. Raw documents use the full extraction rail. Existing parser output uses the lighter FastScript adjudication rail.
Parser Output Import
POST /v1/connectors/parser-output/importSend parsed markdown, text, JSON, tables, or parser events. PolDex skips raw-document reading and applies a 50% parsed-output credit discount, minimum 1 credit.
Architecture
Raw document in your stack. Insurance truth from PolDex.
The important boundary is simple: parsers make documents machine-readable. FastScript decides what becomes insurance truth.
Raw PDF or scan -> Your parser -> Parsed text / markdown / JSON -> PolDex FastScript -> Evidence-backed insurance JSON
Parse where you already trust
Use your existing parser for OCR, layout extraction, privacy controls, compliance workflows, or internal review.
Send parsed output
Post text, markdown, JSON, tables, or adapted parser events into PolDex through API and connector rails.
FastScript adjudicates
PolDex applies insurance schemas, evidence rules, conflicts, truth states, abstention, and benchmark-gated behavior.
Return insurance truth
Downstream systems receive structured JSON plus CSV/XLSX artifacts, webhooks, connector events, and signed result links.
Parser Output Compatible
Works on top of the document stack customers already bought.
The generic parser-output rail accepts customer-provided text, markdown, JSON, tables, and parser event payloads. Reducto remains a named convenience rail for teams already standardized on Reducto Parse.
HyperscienceDocugamiReductoUnstructuredLlamaParseAWS TextractGoogle Document AIAzure Document IntelligenceIn-house OCR
Why It Matters
The underneath strategy makes PolDex easier to adopt.
Privacy-first rail
On parser-output workflows, PolDex does not need raw PDFs. Customers can keep source documents inside their existing parser, DMS, or compliance boundary.
No parser migration
Teams do not need to rip out Hyperscience, Docugami, Reducto, Unstructured, LlamaParse, or an internal parser to use PolDex.
Insurance control layer
PolDex is the insurance truth layer: schema enforcement, evidence grounding, conflict resolution, normalization, and abstention.
Reader stays replaceable
Parser choice can change without changing the customer-facing insurance result contract, because FastScript owns the adjudication layer.
Input Rail
Send parsed output, not necessarily raw files.
Use the generic parser-output import when the customer already has Hyperscience, Docugami, Unstructured, LlamaParse, cloud OCR, or an internal parser. The Reducto endpoint remains a named rail for Reducto-specific workflows.
curl -X POST https://poldex-api.plain-water-cd0f.workers.dev/v1/connectors/parser-output/import \
-H "x-api-key: pd_live_YOUR_KEY" \
-H "content-type: application/json" \
-d '{
"schema": "coi",
"source_parser": "docugami",
"client_ref": "dg-parse-001",
"metadata": {
"parser_job_id": "job_2026_001",
"page_count": 3
},
"markdown": "CERTIFICATE OF LIABILITY INSURANCE..."
}'Parser-output mode returns credit_policy and credits_captured in the response. Because the customer already handled document reading, this rail uses the parsed-output discount instead of full raw-document extraction pricing.
The insurance truth layer you plug into your current stack.
Keep the parser. Add FastScript where insurance facts need to become reliable data.