@talonic/docs 0.20.18 → 0.20.20

package/dist/content.js

@@ -1,5 +1,5 @@
 // src/content/helpers.ts
-function deriveBreadcrumbs(sections52, leafId, domain) {
+function deriveBreadcrumbs(sections55, leafId, domain) {
   const domainLabels = {
     api: "API Reference",
     platform: "Platform Guide",
@@ -10,20 +10,20 @@ function deriveBreadcrumbs(sections52, leafId, domain) {
     label: domainLabels[domain] ?? domain,
     slug: domain
   };
-  for (const group of sections52) {
+  for (const group of sections55) {
     const child = group.children?.find((c) => c.id === leafId);
     if (child) {
       return [root, { label: group.label, slug: group.id }, { label: child.label, slug: child.id }];
     }
   }
-  const topLevel = sections52.find((s) => s.id === leafId);
+  const topLevel = sections55.find((s) => s.id === leafId);
   if (topLevel) {
     return [root, { label: topLevel.label, slug: topLevel.id }];
   }
   return [root];
 }
-function derivePrevNext(sections52, leafId) {
-  const flat = sections52.flatMap(
+function derivePrevNext(sections55, leafId) {
+  const flat = sections55.flatMap(
     (s) => s.children ?? [{ id: s.id, label: s.label }]
   );
   const idx = flat.findIndex((c) => c.id === leafId);
@@ -250,6 +250,29 @@ var API_NAV_SECTIONS = [
   { id: "registry", label: "Registry", children: [
     { id: "registry-query", label: "Query" }
   ] },
+  { id: "data-products", label: "Data Products", children: [
+    { id: "list-data-products", label: "List Data Products" },
+    { id: "get-data-product", label: "Get Data Product" },
+    { id: "delete-data-product", label: "Delete Data Product" },
+    { id: "get-data-product-results", label: "Get Results" }
+  ] },
+  { id: "data-policies", label: "Data Policies", children: [
+    { id: "list-data-policies", label: "List Policies" },
+    { id: "create-data-policy", label: "Create Policy" },
+    { id: "get-data-policy", label: "Get Policy" },
+    { id: "update-data-policy", label: "Update Policy" },
+    { id: "delete-data-policy", label: "Delete Policy" },
+    { id: "list-data-policy-versions", label: "List Versions" },
+    { id: "list-data-policy-fields", label: "List Fields" },
+    { id: "list-data-policy-rules", label: "List Rules" }
+  ] },
+  { id: "record-sets", label: "Record Sets", children: [
+    { id: "list-record-sets", label: "List Record Sets" },
+    { id: "get-record-set", label: "Get Record Set" },
+    { id: "list-record-set-fields", label: "List Fields" },
+    { id: "list-record-set-records", label: "List Records" },
+    { id: "export-record-set", label: "Export" }
+  ] },
   { id: "billing", label: "Billing", children: [
     { id: "billing-settings", label: "Settings" },
     { id: "billing-topup", label: "Auto Top-Up" },
@@ -25906,99 +25929,1593 @@ var sections49 = [
   }
 ];
 
-// src/content/sdk/sections.json
-var sections_default = [
+// src/content/api/data-products.ts
+var sections50 = [
   {
-    slug: "sdk-introduction",
-    parentSlug: "sdk-overview",
-    title: "Introduction",
-    seoTitle: "Node SDK Introduction \u2014 Talonic Docs",
-    description: "Official Talonic SDK for Node.js and TypeScript. Extract structured, schema-validated data from any document with a single function call.",
+    slug: "list-data-products",
+    parentSlug: "data-products",
+    title: "List Data Products",
+    seoTitle: "List Data Products Endpoint \u2014 Talonic Docs",
+    description: "List assembled data products with cursor-based pagination. Filter by status to find active, draft, or archived products across your workspace.",
     content: [
-      { type: "paragraph", text: "The `@talonic/node` SDK is the official Node.js and TypeScript client for the Talonic API. Extract structured, schema-validated data from any document with a single function call." },
-      { type: "paragraph", text: "Zero runtime dependencies. Requires Node.js 18 or newer." },
-      { type: "callout", text: "Looking for the AI agent path? [`@talonic/mcp`](https://github.com/talonicdev/talonic-mcp) wraps this SDK as a Model Context Protocol server. Install it into Claude Desktop, Cursor, Cline, Continue, or Cowork and any MCP-aware agent can extract documents directly." }
+      { type: "paragraph", text: "Data products are the final assembled output datasets produced by the Talonic pipeline. Each data product represents a structured, validated, and optionally resolved collection of records ready for downstream consumption. Data products are created automatically when a job run completes validation, or manually via the platform UI." },
+      { type: "paragraph", text: "Use this endpoint to list all data products in your workspace. Results are returned in reverse chronological order by default and support cursor-based pagination. You can filter by status to find only active products, or locate draft and archived products for lifecycle management." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/data-products",
+        summary: "List data products with optional status filter and cursor-based pagination.",
+        description: "Requires read scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Query parameters",
+            params: [
+              { name: "status", type: "string", description: "Filter by product status (e.g. active, draft, archived)." },
+              { name: "limit", type: "integer", default: "20", description: "Maximum number of items to return (1-100)." },
+              { name: "cursor", type: "string", description: "Opaque pagination cursor from a previous response." },
+              { name: "order", type: "string", default: "desc", description: "Sort order by creation date (asc | desc)." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "list-data-products-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "data", type: "array", description: "Array of data product objects." },
+          { name: "data[].id", type: "string", description: "Data product UUID." },
+          { name: "data[].name", type: "string", description: "Human-readable product name." },
+          { name: "data[].description", type: "string | null", description: "Optional description of the data product." },
+          { name: "data[].schema_id", type: "string", description: "UUID of the user schema that defines the output columns." },
+          { name: "data[].run_id", type: "string", description: "UUID of the source job run that produced the data." },
+          { name: "data[].status", type: "string", description: "Product status: active, draft, or archived." },
+          { name: "data[].created_at", type: "string", description: "ISO 8601 creation timestamp." },
+          { name: "data[].updated_at", type: "string", description: "ISO 8601 last update timestamp." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s https://api.talonic.ai/v1/data-products?status=active&limit=10 \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "data": [
+    {
+      "id": "d1a2b3c4-e5f6-7890-abcd-ef1234567890",
+      "name": "Q4 Invoice Extract",
+      "description": "Structured invoice data from Q4 batch",
+      "schema_id": "s1a2b3c4-e5f6-7890-abcd-ef1234567890",
+      "run_id": "r1a2b3c4-e5f6-7890-abcd-ef1234567890",
+      "status": "active",
+      "created_at": "2024-11-01T14:20:00.000Z",
+      "updated_at": "2024-11-01T14:25:30.000Z"
+    }
+  ],
+  "cursor": "eyJpZCI6ImQxYTJiM2M0In0="
+}`
+      },
+      { type: "heading", level: 2, id: "list-data-products-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
     ],
     related: [
-      { label: "Installation", slug: "sdk-installation" },
-      { label: "Quick Start", slug: "sdk-quickstart" },
-      { label: "MCP Server", slug: "mcp-introduction" }
+      { label: "Resolutions", slug: "list-resolutions" },
+      { label: "Jobs", slug: "list-jobs" },
+      { label: "Schemas", slug: "list-schemas" }
     ],
     faq: [
-      { question: "What is the Talonic Node SDK?", answer: "The official Node.js and TypeScript client for extracting structured, schema-validated data from documents via the Talonic API." },
-      { question: "What Node.js version is required?", answer: "Node.js 18 or newer. The SDK has zero runtime dependencies." }
+      { question: "What is a data product?", answer: "A data product is the final assembled dataset from the Talonic pipeline. It contains structured, validated, and optionally resolved records ready for export or delivery. Each product is tied to a schema and a source job run." },
+      { question: "How do data product statuses work?", answer: "Data products start as `draft` during assembly, move to `active` once validated and ready for consumption, and can be `archived` when no longer needed. Only `active` products are included in delivery bindings by default." },
+      { question: "Can I filter data products by schema?", answer: "The list endpoint currently supports filtering by `status`. To find products for a specific schema, list all products and filter client-side by `schema_id`, or use the filter/search endpoints for more advanced queries." }
     ],
-    mentions: ["Node.js", "TypeScript", "SDK", "npm", "document extraction"]
+    mentions: ["data product", "export", "structured output"]
   },
   {
-    slug: "sdk-installation",
-    parentSlug: "sdk-overview",
-    title: "Installation",
-    seoTitle: "Install Talonic Node SDK \u2014 Talonic Docs",
-    description: "Install the @talonic/node package via npm. Zero runtime dependencies, requires Node.js 18+.",
+    slug: "get-data-product",
+    parentSlug: "data-products",
+    title: "Get Data Product",
+    seoTitle: "Get Data Product Endpoint \u2014 Talonic Docs",
+    description: "Retrieve a single data product by ID with its schema reference, source run, status, and metadata. Requires read scope for the workspace.",
     content: [
-      { type: "code", language: "bash", title: "Install via npm", code: "npm install @talonic/node" },
-      { type: "paragraph", text: "Requires Node.js 18 or newer. Zero runtime dependencies." }
+      { type: "paragraph", text: "Retrieve the full details of a specific data product by its UUID. The response includes the product name, description, associated schema and source run references, current status, and timestamps. Use this endpoint to inspect a product before fetching its result rows." },
+      { type: "paragraph", text: "The `schema_id` field references the user schema that defines the output columns. The `run_id` field references the job run that produced the underlying extracted data. Together these let you trace the full lineage of the data product back to the original documents." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/data-products/{id}",
+        summary: "Get a single data product by ID.",
+        description: "Requires read scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Data product UUID." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "get-data-product-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "id", type: "string", description: "Data product UUID." },
+          { name: "name", type: "string", description: "Human-readable product name." },
+          { name: "description", type: "string | null", description: "Optional description." },
+          { name: "schema_id", type: "string", description: "UUID of the user schema." },
+          { name: "run_id", type: "string", description: "UUID of the source job run." },
+          { name: "status", type: "string", description: "Product status: active, draft, or archived." },
+          { name: "created_at", type: "string", description: "ISO 8601 creation timestamp." },
+          { name: "updated_at", type: "string", description: "ISO 8601 last update timestamp." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s https://api.talonic.ai/v1/data-products/d1a2b3c4-e5f6-7890-abcd-ef1234567890 \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "id": "d1a2b3c4-e5f6-7890-abcd-ef1234567890",
+  "name": "Q4 Invoice Extract",
+  "description": "Structured invoice data from Q4 batch",
+  "schema_id": "s1a2b3c4-e5f6-7890-abcd-ef1234567890",
+  "run_id": "r1a2b3c4-e5f6-7890-abcd-ef1234567890",
+  "status": "active",
+  "created_at": "2024-11-01T14:20:00.000Z",
+  "updated_at": "2024-11-01T14:25:30.000Z"
+}`
+      },
+      { type: "heading", level: 2, id: "get-data-product-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Data product not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
     ],
     related: [
-      { label: "Authentication", slug: "sdk-authentication" },
-      { label: "Quick Start", slug: "sdk-quickstart" }
+      { label: "List Data Products", slug: "list-data-products" },
+      { label: "Get Data Product Results", slug: "get-data-product-results" },
+      { label: "Jobs", slug: "list-jobs" }
     ],
     faq: [
-      { question: "How do I install the Talonic Node SDK?", answer: "Run npm install @talonic/node. Requires Node.js 18 or newer. Zero runtime dependencies." }
+      { question: "How do I trace a data product back to its source documents?", answer: "The `run_id` field references the job run. Use `GET /v1/jobs/{run_id}` to find the job details, including the list of documents that were processed. The `schema_id` field tells you which output schema was used." },
+      { question: "What does the schema_id represent?", answer: "The `schema_id` references the user schema that defines the output columns and field mappings for this data product. It determines which fields appear in the result rows and how they are named." }
     ],
-    mentions: ["npm", "Node.js"]
+    mentions: ["data product", "export", "structured output", "lineage"]
   },
   {
-    slug: "sdk-authentication",
-    parentSlug: "sdk-overview",
-    title: "Authentication",
-    seoTitle: "SDK Authentication \u2014 Talonic Docs",
-    description: "Get a Talonic API key and configure it for the Node SDK. Each workspace is isolated with private documents and schemas.",
+    slug: "delete-data-product",
+    parentSlug: "data-products",
+    title: "Delete Data Product",
+    seoTitle: "Delete Data Product Endpoint \u2014 Talonic Docs",
+    description: "Permanently delete a data product and its result rows. Requires write scope. The source job run and extracted data are not affected.",
     content: [
-      { type: "paragraph", text: "Every user runs against their own Talonic workspace, so each user needs their own key. Workspaces are isolated; your documents and schemas are private to you." },
-      { type: "list", ordered: true, items: [
-        "Sign up at [https://app.talonic.com](https://app.talonic.com). Free tier: 50 extractions per day, no credit card.",
-        "Settings → API Keys → Create New Key.",
-        "Copy the `tlnc_` value.",
-        "Set it as the `TALONIC_API_KEY` environment variable, or pass it directly to the client constructor."
-      ] },
-      { type: "callout", text: "Keep your API key secret. Do not expose it in client-side code or version control." }
+      { type: "paragraph", text: "Permanently delete a data product and all associated result rows. This action is irreversible. The source job run, its documents, and extracted data are not affected by this operation. Use this to remove products that are no longer needed or to clean up test data." },
+      { type: "callout", variant: "warning", text: "Deletion is permanent. All result rows associated with this data product are removed. Any active delivery bindings referencing this product will stop delivering data." },
+      {
+        type: "endpoint",
+        method: "DELETE",
+        path: "/v1/data-products/{id}",
+        summary: "Delete a data product and its results.",
+        description: "Requires write scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Data product UUID." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "delete-data-product-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "deleted", type: "boolean", description: "Always true on success." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s -X DELETE https://api.talonic.ai/v1/data-products/d1a2b3c4-e5f6-7890-abcd-ef1234567890 \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "deleted": true
+}`
+      },
+      { type: "heading", level: 2, id: "delete-data-product-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Data product not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
     ],
     related: [
-      { label: "Configuration", slug: "sdk-configuration" },
-      { label: "API Authentication", slug: "authentication" }
+      { label: "List Data Products", slug: "list-data-products" },
+      { label: "Resolutions", slug: "list-resolutions" },
+      { label: "Schemas", slug: "list-schemas" }
     ],
     faq: [
-      { question: "Where do I get a Talonic API key?", answer: "Sign up at app.talonic.com, go to Settings \u2192 API Keys \u2192 Create New Key. Free tier includes 50 extractions per day." }
+      { question: "Does deleting a data product affect the source job run?", answer: "No. The source job run, its documents, and all extracted data are completely unaffected. Only the data product and its assembled result rows are removed." },
+      { question: "What happens to delivery bindings when a product is deleted?", answer: "Any delivery bindings that reference the deleted product will stop delivering data. The bindings themselves are not deleted, but they will produce errors on the next delivery attempt. Update or remove affected bindings after deleting a product." }
     ],
-    mentions: ["API key", "authentication", "workspace"]
+    mentions: ["data product", "export", "structured output"]
   },
   {
-    slug: "sdk-quickstart",
-    parentSlug: "sdk-overview",
-    title: "Quick Start",
-    seoTitle: "Node SDK Quick Start \u2014 Talonic Docs",
-    description: "Extract structured data from a document in five lines of TypeScript using the Talonic Node SDK.",
+    slug: "get-data-product-results",
+    parentSlug: "data-products",
+    title: "Get Data Product Results",
+    seoTitle: "Get Data Product Results \u2014 Talonic Docs",
+    description: "Retrieve the structured result rows of a data product with cursor-based pagination. Each row contains field values aligned to the product schema.",
     content: [
-      { type: "code", language: "typescript", title: "Extract an invoice", code: 'import { Talonic } from "@talonic/node"\n\nconst talonic = new Talonic({ apiKey: process.env.TALONIC_API_KEY! })\n\nconst result = await talonic.extract({\n  file_path: "./invoice.pdf",\n  schema: {\n    vendor_name: "string",\n    invoice_number: "string",\n    total_amount: "number",\n    due_date: "date",\n  },\n})\n\nconsole.log(result.data)\n// { vendor_name: "Acme Corp", invoice_number: "INV-2024-0847", total_amount: 14250, due_date: "2024-03-15" }' }
+      { type: "paragraph", text: "Retrieve the assembled result rows for a data product. Each row contains the structured field values aligned to the product's schema, with one row per document or entity. Results are returned with cursor-based pagination for efficient traversal of large datasets." },
+      { type: "paragraph", text: "The result rows represent the final output of the Talonic pipeline: extracted, validated, and optionally resolved field values. Each row maps field names (from the associated user schema) to their values. Use this endpoint to programmatically consume the structured data for downstream integration, analytics, or export." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/data-products/{id}/results",
+        summary: "Get paginated result rows for a data product.",
+        description: "Requires read scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Data product UUID." }
+            ]
+          },
+          {
+            type: "param-table",
+            title: "Query parameters",
+            params: [
+              { name: "limit", type: "integer", default: "20", description: "Maximum number of rows to return (1-100)." },
+              { name: "cursor", type: "string", description: "Opaque pagination cursor from a previous response." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "get-data-product-results-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "data", type: "array", description: "Array of result row objects." },
+          { name: "data[].id", type: "string", description: "Result row UUID." },
+          { name: "data[].document_id", type: "string", description: "Source document UUID." },
+          { name: "data[].fields", type: "object", description: "Key-value map of field names to extracted values." },
+          { name: "data[].created_at", type: "string", description: "ISO 8601 timestamp." },
+          { name: "cursor", type: "string | null", description: "Pagination cursor for the next page, or null if no more results." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s "https://api.talonic.ai/v1/data-products/d1a2b3c4-e5f6-7890-abcd-ef1234567890/results?limit=10" \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "data": [
+    {
+      "id": "row-1a2b3c4d-e5f6-7890-abcd-ef1234567890",
+      "document_id": "doc-a1b2c3d4-e5f6-7890-abcd-ef1234567890",
+      "fields": {
+        "invoice_number": "INV-2024-0042",
+        "vendor_name": "Acme Corp",
+        "total_amount": "1250.00",
+        "currency": "USD",
+        "invoice_date": "2024-10-15"
+      },
+      "created_at": "2024-11-01T14:25:00.000Z"
+    }
+  ],
+  "cursor": "eyJpZCI6InJvdy0xYTJiIn0="
+}`
+      },
+      { type: "heading", level: 2, id: "get-data-product-results-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Data product not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
     ],
     related: [
-      { label: "Extract", slug: "sdk-extract" },
-      { label: "Configuration", slug: "sdk-configuration" },
-      { label: "Schemas", slug: "sdk-schemas" }
+      { label: "Get Data Product", slug: "get-data-product" },
+      { label: "Resolutions", slug: "list-resolutions" },
+      { label: "Jobs", slug: "list-jobs" },
+      { label: "Schemas", slug: "list-schemas" }
     ],
     faq: [
-      { question: "How do I extract data from a document with the SDK?", answer: "Call talonic.extract() with a file_path and a schema defining the fields you want. Returns structured JSON with confidence scores." }
+      { question: "What format are the result row fields in?", answer: "The `fields` object is a flat key-value map where keys are field names from the associated schema and values are strings. All values are returned as strings regardless of the field type defined in the schema." },
+      { question: "How do I export all results at once?", answer: "Paginate through all results using the `cursor` parameter until `cursor` is null. Alternatively, use the platform UI to export the data product as CSV or XLSX, which handles pagination automatically." },
+      { question: "Are resolved values included in the results?", answer: "Yes. If a resolution run was executed against the source job, the result rows contain the resolved canonical values rather than the raw extracted values. The data product always reflects the latest resolved state." }
     ],
-    mentions: ["extract", "schema", "TypeScript", "quickstart"]
-  },
+    mentions: ["data product", "export", "structured output", "result rows"]
+  }
+];
+
+// src/content/api/data-policies.ts
+var sections51 = [
   {
-    slug: "sdk-configuration",
-    parentSlug: "sdk-client",
-    title: "Configuration",
-    seoTitle: "SDK Configuration \u2014 Talonic Docs",
-    description: "Configure the Talonic client with API key, base URL, timeout, max retries, and custom fetch function.",
+    slug: "list-data-policies",
+    parentSlug: "data-policies",
+    title: "List Data Policies",
+    seoTitle: "List Data Policies Endpoint \u2014 Talonic Docs",
+    description: "List data policies that define field transformations, normalization rules, and lookup cascades applied during resolution runs. Supports cursor-based pagination.",
+    content: [
+      { type: "paragraph", text: "Data policies are first-class, versioned configuration objects that define how field values are transformed, normalized, and validated during resolution runs. Each policy contains a set of fields (output contract) and rules (transformation logic) that are compiled into an executable pipeline. Policies support 14 rule types including direct mapping, lookup cascades, Lua scripting, and deterministic computation." },
+      { type: "paragraph", text: "Use this endpoint to list all data policies in your workspace. Policies are returned in reverse chronological order by default. Each policy can have multiple versions, allowing you to iterate on transformation logic without affecting in-flight resolution runs that have already captured a policy snapshot." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/data-policies",
+        summary: "List all data policies in the workspace.",
+        description: "Requires read scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Query parameters",
+            params: [
+              { name: "limit", type: "integer", default: "20", description: "Maximum number of items to return (1-100)." },
+              { name: "cursor", type: "string", description: "Opaque pagination cursor from a previous response." },
+              { name: "order", type: "string", default: "desc", description: "Sort order by creation date (asc | desc)." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "list-data-policies-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "data", type: "array", description: "Array of data policy objects." },
+          { name: "data[].id", type: "string", description: "Data policy UUID." },
+          { name: "data[].name", type: "string", description: "Human-readable policy name." },
+          { name: "data[].description", type: "string | null", description: "Optional description of the policy purpose." },
+          { name: "data[].version", type: "integer", description: "Current version number." },
+          { name: "data[].created_at", type: "string", description: "ISO 8601 creation timestamp." },
+          { name: "data[].updated_at", type: "string", description: "ISO 8601 last update timestamp." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s https://api.talonic.ai/v1/data-policies \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "data": [
+    {
+      "id": "p1a2b3c4-e5f6-7890-abcd-ef1234567890",
+      "name": "Invoice Normalization",
+      "description": "Standardize currency codes, country names, and date formats for invoice processing",
+      "version": 3,
+      "created_at": "2024-10-01T09:00:00.000Z",
+      "updated_at": "2024-10-15T11:30:00.000Z"
+    }
+  ],
+  "cursor": "eyJpZCI6InAxYTJiM2M0In0="
+}`
+      },
+      { type: "heading", level: 2, id: "list-data-policies-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "Create Data Policy", slug: "create-data-policy" },
+      { label: "Resolutions", slug: "list-resolutions" },
+      { label: "Schemas", slug: "list-schemas" }
+    ],
+    faq: [
+      { question: "What do data policies do?", answer: "Data policies define how field values are transformed and normalized during resolution runs. They support lookup cascades (mapping raw values to canonical codes via reference tables), Lua scripting for custom logic, deterministic computation, and direct field-to-field mappings." },
+      { question: "How do data policy versions work?", answer: "Each time you update a policy, a new version is created. Resolution runs capture a snapshot of the active version at run time, so changes to a policy do not retroactively affect completed runs. You can inspect which version was used via the `policy_snapshot` field on a resolution run." },
+      { question: "Can I have multiple data policies per workspace?", answer: "Yes. You can create multiple policies for different transformation scenarios. Each resolution run uses the active policy configured for the workspace, but you can switch between policies as needed." }
+    ],
+    mentions: ["data policy", "transformation", "normalization", "lookup", "Lua"]
+  },
+  {
+    slug: "create-data-policy",
+    parentSlug: "data-policies",
+    title: "Create Data Policy",
+    seoTitle: "Create Data Policy Endpoint \u2014 Talonic Docs",
+    description: "Create a new data policy to define field transformation rules, lookup cascades, and normalization logic for resolution runs.",
+    content: [
+      { type: "paragraph", text: "Create a new data policy in your workspace. The policy starts at version 1 with no fields or rules. After creation, use the fields and rules endpoints to define the output contract and transformation logic. Policies are not active until they contain at least one field and one rule." },
+      { type: "paragraph", text: "The policy name should be descriptive enough to distinguish it from other policies in the workspace. Use the optional description field to document the policy purpose, the types of documents it targets, and the normalization strategies it employs." },
+      {
+        type: "endpoint",
+        method: "POST",
+        path: "/v1/data-policies",
+        summary: "Create a new data policy.",
+        description: "Requires write scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Body parameters",
+            params: [
+              { name: "name", type: "string", required: true, description: "Human-readable policy name." },
+              { name: "description", type: "string", description: "Optional description of the policy purpose." }
+            ]
+          },
+          {
+            type: "code",
+            title: "Request body",
+            code: `{
+  "name": "Invoice Normalization",
+  "description": "Standardize currency codes, country names, and date formats"
+}`
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "create-data-policy-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields (201 Created)",
+        params: [
+          { name: "id", type: "string", description: "Data policy UUID." },
+          { name: "name", type: "string", description: "Policy name." },
+          { name: "description", type: "string | null", description: "Policy description." },
+          { name: "version", type: "integer", description: "Initial version, always 1." },
+          { name: "created_at", type: "string", description: "ISO 8601 creation timestamp." },
+          { name: "updated_at", type: "string", description: "ISO 8601 last update timestamp." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s -X POST https://api.talonic.ai/v1/data-policies \\
+  -H "Authorization: Bearer tlnc_your_api_key" \\
+  -H "Content-Type: application/json" \\
+  -d '{"name":"Invoice Normalization","description":"Standardize currency codes, country names, and date formats"}'`
+      },
+      {
+        type: "code",
+        title: "Response (201 Created)",
+        code: `{
+  "id": "p1a2b3c4-e5f6-7890-abcd-ef1234567890",
+  "name": "Invoice Normalization",
+  "description": "Standardize currency codes, country names, and date formats",
+  "version": 1,
+  "created_at": "2024-10-01T09:00:00.000Z",
+  "updated_at": "2024-10-01T09:00:00.000Z"
+}`
+      },
+      { type: "heading", level: 2, id: "create-data-policy-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "400", type: "bad_request", description: "Invalid request body or missing required fields." },
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "List Data Policies", slug: "list-data-policies" },
+      { label: "List Data Policy Fields", slug: "list-data-policy-fields" },
+      { label: "List Data Policy Rules", slug: "list-data-policy-rules" }
+    ],
+    faq: [
+      { question: "What should I do after creating a policy?", answer: "After creating a policy, define its output fields via the fields endpoint and add transformation rules via the rules endpoint. The policy is not active until it has at least one field and one rule configured." },
+      { question: "Can I rename a policy after creation?", answer: "Yes. Use the `PATCH /v1/data-policies/{id}` endpoint to update the name or description. Renaming does not affect the policy version or any existing resolution runs that captured a snapshot of this policy." }
+    ],
+    mentions: ["data policy", "transformation", "normalization"]
+  },
+  {
+    slug: "get-data-policy",
+    parentSlug: "data-policies",
+    title: "Get Data Policy",
+    seoTitle: "Get Data Policy Endpoint \u2014 Talonic Docs",
+    description: "Retrieve a data policy by ID with its fields and rules inlined. Returns the full policy configuration needed to understand transformation behavior.",
+    content: [
+      { type: "paragraph", text: "Retrieve the complete details of a specific data policy, including its fields and rules inlined in the response. This provides a single-call view of the entire policy configuration, which is useful for auditing transformation logic or debugging resolution results." },
+      { type: "paragraph", text: "The inlined fields define the output contract: which field keys the policy declares and their expected types. The inlined rules define the transformation logic: lookup tables, direct mappings, Lua scripts, and computed fields. Rules are executed in topological order based on their declared dependencies." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/data-policies/{id}",
+        summary: "Get a data policy with fields and rules inlined.",
+        description: "Requires read scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Data policy UUID." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "get-data-policy-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "id", type: "string", description: "Data policy UUID." },
+          { name: "name", type: "string", description: "Policy name." },
+          { name: "description", type: "string | null", description: "Policy description." },
+          { name: "version", type: "integer", description: "Current version number." },
+          { name: "fields", type: "array", description: "Array of declared output field objects." },
+          { name: "rules", type: "array", description: "Array of transformation rule objects." },
+          { name: "created_at", type: "string", description: "ISO 8601 creation timestamp." },
+          { name: "updated_at", type: "string", description: "ISO 8601 last update timestamp." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s https://api.talonic.ai/v1/data-policies/p1a2b3c4-e5f6-7890-abcd-ef1234567890 \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "id": "p1a2b3c4-e5f6-7890-abcd-ef1234567890",
+  "name": "Invoice Normalization",
+  "description": "Standardize currency codes, country names, and date formats",
+  "version": 3,
+  "fields": [
+    { "field_key": "country_code", "type": "string" },
+    { "field_key": "currency", "type": "string" }
+  ],
+  "rules": [
+    {
+      "id": "r1a2b3c4-0001",
+      "field_key": "country_code",
+      "rule_type": "lookup",
+      "config": { "table": "country_codes", "source_field": "country" }
+    }
+  ],
+  "created_at": "2024-10-01T09:00:00.000Z",
+  "updated_at": "2024-10-15T11:30:00.000Z"
+}`
+      },
+      { type: "heading", level: 2, id: "get-data-policy-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Data policy not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "List Data Policies", slug: "list-data-policies" },
+      { label: "Resolutions", slug: "list-resolutions" },
+      { label: "Schemas", slug: "list-schemas" }
+    ],
+    faq: [
+      { question: "Why are fields and rules inlined in the response?", answer: "Inlining fields and rules gives you a complete view of the policy in a single API call. This is useful for auditing the full transformation pipeline, debugging resolution results, or exporting the policy configuration for version control." },
+      { question: "What are the supported rule types?", answer: "Data policies support 14 rule types including `lookup` (reference table matching), `direct` (field-to-field mapping), `lua` (custom Lua scripting), `compute` (deterministic formulas), `format` (string formatting), and more. Each rule type has its own configuration schema." },
+      { question: "How does rule execution order work?", answer: "Rules are compiled into a topological order based on their declared field dependencies. A rule that reads from `country` and writes to `country_code` will always execute after the rule that produces `country`. Circular dependencies are detected at compile time." }
+    ],
+    mentions: ["data policy", "transformation", "normalization", "lookup", "Lua"]
+  },
+  {
+    slug: "update-data-policy",
+    parentSlug: "data-policies",
+    title: "Update Data Policy",
+    seoTitle: "Update Data Policy Endpoint \u2014 Talonic Docs",
+    description: "Update the name or description of a data policy. Creates a new version. Does not affect in-flight resolution runs that captured a previous snapshot.",
+    content: [
+      { type: "paragraph", text: "Update the metadata of an existing data policy. You can change the name, description, or both. Each update increments the policy version number. Resolution runs that have already captured a policy snapshot are not affected by this change." },
+      { type: "paragraph", text: "To modify the transformation logic (fields and rules), use the dedicated fields and rules endpoints. This endpoint only updates the policy-level metadata. Version increments from metadata changes are tracked separately from version increments caused by field or rule modifications." },
+      {
+        type: "endpoint",
+        method: "PATCH",
+        path: "/v1/data-policies/{id}",
+        summary: "Update data policy name or description.",
+        description: "Requires write scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Data policy UUID." }
+            ]
+          },
+          {
+            type: "param-table",
+            title: "Body parameters",
+            params: [
+              { name: "name", type: "string", description: "Updated policy name." },
+              { name: "description", type: "string", description: "Updated policy description." }
+            ]
+          },
+          {
+            type: "code",
+            title: "Request body",
+            code: `{
+  "name": "Invoice Normalization v2",
+  "description": "Updated: added VAT normalization rules"
+}`
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "update-data-policy-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "id", type: "string", description: "Data policy UUID." },
+          { name: "name", type: "string", description: "Updated policy name." },
+          { name: "description", type: "string | null", description: "Updated policy description." },
+          { name: "version", type: "integer", description: "Incremented version number." },
+          { name: "created_at", type: "string", description: "ISO 8601 creation timestamp." },
+          { name: "updated_at", type: "string", description: "ISO 8601 last update timestamp." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s -X PATCH https://api.talonic.ai/v1/data-policies/p1a2b3c4-e5f6-7890-abcd-ef1234567890 \\
+  -H "Authorization: Bearer tlnc_your_api_key" \\
+  -H "Content-Type: application/json" \\
+  -d '{"name":"Invoice Normalization v2"}'`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "id": "p1a2b3c4-e5f6-7890-abcd-ef1234567890",
+  "name": "Invoice Normalization v2",
+  "description": "Updated: added VAT normalization rules",
+  "version": 4,
+  "created_at": "2024-10-01T09:00:00.000Z",
+  "updated_at": "2024-10-20T08:15:00.000Z"
+}`
+      },
+      { type: "heading", level: 2, id: "update-data-policy-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "400", type: "bad_request", description: "Invalid request body." },
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Data policy not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "Get Data Policy", slug: "get-data-policy" },
+      { label: "List Data Policy Versions", slug: "list-data-policy-versions" }
+    ],
+    faq: [
+      { question: "Does updating a policy affect running resolutions?", answer: "No. Resolution runs capture a snapshot of the policy at creation time. Updates to the policy only affect future resolution runs. Completed and in-flight runs are unaffected." },
+      { question: "Do I need to provide all fields in the update?", answer: "No. This is a PATCH endpoint, so you only need to include the fields you want to change. Omitted fields retain their current values." }
+    ],
+    mentions: ["data policy", "transformation", "normalization"]
+  },
+  {
+    slug: "delete-data-policy",
+    parentSlug: "data-policies",
+    title: "Delete Data Policy",
+    seoTitle: "Delete Data Policy Endpoint \u2014 Talonic Docs",
+    description: "Permanently delete a data policy and all its versions, fields, and rules. Requires write scope. Does not affect resolution runs that captured a snapshot.",
+    content: [
+      { type: "paragraph", text: "Permanently delete a data policy and all associated versions, fields, and rules. This action is irreversible. Resolution runs that previously captured a snapshot of this policy are not affected, as they retain their own copy of the policy configuration." },
+      { type: "callout", variant: "warning", text: "Deletion is permanent. All versions, fields, and rules of this policy are removed. Future resolution runs will not be able to reference this policy. Existing resolution runs with captured snapshots are unaffected." },
+      {
+        type: "endpoint",
+        method: "DELETE",
+        path: "/v1/data-policies/{id}",
+        summary: "Delete a data policy and all its versions.",
+        description: "Requires write scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Data policy UUID." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "delete-data-policy-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "deleted", type: "boolean", description: "Always true on success." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s -X DELETE https://api.talonic.ai/v1/data-policies/p1a2b3c4-e5f6-7890-abcd-ef1234567890 \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "deleted": true
+}`
+      },
+      { type: "heading", level: 2, id: "delete-data-policy-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Data policy not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "List Data Policies", slug: "list-data-policies" },
+      { label: "Resolutions", slug: "list-resolutions" }
+    ],
+    faq: [
+      { question: "Does deleting a policy break existing resolution runs?", answer: "No. Resolution runs capture a complete policy snapshot at creation time. Deleting the policy only prevents future resolution runs from using it. All historical results are preserved." },
+      { question: "Can I recover a deleted policy?", answer: "No. Deletion is permanent. If you need to preserve a policy configuration, export it via the GET endpoint before deleting. You can inspect historical snapshots in completed resolution runs." }
+    ],
+    mentions: ["data policy", "transformation", "normalization"]
+  },
+  {
+    slug: "list-data-policy-versions",
+    parentSlug: "data-policies",
+    title: "List Data Policy Versions",
+    seoTitle: "List Data Policy Versions \u2014 Talonic Docs",
+    description: "List all versions of a data policy showing the change history. Each version captures the policy state at a point in time for auditability.",
+    content: [
+      { type: "paragraph", text: "Retrieve the version history of a data policy. Each version represents a snapshot of the policy configuration at a point in time. Versions are created automatically when the policy, its fields, or its rules are modified. This endpoint is useful for auditing changes and understanding how the policy evolved." },
+      { type: "paragraph", text: "Version numbers are monotonically increasing integers starting at 1. When a resolution run captures a policy snapshot, it records the version number, allowing you to correlate resolution results with the specific policy configuration that produced them." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/data-policies/{id}/versions",
+        summary: "List all versions of a data policy.",
+        description: "Requires read scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Data policy UUID." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "list-data-policy-versions-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "data", type: "array", description: "Array of version objects." },
+          { name: "data[].version", type: "integer", description: "Version number." },
+          { name: "data[].created_at", type: "string", description: "ISO 8601 timestamp when this version was created." },
+          { name: "data[].fields_count", type: "integer", description: "Number of declared fields in this version." },
+          { name: "data[].rules_count", type: "integer", description: "Number of rules in this version." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s https://api.talonic.ai/v1/data-policies/p1a2b3c4-e5f6-7890-abcd-ef1234567890/versions \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "data": [
+    { "version": 3, "created_at": "2024-10-15T11:30:00.000Z", "fields_count": 5, "rules_count": 8 },
+    { "version": 2, "created_at": "2024-10-10T14:00:00.000Z", "fields_count": 4, "rules_count": 6 },
+    { "version": 1, "created_at": "2024-10-01T09:00:00.000Z", "fields_count": 2, "rules_count": 3 }
+  ]
+}`
+      },
+      { type: "heading", level: 2, id: "list-data-policy-versions-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Data policy not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "Get Data Policy", slug: "get-data-policy" },
+      { label: "Update Data Policy", slug: "update-data-policy" }
+    ],
+    faq: [
+      { question: "How do I see which version a resolution used?", answer: "The `policy_snapshot` field on a resolution run includes the version number. Compare it with the version list to understand which configuration was active when the resolution executed." },
+      { question: "Can I roll back to a previous version?", answer: "The API does not support direct rollback. To revert to a previous configuration, inspect the version history and manually re-apply the desired fields and rules, which creates a new version with the old configuration." }
+    ],
+    mentions: ["data policy", "transformation", "normalization", "version history"]
+  },
+  {
+    slug: "list-data-policy-fields",
+    parentSlug: "data-policies",
+    title: "List Data Policy Fields",
+    seoTitle: "List Data Policy Fields \u2014 Talonic Docs",
+    description: "List the declared output fields of a data policy. Fields define the output contract \u2014 which field keys the policy produces and their expected types.",
+    content: [
+      { type: "paragraph", text: "Retrieve the list of declared output fields for a data policy. Fields define the output contract: the set of field keys that the policy executor will emit in resolution results. Only fields declared here appear in the final output; any intermediate values computed by rules but not declared as fields are discarded." },
+      { type: "paragraph", text: "Each field has a `field_key` (the output column name) and a `type` (the expected value type). The field list is used by the resolution pipeline to validate output schema compliance and by the results UI to determine which columns to display. Adding or removing fields creates a new policy version." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/data-policies/{id}/fields",
+        summary: "List declared output fields for a data policy.",
+        description: "Requires read scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Data policy UUID." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "list-data-policy-fields-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "data", type: "array", description: "Array of field definition objects." },
+          { name: "data[].field_key", type: "string", description: "Output field key name." },
+          { name: "data[].type", type: "string", description: "Expected value type (e.g. string, number, date)." },
+          { name: "data[].description", type: "string | null", description: "Optional field description." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s https://api.talonic.ai/v1/data-policies/p1a2b3c4-e5f6-7890-abcd-ef1234567890/fields \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "data": [
+    { "field_key": "country_code", "type": "string", "description": "ISO 3166-1 alpha-2 country code" },
+    { "field_key": "currency", "type": "string", "description": "ISO 4217 currency code" },
+    { "field_key": "invoice_date", "type": "date", "description": "Normalized invoice date in ISO 8601 format" }
+  ]
+}`
+      },
+      { type: "heading", level: 2, id: "list-data-policy-fields-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Data policy not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "Get Data Policy", slug: "get-data-policy" },
+      { label: "List Data Policy Rules", slug: "list-data-policy-rules" },
+      { label: "Schemas", slug: "list-schemas" }
+    ],
+    faq: [
+      { question: "What happens to values not declared as fields?", answer: "Values computed by rules but not declared as policy fields are treated as intermediate variables and discarded from the final output. Only declared fields appear in resolution results. You can enable `emit_undeclared_cells` in pipeline config for debugging." },
+      { question: "Can Lua scripts write to undeclared field keys?", answer: "Yes. Lua scripts can write to any cell name as a temporary variable for intermediate computation. However, only values written to declared field keys will appear in the resolution output. Undeclared keys are silently dropped unless debugging is enabled." }
+    ],
+    mentions: ["data policy", "transformation", "normalization", "Lua"]
+  },
+  {
+    slug: "list-data-policy-rules",
+    parentSlug: "data-policies",
+    title: "List Data Policy Rules",
+    seoTitle: "List Data Policy Rules \u2014 Talonic Docs",
+    description: "List the transformation rules of a data policy. Rules define lookup cascades, Lua scripts, direct mappings, and computed fields executed during resolution.",
+    content: [
+      { type: "paragraph", text: "Retrieve the list of transformation rules for a data policy. Rules define the actual transformation logic executed during resolution: lookup cascades against reference tables, Lua scripting for custom logic, direct field-to-field mappings, deterministic computations, and format transformations. Rules are compiled into topological execution order based on field dependencies." },
+      { type: "paragraph", text: "Each rule targets a specific output field and declares its rule type and configuration. The policy compiler validates that all rule dependencies are satisfiable and detects circular references at compile time. During execution, rules are processed in dependency order so that upstream fields are resolved before downstream rules that read from them." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/data-policies/{id}/rules",
+        summary: "List transformation rules for a data policy.",
+        description: "Requires read scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Data policy UUID." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "list-data-policy-rules-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "data", type: "array", description: "Array of rule objects." },
+          { name: "data[].id", type: "string", description: "Rule UUID." },
+          { name: "data[].field_key", type: "string", description: "Target output field key." },
+          { name: "data[].rule_type", type: "string", description: "Rule type (e.g. lookup, direct, lua, compute, format)." },
+          { name: "data[].config", type: "object", description: "Rule-type-specific configuration." },
+          { name: "data[].order", type: "integer", description: "Execution order within the compiled pipeline." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s https://api.talonic.ai/v1/data-policies/p1a2b3c4-e5f6-7890-abcd-ef1234567890/rules \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "data": [
+    {
+      "id": "r1a2b3c4-0001",
+      "field_key": "country_code",
+      "rule_type": "lookup",
+      "config": {
+        "table": "country_codes",
+        "source_field": "country",
+        "fallback": "llm"
+      },
+      "order": 1
+    },
+    {
+      "id": "r1a2b3c4-0002",
+      "field_key": "currency",
+      "rule_type": "lua",
+      "config": {
+        "script": "if cell.country_code == 'US' then return 'USD' end"
+      },
+      "order": 2
+    }
+  ]
+}`
+      },
+      { type: "heading", level: 2, id: "list-data-policy-rules-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Data policy not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "Get Data Policy", slug: "get-data-policy" },
+      { label: "List Data Policy Fields", slug: "list-data-policy-fields" },
+      { label: "Resolutions", slug: "list-resolutions" }
+    ],
+    faq: [
+      { question: "What rule types are available?", answer: "Data policies support 14 rule types: `lookup` (reference table matching with optional LLM fallback), `direct` (field-to-field mapping), `lua` (custom Lua scripting), `compute` (deterministic formulas), `format` (string formatting), `coalesce`, `constant`, `concat`, `split`, `regex`, `date_format`, `number_format`, `conditional`, and `aggregate`." },
+      { question: "How does the Lua scripting work?", answer: "Lua rules execute a sandboxed Lua 5.4 chunk for each record. The script receives a `cell` table with the current record values and can call chain methods like `:matches()` for reference table matching and `:llm_match()` for LLM-powered fuzzy matching. Scripts return the resolved value for the target field." },
+      { question: "What happens when a lookup fails to find a match?", answer: "Lookup rules support a 3-tier cascade: string normalization (exact match after lowercasing and trimming), token-based fuzzy matching (Jaccard similarity), and an optional LLM fallback (Haiku). If all tiers fail, the field retains its original value. Configure the `fallback` option in the rule config to control cascade behavior." }
+    ],
+    mentions: ["data policy", "transformation", "normalization", "lookup", "Lua"]
+  }
+];
+
+// src/content/api/record-sets.ts
+var sections52 = [
+  {
+    slug: "list-record-sets",
+    parentSlug: "record-sets",
+    title: "List Record Sets",
+    seoTitle: "List Record Sets Endpoint \u2014 Talonic Docs",
+    description: "List record sets across the value plane with cursor-based pagination. Filter by layer (capture, structured, resolved, product) to find sets at specific pipeline stages.",
+    content: [
+      { type: "paragraph", text: "Record sets are the core storage abstraction in the Talonic value plane. Each record set is a table-like collection of records at a specific layer of the pipeline. The value plane organizes data into four layers: **capture** (raw OCR output), **structured** (extracted field values), **resolved** (normalized canonical values), and **product** (final assembled output). Record sets at each layer share the same underlying cell storage model but represent progressively refined data." },
+      { type: "paragraph", text: "Use this endpoint to list all record sets in your workspace. Filter by `layer` to find sets at a specific pipeline stage, or by `source_type` to locate sets from a particular origin. Results support cursor-based pagination and can be sorted by creation date. Each record set contains typed cells with confidence scores and provenance metadata." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/record-sets",
+        summary: "List record sets with optional layer and source type filters.",
+        description: "Requires read scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Query parameters",
+            params: [
+              { name: "layer", type: "string", description: "Filter by value plane layer: capture, structured, resolved, or product." },
+              { name: "source_type", type: "string", description: "Filter by source type (e.g. extraction, resolution, job)." },
+              { name: "limit", type: "integer", default: "20", description: "Maximum number of items to return (1-100)." },
+              { name: "cursor", type: "string", description: "Opaque pagination cursor from a previous response." },
+              { name: "order", type: "string", default: "desc", description: "Sort order by creation date (asc | desc)." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "list-record-sets-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "data", type: "array", description: "Array of record set objects." },
+          { name: "data[].id", type: "string", description: "Record set UUID." },
+          { name: "data[].name", type: "string", description: "Human-readable record set name." },
+          { name: "data[].layer", type: "string", description: "Value plane layer: capture, structured, resolved, or product." },
+          { name: "data[].source_type", type: "string", description: "Origin type that created this record set." },
+          { name: "data[].source_id", type: "string", description: "UUID of the source entity (e.g. extraction run, job run)." },
+          { name: "data[].record_count", type: "integer", description: "Total number of records in this set." },
+          { name: "data[].field_count", type: "integer", description: "Number of fields defined on this set." },
+          { name: "data[].created_at", type: "string", description: "ISO 8601 creation timestamp." },
+          { name: "data[].updated_at", type: "string", description: "ISO 8601 last update timestamp." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s "https://api.talonic.ai/v1/record-sets?layer=resolved&limit=10" \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "data": [
+    {
+      "id": "rs-a1b2c3d4-e5f6-7890-abcd-ef1234567890",
+      "name": "Resolution Run 2024-10-15",
+      "layer": "resolved",
+      "source_type": "resolution",
+      "source_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
+      "record_count": 142,
+      "field_count": 12,
+      "created_at": "2024-10-15T11:30:00.000Z",
+      "updated_at": "2024-10-15T11:35:42.000Z"
+    }
+  ],
+  "cursor": "eyJpZCI6InJzLWExYjJjM2Q0In0="
+}`
+      },
+      { type: "heading", level: 2, id: "list-record-sets-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "Extractions", slug: "list-extractions" },
+      { label: "Jobs", slug: "list-jobs" },
+      { label: "Resolutions", slug: "list-resolutions" }
+    ],
+    faq: [
+      { question: "What are the value plane layers?", answer: "The value plane has four layers: **capture** (raw OCR/parsed text), **structured** (extracted field values from Claude), **resolved** (normalized canonical values from resolution), and **product** (final assembled output ready for delivery). Each layer represents a progressive refinement of the data." },
+      { question: "What is a record set?", answer: "A record set is a table-like collection of records at a specific value plane layer. It provides typed cell storage where each cell holds a value, confidence score, status, and provenance trace. Record sets are the primary read model for hot table operations in the platform." },
+      { question: "How do record sets relate to jobs and resolutions?", answer: "Job runs produce record sets at the **structured** layer (extracted values). Resolution runs consume structured record sets and produce new record sets at the **resolved** layer (normalized values). The `source_type` and `source_id` fields let you trace each record set back to its origin." }
+    ],
+    mentions: ["record set", "value plane", "cell", "provenance", "confidence"]
+  },
+  {
+    slug: "get-record-set",
+    parentSlug: "record-sets",
+    title: "Get Record Set",
+    seoTitle: "Get Record Set Endpoint \u2014 Talonic Docs",
+    description: "Retrieve a single record set by ID with its layer, source reference, record count, and field count. Requires read scope for the workspace.",
+    content: [
+      { type: "paragraph", text: "Retrieve the full metadata of a specific record set by its UUID. The response includes the value plane layer, source entity reference, record and field counts, and timestamps. Use this endpoint to inspect a record set before fetching its fields or records." },
+      { type: "paragraph", text: "The `layer` field tells you where this record set sits in the pipeline progression. The `source_type` and `source_id` fields let you trace the record set back to the extraction, job, or resolution run that created it. The `record_count` and `field_count` give you a quick summary of the dataset size without fetching the actual records." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/record-sets/{id}",
+        summary: "Get a single record set by ID.",
+        description: "Requires read scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Record set UUID." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "get-record-set-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "id", type: "string", description: "Record set UUID." },
+          { name: "name", type: "string", description: "Human-readable name." },
+          { name: "layer", type: "string", description: "Value plane layer: capture, structured, resolved, or product." },
+          { name: "source_type", type: "string", description: "Origin type that created this record set." },
+          { name: "source_id", type: "string", description: "UUID of the source entity." },
+          { name: "record_count", type: "integer", description: "Total number of records." },
+          { name: "field_count", type: "integer", description: "Number of defined fields." },
+          { name: "created_at", type: "string", description: "ISO 8601 creation timestamp." },
+          { name: "updated_at", type: "string", description: "ISO 8601 last update timestamp." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s https://api.talonic.ai/v1/record-sets/rs-a1b2c3d4-e5f6-7890-abcd-ef1234567890 \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "id": "rs-a1b2c3d4-e5f6-7890-abcd-ef1234567890",
+  "name": "Resolution Run 2024-10-15",
+  "layer": "resolved",
+  "source_type": "resolution",
+  "source_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
+  "record_count": 142,
+  "field_count": 12,
+  "created_at": "2024-10-15T11:30:00.000Z",
+  "updated_at": "2024-10-15T11:35:42.000Z"
+}`
+      },
+      { type: "heading", level: 2, id: "get-record-set-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Record set not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "List Record Sets", slug: "list-record-sets" },
+      { label: "List Record Set Fields", slug: "list-record-set-fields" },
+      { label: "List Record Set Records", slug: "list-record-set-records" }
+    ],
+    faq: [
+      { question: "How do I find which job or resolution produced a record set?", answer: 'The `source_type` field indicates the origin (e.g. "extraction", "job", "resolution") and the `source_id` is the UUID of that entity. Use the corresponding GET endpoint to retrieve the full source details.' },
+      { question: "What does record_count represent?", answer: "The `record_count` is the total number of records (rows) in the set. For structured and resolved layers, each record typically corresponds to one document. For the product layer, records correspond to assembled output rows." }
+    ],
+    mentions: ["record set", "value plane", "cell", "provenance", "confidence"]
+  },
+  {
+    slug: "list-record-set-fields",
+    parentSlug: "record-sets",
+    title: "List Record Set Fields",
+    seoTitle: "List Record Set Fields \u2014 Talonic Docs",
+    description: "List the field definitions of a record set including field keys, types, and display metadata. Fields define the columns available in the record set.",
+    content: [
+      { type: "paragraph", text: "Retrieve the field definitions for a record set. Fields define the columns available in the set \u2014 each field has a key (column name), a type, and optional display metadata. The field list determines which cell values can appear in each record and how they should be interpreted by downstream consumers." },
+      { type: "paragraph", text: "Field definitions are derived from the schema used during extraction or the policy used during resolution. They include the field key, value type, and optional attributes like display name and description. Use this endpoint to understand the shape of the data before fetching records, or to build dynamic table UIs that adapt to the field list." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/record-sets/{id}/fields",
+        summary: "List field definitions for a record set.",
+        description: "Requires read scope.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Record set UUID." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "list-record-set-fields-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "data", type: "array", description: "Array of field definition objects." },
+          { name: "data[].field_key", type: "string", description: "Field key (column name)." },
+          { name: "data[].type", type: "string", description: "Value type (e.g. string, number, date, boolean)." },
+          { name: "data[].display_name", type: "string | null", description: "Human-readable display name." },
+          { name: "data[].description", type: "string | null", description: "Optional field description." },
+          { name: "data[].ordinal", type: "integer", description: "Display order position." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s https://api.talonic.ai/v1/record-sets/rs-a1b2c3d4-e5f6-7890-abcd-ef1234567890/fields \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "data": [
+    { "field_key": "invoice_number", "type": "string", "display_name": "Invoice Number", "description": null, "ordinal": 0 },
+    { "field_key": "vendor_name", "type": "string", "display_name": "Vendor Name", "description": null, "ordinal": 1 },
+    { "field_key": "total_amount", "type": "number", "display_name": "Total Amount", "description": "Invoice total in local currency", "ordinal": 2 },
+    { "field_key": "invoice_date", "type": "date", "display_name": "Invoice Date", "description": null, "ordinal": 3 }
+  ]
+}`
+      },
+      { type: "heading", level: 2, id: "list-record-set-fields-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Record set not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "Get Record Set", slug: "get-record-set" },
+      { label: "List Record Set Records", slug: "list-record-set-records" },
+      { label: "Extractions", slug: "list-extractions" }
+    ],
+    faq: [
+      { question: "Where do the field definitions come from?", answer: "Field definitions are derived from the user schema (for structured layer sets) or the data policy output contract (for resolved layer sets). They represent the declared columns that the extraction or resolution pipeline was configured to produce." },
+      { question: "Can different record sets have different fields?", answer: "Yes. Each record set has its own independent field definitions. A structured record set may have different fields than a resolved record set, even if they originate from the same documents, because the resolution policy may add, rename, or transform fields." }
+    ],
+    mentions: ["record set", "value plane", "cell", "field definition"]
+  },
+  {
+    slug: "list-record-set-records",
+    parentSlug: "record-sets",
+    title: "List Record Set Records",
+    seoTitle: "List Record Set Records \u2014 Talonic Docs",
+    description: "List records in a record set with offset-based pagination. Each record contains typed cells with values, confidence scores, and provenance metadata.",
+    content: [
+      { type: "paragraph", text: "Retrieve the records (rows) in a record set with offset-based pagination. Each record contains a set of typed cells keyed by field name. Every cell carries a value, a confidence score (0-1), a status indicator, and provenance metadata tracing the value back to its source. This endpoint is the primary way to read structured data from the value plane." },
+      { type: "paragraph", text: "Unlike the cursor-based pagination used by most list endpoints, record set records use offset-based pagination with `page` and `limit` parameters. This is intentional: record sets are table-like structures where random access by page number is a common use case for building paginated table UIs. The total record count is available on the parent record set object." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/record-sets/{id}/records",
+        summary: "List records with typed cells, confidence, and provenance.",
+        description: "Requires read scope. Uses offset-based pagination.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Record set UUID." }
+            ]
+          },
+          {
+            type: "param-table",
+            title: "Query parameters",
+            params: [
+              { name: "page", type: "integer", default: "1", description: "Page number (1-indexed)." },
+              { name: "limit", type: "integer", default: "20", description: "Number of records per page (1-100)." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "list-record-set-records-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "data", type: "array", description: "Array of record objects." },
+          { name: "data[].id", type: "string", description: "Record UUID." },
+          { name: "data[].document_id", type: "string | null", description: "Associated document UUID, if applicable." },
+          { name: "data[].cells", type: "object", description: "Map of field_key to cell objects." },
+          { name: "data[].cells[key].value", type: "string | null", description: "The cell value." },
+          { name: "data[].cells[key].confidence", type: "number", description: "Confidence score (0-1)." },
+          { name: "data[].cells[key].status", type: "string", description: "Cell status (e.g. extracted, resolved, manual)." },
+          { name: "data[].cells[key].source", type: "string | null", description: "Provenance trace indicating how the value was produced." },
+          { name: "page", type: "integer", description: "Current page number." },
+          { name: "total", type: "integer", description: "Total number of records." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s "https://api.talonic.ai/v1/record-sets/rs-a1b2c3d4-e5f6-7890-abcd-ef1234567890/records?page=1&limit=10" \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "data": [
+    {
+      "id": "rec-f1e2d3c4-b5a6-7890-fedc-ba0987654321",
+      "document_id": "doc-a1b2c3d4-e5f6-7890-abcd-ef1234567890",
+      "cells": {
+        "invoice_number": {
+          "value": "INV-2024-0042",
+          "confidence": 0.97,
+          "status": "extracted",
+          "source": "chunk:3/line:12"
+        },
+        "country_code": {
+          "value": "DE",
+          "confidence": 0.95,
+          "status": "resolved",
+          "source": "lookup:country_codes"
+        }
+      }
+    }
+  ],
+  "page": 1,
+  "total": 142
+}`
+      },
+      { type: "heading", level: 2, id: "list-record-set-records-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Record set not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "Get Record Set", slug: "get-record-set" },
+      { label: "List Record Set Fields", slug: "list-record-set-fields" },
+      { label: "Export Record Set", slug: "export-record-set" }
+    ],
+    faq: [
+      { question: "What does the confidence score mean?", answer: "The confidence score (0-1) reflects how certain the system is about the cell value. Values extracted directly from documents with high textual clarity score 0.9+. Values resolved via fuzzy matching or LLM fallback score lower (0.5-0.9). Manually entered values always have confidence 1.0." },
+      { question: "What does the source field contain?", answer: "The `source` field is a provenance trace that indicates how the value was produced. For extracted values, it references the document chunk and line. For resolved values, it references the lookup table or resolution strategy. For computed values, it names the computation formula." },
+      { question: "Why does this endpoint use page/limit instead of cursor?", answer: "Record sets are table-like structures where random access by page number is common in table UIs. Offset pagination supports jumping to arbitrary pages, which cursor pagination does not. For sequential traversal of very large sets, use the export endpoint instead." }
+    ],
+    mentions: ["record set", "value plane", "cell", "provenance", "confidence"]
+  },
+  {
+    slug: "export-record-set",
+    parentSlug: "record-sets",
+    title: "Export Record Set",
+    seoTitle: "Export Record Set Endpoint \u2014 Talonic Docs",
+    description: "Export a complete record set as a downloadable file. Returns all records with their cell values in a single response for bulk consumption.",
+    content: [
+      { type: "paragraph", text: "Export the complete contents of a record set as a downloadable payload. Unlike the paginated records endpoint, this returns all records in a single response, making it suitable for bulk data consumption, ETL pipelines, and offline analysis. The export includes all cell values for every record in the set." },
+      { type: "paragraph", text: "The export endpoint streams the full record set content without pagination. For large record sets, this may result in significant response payloads. Consider using the paginated records endpoint for interactive use cases, and reserve this endpoint for batch export workflows where you need the complete dataset in one call." },
+      {
+        type: "endpoint",
+        method: "GET",
+        path: "/v1/record-sets/{id}/export",
+        summary: "Export a complete record set.",
+        description: "Requires read scope. Returns all records without pagination.",
+        blocks: [
+          {
+            type: "param-table",
+            title: "Path parameters",
+            params: [
+              { name: "id", type: "uuid", required: true, description: "Record set UUID." }
+            ]
+          }
+        ]
+      },
+      { type: "heading", level: 2, id: "export-record-set-response", text: "Response" },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          { name: "data", type: "array", description: "Complete array of record objects with cells." },
+          { name: "fields", type: "array", description: "Array of field definitions for column ordering." },
+          { name: "record_count", type: "integer", description: "Total number of exported records." },
+          { name: "exported_at", type: "string", description: "ISO 8601 timestamp of the export." }
+        ]
+      },
+      {
+        type: "code",
+        title: "curl",
+        language: "bash",
+        code: `curl -s https://api.talonic.ai/v1/record-sets/rs-a1b2c3d4-e5f6-7890-abcd-ef1234567890/export \\
+  -H "Authorization: Bearer tlnc_your_api_key"`
+      },
+      {
+        type: "code",
+        title: "Response",
+        code: `{
+  "data": [
+    {
+      "id": "rec-f1e2d3c4-b5a6-7890-fedc-ba0987654321",
+      "document_id": "doc-a1b2c3d4-e5f6-7890-abcd-ef1234567890",
+      "cells": {
+        "invoice_number": { "value": "INV-2024-0042", "confidence": 0.97 },
+        "country_code": { "value": "DE", "confidence": 0.95 }
+      }
+    }
+  ],
+  "fields": [
+    { "field_key": "invoice_number", "type": "string", "ordinal": 0 },
+    { "field_key": "country_code", "type": "string", "ordinal": 1 }
+  ],
+  "record_count": 142,
+  "exported_at": "2024-10-16T09:00:00.000Z"
+}`
+      },
+      { type: "heading", level: 2, id: "export-record-set-errors", text: "Errors" },
+      {
+        type: "param-table",
+        title: "Error responses",
+        params: [
+          { name: "401", type: "unauthorized", description: "Missing or invalid API key." },
+          { name: "404", type: "not_found", description: "Record set not found or does not belong to your organization." },
+          { name: "429", type: "rate_limited", description: "Too many requests. Retry after the period indicated in the Retry-After header." }
+        ]
+      }
+    ],
+    related: [
+      { label: "List Record Set Records", slug: "list-record-set-records" },
+      { label: "Extractions", slug: "list-extractions" },
+      { label: "Jobs", slug: "list-jobs" },
+      { label: "Resolutions", slug: "list-resolutions" }
+    ],
+    faq: [
+      { question: "When should I use export vs the paginated records endpoint?", answer: "Use the export endpoint for batch workflows, ETL pipelines, and offline analysis where you need the complete dataset in one call. Use the paginated records endpoint for interactive UIs, incremental processing, or when working with very large record sets where memory is a concern." },
+      { question: "Does the export include confidence and provenance?", answer: "The export includes cell values and confidence scores. Full provenance traces are available through the paginated records endpoint. The export is optimized for bulk data consumption rather than detailed audit trails." },
+      { question: "Is there a size limit on exports?", answer: "There is no hard limit, but very large record sets (10,000+ records) may result in large response payloads and slower response times. For extremely large datasets, consider using the delivery pipeline to push data to an S3 bucket or SFTP server instead." }
+    ],
+    mentions: ["record set", "value plane", "cell", "export", "bulk"]
+  }
+];
+
+// src/content/sdk/sections.json
+var sections_default = [
+  {
+    slug: "sdk-introduction",
+    parentSlug: "sdk-overview",
+    title: "Introduction",
+    seoTitle: "Node SDK Introduction \u2014 Talonic Docs",
+    description: "Official Talonic SDK for Node.js and TypeScript. Extract structured, schema-validated data from any document with a single function call.",
+    content: [
+      { type: "paragraph", text: "The `@talonic/node` SDK is the official Node.js and TypeScript client for the Talonic API. Extract structured, schema-validated data from any document with a single function call." },
+      { type: "paragraph", text: "Zero runtime dependencies. Requires Node.js 18 or newer." },
+      { type: "callout", text: "Looking for the AI agent path? [`@talonic/mcp`](https://github.com/talonicdev/talonic-mcp) wraps this SDK as a Model Context Protocol server. Install it into Claude Desktop, Cursor, Cline, Continue, or Cowork and any MCP-aware agent can extract documents directly." }
+    ],
+    related: [
+      { label: "Installation", slug: "sdk-installation" },
+      { label: "Quick Start", slug: "sdk-quickstart" },
+      { label: "MCP Server", slug: "mcp-introduction" }
+    ],
+    faq: [
+      { question: "What is the Talonic Node SDK?", answer: "The official Node.js and TypeScript client for extracting structured, schema-validated data from documents via the Talonic API." },
+      { question: "What Node.js version is required?", answer: "Node.js 18 or newer. The SDK has zero runtime dependencies." }
+    ],
+    mentions: ["Node.js", "TypeScript", "SDK", "npm", "document extraction"]
+  },
+  {
+    slug: "sdk-installation",
+    parentSlug: "sdk-overview",
+    title: "Installation",
+    seoTitle: "Install Talonic Node SDK \u2014 Talonic Docs",
+    description: "Install the @talonic/node package via npm. Zero runtime dependencies, requires Node.js 18+.",
+    content: [
+      { type: "code", language: "bash", title: "Install via npm", code: "npm install @talonic/node" },
+      { type: "paragraph", text: "Requires Node.js 18 or newer. Zero runtime dependencies." }
+    ],
+    related: [
+      { label: "Authentication", slug: "sdk-authentication" },
+      { label: "Quick Start", slug: "sdk-quickstart" }
+    ],
+    faq: [
+      { question: "How do I install the Talonic Node SDK?", answer: "Run npm install @talonic/node. Requires Node.js 18 or newer. Zero runtime dependencies." }
+    ],
+    mentions: ["npm", "Node.js"]
+  },
+  {
+    slug: "sdk-authentication",
+    parentSlug: "sdk-overview",
+    title: "Authentication",
+    seoTitle: "SDK Authentication \u2014 Talonic Docs",
+    description: "Get a Talonic API key and configure it for the Node SDK. Each workspace is isolated with private documents and schemas.",
+    content: [
+      { type: "paragraph", text: "Every user runs against their own Talonic workspace, so each user needs their own key. Workspaces are isolated; your documents and schemas are private to you." },
+      { type: "list", ordered: true, items: [
+        "Sign up at [https://app.talonic.com](https://app.talonic.com). Free tier: 50 extractions per day, no credit card.",
+        "Settings → API Keys → Create New Key.",
+        "Copy the `tlnc_` value.",
+        "Set it as the `TALONIC_API_KEY` environment variable, or pass it directly to the client constructor."
+      ] },
+      { type: "callout", text: "Keep your API key secret. Do not expose it in client-side code or version control." }
+    ],
+    related: [
+      { label: "Configuration", slug: "sdk-configuration" },
+      { label: "API Authentication", slug: "authentication" }
+    ],
+    faq: [
+      { question: "Where do I get a Talonic API key?", answer: "Sign up at app.talonic.com, go to Settings \u2192 API Keys \u2192 Create New Key. Free tier includes 50 extractions per day." }
+    ],
+    mentions: ["API key", "authentication", "workspace"]
+  },
+  {
+    slug: "sdk-quickstart",
+    parentSlug: "sdk-overview",
+    title: "Quick Start",
+    seoTitle: "Node SDK Quick Start \u2014 Talonic Docs",
+    description: "Extract structured data from a document in five lines of TypeScript using the Talonic Node SDK.",
+    content: [
+      { type: "code", language: "typescript", title: "Extract an invoice", code: 'import { Talonic } from "@talonic/node"\n\nconst talonic = new Talonic({ apiKey: process.env.TALONIC_API_KEY! })\n\nconst result = await talonic.extract({\n  file_path: "./invoice.pdf",\n  schema: {\n    vendor_name: "string",\n    invoice_number: "string",\n    total_amount: "number",\n    due_date: "date",\n  },\n})\n\nconsole.log(result.data)\n// { vendor_name: "Acme Corp", invoice_number: "INV-2024-0847", total_amount: 14250, due_date: "2024-03-15" }' }
+    ],
+    related: [
+      { label: "Extract", slug: "sdk-extract" },
+      { label: "Configuration", slug: "sdk-configuration" },
+      { label: "Schemas", slug: "sdk-schemas" }
+    ],
+    faq: [
+      { question: "How do I extract data from a document with the SDK?", answer: "Call talonic.extract() with a file_path and a schema defining the fields you want. Returns structured JSON with confidence scores." }
+    ],
+    mentions: ["extract", "schema", "TypeScript", "quickstart"]
+  },
+  {
+    slug: "sdk-configuration",
+    parentSlug: "sdk-client",
+    title: "Configuration",
+    seoTitle: "SDK Configuration \u2014 Talonic Docs",
+    description: "Configure the Talonic client with API key, base URL, timeout, max retries, and custom fetch function.",
     content: [
       { type: "code", language: "typescript", title: "Client configuration", code: 'const talonic = new Talonic({\n  apiKey: process.env.TALONIC_API_KEY!,\n  baseUrl: "https://api.talonic.com",   // default\n  timeout: 60_000,                       // ms; default 60s\n  maxRetries: 3,                         // 429, 500, 502, 503, 504, network, timeout\n  fetch: customFetch,                    // optional override (e.g. for testing)\n})' },
       { type: "param-table", title: "Constructor options", params: [
@@ -26174,7 +27691,7 @@ talonic --help` }
 ];
 
 // src/content/sdk/index.ts
-var sections50 = sections_default;
+var sections53 = sections_default;
 
 // src/content/mcp/sections.json
 var sections_default2 = [
@@ -26183,11 +27700,11 @@ var sections_default2 = [
     parentSlug: "mcp-overview",
     title: "Introduction",
     seoTitle: "MCP Server Introduction \u2014 Talonic Docs",
-    description: "Official Talonic MCP server for AI agents. Eight tools and two resources for structured document extraction via the Model Context Protocol.",
+    description: "Official Talonic MCP server for AI agents. Nine tools and two resources for structured document extraction via the Model Context Protocol.",
     content: [
       {
         type: "paragraph",
-        text: "The `@talonic/mcp` package is the official Talonic MCP server. It gives AI agents eight tools and two resources for extracting structured, schema-validated data from any document via the [Model Context Protocol](https://modelcontextprotocol.io)."
+        text: "The `@talonic/mcp` package is the official Talonic MCP server. It gives AI agents nine tools and two resources for extracting structured, schema-validated data from any document via the [Model Context Protocol](https://modelcontextprotocol.io)."
       },
       {
         type: "paragraph",
@@ -26195,7 +27712,7 @@ var sections_default2 = [
       },
       {
         type: "paragraph",
-        text: "With this MCP server installed, the agent has a `talonic_extract` tool that returns schema-validated JSON with per-field confidence scores, a detected document type, and stable IDs for follow-up calls. The other seven tools cover the rest of the workflow: searching the workspace, filtering by extracted field values, fetching document metadata, getting OCR markdown, listing saved schemas, saving new ones, and reading the workspace credit balance for budget-aware behaviour."
+        text: "With this MCP server installed, the agent has a `talonic_extract` tool that returns schema-validated JSON with per-field confidence scores, a detected document type, and stable IDs for follow-up calls. The other eight tools cover the rest of the workflow: searching the workspace, filtering by extracted field values, fetching document metadata, getting OCR markdown, listing saved schemas, saving new ones, and reading the workspace credit balance for budget-aware behaviour."
       },
       {
         type: "callout",
@@ -26203,17 +27720,31 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "Installation", slug: "mcp-installation" },
-      { label: "Tools", slug: "mcp-talonic-extract" },
-      { label: "Node SDK", slug: "sdk-introduction" }
+      {
+        label: "Installation",
+        slug: "mcp-installation"
+      },
+      {
+        label: "Tools",
+        slug: "mcp-talonic-extract"
+      },
+      {
+        label: "Node SDK",
+        slug: "sdk-introduction"
+      }
     ],
     faq: [
       {
         question: "What is the Talonic MCP server?",
-        answer: "An official Model Context Protocol server that gives AI agents eight tools for document extraction, search, filtering, schema management, and credit-balance lookup via the Talonic API."
+        answer: "An official Model Context Protocol server that gives AI agents nine tools for document extraction, search, filtering, schema management, and credit-balance lookup via the Talonic API."
       }
     ],
-    mentions: ["MCP", "Model Context Protocol", "AI agents", "document extraction"]
+    mentions: [
+      "MCP",
+      "Model Context Protocol",
+      "AI agents",
+      "document extraction"
+    ]
   },
   {
     slug: "mcp-installation",
@@ -26226,6 +27757,10 @@ var sections_default2 = [
         type: "paragraph",
         text: "Three install paths. Pick the one that matches your client."
       },
+      {
+        type: "callout",
+        text: "The hosted MCP endpoint accepts both `https://mcp.talonic.com/mcp` and the bare origin `https://mcp.talonic.com` as connector URLs \u2014 POST/DELETE/SSE traffic at either path routes through the same Streamable HTTP transport. Plain `GET /` still returns the discovery JSON. Either URL works in any of the install paths below."
+      },
       {
         type: "heading",
         level: 3,
@@ -26252,7 +27787,11 @@ var sections_default2 = [
         type: "callout",
         text: "The connector does not need an API key in its config. Token rotation is handled by the OAuth flow."
       },
-      { type: "heading", level: 3, text: "Local stdio (npx)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Local stdio (npx)"
+      },
       {
         type: "paragraph",
         text: "Recommended for IDE-style clients (Claude Desktop, Cursor, Cline, Continue, Cowork). Runs on your machine via stdio; requires Node.js 18 or later. Uses a `TALONIC_API_KEY` from `app.talonic.com`."
@@ -26288,9 +27827,18 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "Claude Desktop", slug: "mcp-claude-desktop" },
-      { label: "Cursor", slug: "mcp-cursor" },
-      { label: "Authentication", slug: "mcp-authentication" }
+      {
+        label: "Claude Desktop",
+        slug: "mcp-claude-desktop"
+      },
+      {
+        label: "Cursor",
+        slug: "mcp-cursor"
+      },
+      {
+        label: "Authentication",
+        slug: "mcp-authentication"
+      }
     ],
     faq: [
       {
@@ -26324,7 +27872,11 @@ var sections_default2 = [
         type: "paragraph",
         text: "Each user runs against their own isolated Talonic workspace. Your documents and schemas are private to you. There are three authentication paths depending on how you connect."
       },
-      { type: "heading", level: 3, text: "OAuth 2.1 (Claude.ai connector, recommended)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "OAuth 2.1 (Claude.ai connector, recommended)"
+      },
       {
         type: "paragraph",
         text: "When you connect via Claude.ai's custom-connector flow, the connector launches an OAuth 2.1 sign-in to `app.talonic.com` (PKCE + Dynamic Client Registration). After consent, the connector exchanges a short-lived bearer token that is rotated automatically; no API key sits in the connector config. The Talonic MCP server validates the token on each request against the API, so revocation propagates immediately."
@@ -26355,7 +27907,11 @@ var sections_default2 = [
           "`?apiKey=tlnc_...` query parameter (only for clients that cannot set custom headers)."
         ]
       },
-      { type: "heading", level: 3, text: "Environment variable (local stdio)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Environment variable (local stdio)"
+      },
       {
         type: "paragraph",
         text: "Set `TALONIC_API_KEY` in the `env` block of your MCP client config. The local server reads it at startup."
@@ -26371,8 +27927,14 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "Installation", slug: "mcp-installation" },
-      { label: "API Authentication", slug: "authentication" }
+      {
+        label: "Installation",
+        slug: "mcp-installation"
+      },
+      {
+        label: "API Authentication",
+        slug: "authentication"
+      }
     ],
     faq: [
       {
@@ -26401,14 +27963,22 @@ var sections_default2 = [
         type: "paragraph",
         text: "Edit `~/Library/Application Support/Claude/claude_desktop_config.json` (macOS) or `%APPDATA%\\Claude\\claude_desktop_config.json` (Windows)."
       },
-      { type: "heading", level: 3, text: "Hosted (recommended)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Hosted (recommended)"
+      },
       {
         type: "code",
         language: "json",
         title: "claude_desktop_config.json",
         code: '{\n  "mcpServers": {\n    "talonic": {\n      "url": "https://mcp.talonic.com/mcp",\n      "headers": {\n        "Authorization": "Bearer tlnc_your_key_here"\n      }\n    }\n  }\n}'
       },
-      { type: "heading", level: 3, text: "Local (npx)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Local (npx)"
+      },
       {
         type: "code",
         language: "json",
@@ -26421,8 +27991,14 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "Cursor", slug: "mcp-cursor" },
-      { label: "Tool Reference", slug: "mcp-talonic-extract" }
+      {
+        label: "Cursor",
+        slug: "mcp-cursor"
+      },
+      {
+        label: "Tool Reference",
+        slug: "mcp-talonic-extract"
+      }
     ],
     faq: [
       {
@@ -26430,7 +28006,11 @@ var sections_default2 = [
         answer: "Edit claude_desktop_config.json, add the Talonic MCP server config (hosted URL or local npx) with your API key, and fully restart Claude Desktop (Cmd+Q on macOS)."
       }
     ],
-    mentions: ["Claude Desktop", "macOS", "Windows"]
+    mentions: [
+      "Claude Desktop",
+      "macOS",
+      "Windows"
+    ]
   },
   {
     slug: "mcp-cursor",
@@ -26443,14 +28023,22 @@ var sections_default2 = [
         type: "paragraph",
         text: "Edit `~/.cursor/mcp.json` (or open Cursor settings \u2192 MCP \u2192 edit config):"
       },
-      { type: "heading", level: 3, text: "Hosted (recommended)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Hosted (recommended)"
+      },
       {
         type: "code",
         language: "json",
         title: "~/.cursor/mcp.json",
         code: '{\n  "mcpServers": {\n    "talonic": {\n      "url": "https://mcp.talonic.com/mcp",\n      "headers": {\n        "Authorization": "Bearer tlnc_your_key_here"\n      }\n    }\n  }\n}'
       },
-      { type: "heading", level: 3, text: "Local (npx)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Local (npx)"
+      },
       {
         type: "code",
         language: "json",
@@ -26459,8 +28047,14 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "Claude Desktop", slug: "mcp-claude-desktop" },
-      { label: "Cline", slug: "mcp-cline" }
+      {
+        label: "Claude Desktop",
+        slug: "mcp-claude-desktop"
+      },
+      {
+        label: "Cline",
+        slug: "mcp-cline"
+      }
     ],
     faq: [
       {
@@ -26468,7 +28062,10 @@ var sections_default2 = [
         answer: "Edit ~/.cursor/mcp.json and add the Talonic MCP server config with your API key. Hosted or local."
       }
     ],
-    mentions: ["Cursor", "IDE"]
+    mentions: [
+      "Cursor",
+      "IDE"
+    ]
   },
   {
     slug: "mcp-cline",
@@ -26481,23 +28078,40 @@ var sections_default2 = [
         type: "paragraph",
         text: "Open the Cline panel \u2192 settings (gear icon) \u2192 MCP Servers \u2192 Edit."
       },
-      { type: "heading", level: 3, text: "Hosted (recommended)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Hosted (recommended)"
+      },
       {
         type: "code",
         language: "json",
         code: '{\n  "mcpServers": {\n    "talonic": {\n      "url": "https://mcp.talonic.com/mcp",\n      "headers": {\n        "Authorization": "Bearer tlnc_your_key_here"\n      }\n    }\n  }\n}'
       },
-      { type: "heading", level: 3, text: "Local (npx)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Local (npx)"
+      },
       {
         type: "code",
         language: "json",
         code: '{\n  "mcpServers": {\n    "talonic": {\n      "command": "npx",\n      "args": ["-y", "@talonic/mcp@latest"],\n      "env": {\n        "TALONIC_API_KEY": "tlnc_your_key_here"\n      }\n    }\n  }\n}'
       },
-      { type: "paragraph", text: "Save and restart the panel." }
+      {
+        type: "paragraph",
+        text: "Save and restart the panel."
+      }
     ],
     related: [
-      { label: "Continue", slug: "mcp-continue" },
-      { label: "Cursor", slug: "mcp-cursor" }
+      {
+        label: "Continue",
+        slug: "mcp-continue"
+      },
+      {
+        label: "Cursor",
+        slug: "mcp-cursor"
+      }
     ],
     faq: [
       {
@@ -26505,7 +28119,10 @@ var sections_default2 = [
         answer: "Open the Cline panel settings, go to MCP Servers, click Edit, and add the Talonic config entry."
       }
     ],
-    mentions: ["Cline", "VS Code"]
+    mentions: [
+      "Cline",
+      "VS Code"
+    ]
   },
   {
     slug: "mcp-continue",
@@ -26518,14 +28135,22 @@ var sections_default2 = [
         type: "paragraph",
         text: "Edit `~/.continue/config.json`. Add to the `mcpServers` array:"
       },
-      { type: "heading", level: 3, text: "Hosted (recommended)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Hosted (recommended)"
+      },
       {
         type: "code",
         language: "json",
         title: "~/.continue/config.json",
         code: '{\n  "name": "talonic",\n  "url": "https://mcp.talonic.com/mcp",\n  "headers": {\n    "Authorization": "Bearer tlnc_your_key_here"\n  }\n}'
       },
-      { type: "heading", level: 3, text: "Local (npx)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Local (npx)"
+      },
       {
         type: "code",
         language: "json",
@@ -26534,8 +28159,14 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "Cowork", slug: "mcp-cowork" },
-      { label: "Cline", slug: "mcp-cline" }
+      {
+        label: "Cowork",
+        slug: "mcp-cowork"
+      },
+      {
+        label: "Cline",
+        slug: "mcp-cline"
+      }
     ],
     faq: [
       {
@@ -26543,7 +28174,11 @@ var sections_default2 = [
         answer: "Edit ~/.continue/config.json and add a Talonic entry to the mcpServers array with your API key."
       }
     ],
-    mentions: ["Continue", "VS Code", "JetBrains"]
+    mentions: [
+      "Continue",
+      "VS Code",
+      "JetBrains"
+    ]
   },
   {
     slug: "mcp-cowork",
@@ -26552,14 +28187,25 @@ var sections_default2 = [
     seoTitle: "MCP Setup for Cowork \u2014 Talonic Docs",
     description: "Configure the Talonic MCP server in Cowork. Hosted and local configs.",
     content: [
-      { type: "paragraph", text: "Open Cowork settings \u2192 MCP Servers \u2192 Add." },
-      { type: "heading", level: 3, text: "Hosted (recommended)" },
+      {
+        type: "paragraph",
+        text: "Open Cowork settings \u2192 MCP Servers \u2192 Add."
+      },
+      {
+        type: "heading",
+        level: 3,
+        text: "Hosted (recommended)"
+      },
       {
         type: "code",
         language: "json",
         code: '{\n  "mcpServers": {\n    "talonic": {\n      "url": "https://mcp.talonic.com/mcp",\n      "headers": {\n        "Authorization": "Bearer tlnc_your_key_here"\n      }\n    }\n  }\n}'
       },
-      { type: "heading", level: 3, text: "Local (npx)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Local (npx)"
+      },
       {
         type: "code",
         language: "json",
@@ -26567,8 +28213,14 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "Claude Desktop", slug: "mcp-claude-desktop" },
-      { label: "Tool Reference", slug: "mcp-talonic-extract" }
+      {
+        label: "Claude Desktop",
+        slug: "mcp-claude-desktop"
+      },
+      {
+        label: "Tool Reference",
+        slug: "mcp-talonic-extract"
+      }
     ],
     faq: [
       {
@@ -26576,7 +28228,9 @@ var sections_default2 = [
         answer: "Open Cowork settings, go to MCP Servers, click Add, and paste the standard Talonic config with your API key."
       }
     ],
-    mentions: ["Cowork"]
+    mentions: [
+      "Cowork"
+    ]
   },
   {
     slug: "mcp-talonic-extract",
@@ -26589,7 +28243,11 @@ var sections_default2 = [
         type: "paragraph",
         text: "Extract structured, schema-validated data from a document. Returns clean JSON matching the schema, with per-field confidence scores and document metadata."
       },
-      { type: "heading", level: 3, text: "When to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -26600,7 +28258,11 @@ var sections_default2 = [
           "You want validated JSON instead of trying to OCR + parse with raw LLM calls."
         ]
       },
-      { type: "heading", level: 3, text: "When NOT to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When NOT to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -26609,7 +28271,11 @@ var sections_default2 = [
           "The user wants to find documents matching a query \u2192 use `talonic_search` or `talonic_filter`."
         ]
       },
-      { type: "heading", level: 3, text: "Input schema" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Input schema"
+      },
       {
         type: "param-table",
         title: "File source (provide exactly one)",
@@ -26678,14 +28344,22 @@ var sections_default2 = [
           }
         ]
       },
-      { type: "heading", level: 3, text: "Response shape" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Response shape"
+      },
       {
         type: "code",
         language: "json",
         title: "Example response",
         code: '{\n  "data": {\n    "vendor_name": "Acme Corp",\n    "invoice_number": "INV-2024-0847",\n    "total_amount": 14250.00,\n    "due_date": "2024-03-15"\n  },\n  "confidence": {\n    "vendor_name": 0.97,\n    "invoice_number": 0.99,\n    "total_amount": 0.94,\n    "due_date": 0.91\n  },\n  "document": {\n    "id": "d_abc123",\n    "filename": "invoice.pdf",\n    "documentType": "invoice",\n    "language": "en",\n    "pageCount": 2\n  },\n  "extraction": {\n    "id": "ext_xyz789",\n    "schemaId": "sch_def456"\n  },\n  "cost": {\n    "costCredits": 1,\n    "costEur": 0.05,\n    "balanceCredits": 999,\n    "cellsResolvedRegistry": 3,\n    "cellsResolvedAi": 1\n  }\n}'
       },
-      { type: "heading", level: 3, text: "Confidence scores and human escalation" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Confidence scores and human escalation"
+      },
       {
         type: "paragraph",
         text: "Each field in the `confidence` object is a float from 0.0 to 1.0. Values above **0.90** are high confidence. Values between **0.70\u20130.90** should be treated with caution \u2014 flag them to the user for verification. Values below **0.70** indicate low confidence \u2014 the agent should ask the user to verify the value or re-extract with more specific instructions."
@@ -26695,17 +28369,29 @@ var sections_default2 = [
         variant: "warning",
         text: "Always provide either a `schema` or `schema_id`. The MCP layer rejects schema-less calls with a validation error before they reach the API."
       },
-      { type: "heading", level: 3, text: "Cost" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Cost"
+      },
       {
         type: "paragraph",
         text: "Each `talonic_extract` call with a new file consumes **one extraction credit**. Re-extracting the same `document_id` with a different schema also consumes one credit. The per-call cost is surfaced on the response under `cost` (`costCredits`, `costEur`, `balanceCredits`, plus a breakdown of how many cells were resolved by the registry vs the AI), parsed from the `X-Talonic-Cost-*` and `X-Talonic-Balance-*` response headers. To avoid unnecessary cost, check if a document has already been extracted before calling again \u2014 use `talonic_search` or `talonic_filter` to find existing results, and `talonic_get_balance` to check your runway before kicking off a large batch."
       },
-      { type: "heading", level: 3, text: "Errors" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Errors"
+      },
       {
         type: "param-table",
         title: "Common errors",
         params: [
-          { name: "unauthorized", type: "401", description: "Invalid or missing API key." },
+          {
+            name: "unauthorized",
+            type: "401",
+            description: "Invalid or missing API key."
+          },
           {
             name: "validation_error",
             type: "422",
@@ -26730,9 +28416,18 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "SDK Extract", slug: "sdk-extract" },
-      { label: "POST /v1/extract", slug: "post-extract" },
-      { label: "Cost & Rate Limits", slug: "mcp-cost-and-limits" }
+      {
+        label: "SDK Extract",
+        slug: "sdk-extract"
+      },
+      {
+        label: "POST /v1/extract",
+        slug: "post-extract"
+      },
+      {
+        label: "Cost & Rate Limits",
+        slug: "mcp-cost-and-limits"
+      }
     ],
     faq: [
       {
@@ -26744,7 +28439,13 @@ var sections_default2 = [
         answer: "Scores above 0.90 are reliable. Between 0.70\u20130.90, flag to the user for verification. Below 0.70, ask the user to verify or re-extract with more specific instructions."
       }
     ],
-    mentions: ["talonic_extract", "file_data", "schema", "confidence", "extraction"]
+    mentions: [
+      "talonic_extract",
+      "file_data",
+      "schema",
+      "confidence",
+      "extraction"
+    ]
   },
   {
     slug: "mcp-talonic-search",
@@ -26757,7 +28458,11 @@ var sections_default2 = [
         type: "paragraph",
         text: "Omnisearch across documents, extracted field values, field names, sources, and schemas in the workspace. Returns ranked results across all entity types in one call."
       },
-      { type: "heading", level: 3, text: "When to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -26769,7 +28474,11 @@ var sections_default2 = [
           "You need to discover canonical field names before using `talonic_filter`."
         ]
       },
-      { type: "heading", level: 3, text: "When NOT to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When NOT to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -26779,7 +28488,11 @@ var sections_default2 = [
           "The user wants to extract data from a new document \u2192 use `talonic_extract`."
         ]
       },
-      { type: "heading", level: 3, text: "Input schema" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Input schema"
+      },
       {
         type: "param-table",
         title: "Parameters",
@@ -26797,28 +28510,48 @@ var sections_default2 = [
           }
         ]
       },
-      { type: "heading", level: 3, text: "Response shape" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Response shape"
+      },
       {
         type: "code",
         language: "json",
         title: "Example response",
-        code: '{\n  "documents": [\n    {\n      "id": "d_abc123",\n      "filename": "acme-invoice-q4.pdf",\n      "documentType": "invoice",\n      "score": 0.92\n    }\n  ],\n  "fieldMatches": [\n    {\n      "resolvedFieldId": "f_ghi789",\n      "displayName": "Vendor Name",\n      "matchedValue": "Acme Corp",\n      "documentCount": 3,\n      "filterable": true\n    }\n  ],\n  "sources": [],\n  "schemas": [\n    {\n      "id": "sch_def456",\n      "name": "Standard Invoice"\n    }\n  ],\n  "fields": [\n    {\n      "id": "f_ghi789",\n      "canonicalName": "vendor.name",\n      "displayName": "Vendor Name",\n      "documentCount": 12,\n      "filterable": true\n    }\n  ]\n}'
+        code: '{\n  "documents": [\n    {\n      "id": "d_abc123",\n      "filename": "acme-invoice-q4.pdf",\n      "documentType": "invoice",\n      "score": 0.92\n    }\n  ],\n  "fieldMatches": [\n    {\n      "resolvedFieldId": "f_ghi789",\n      "displayName": "Vendor Name",\n      "matchedValue": "Acme Corp",\n      "documentCount": 3,\n      "filterable": true,\n      "dataType": "string"\n    }\n  ],\n  "sources": [],\n  "schemas": [\n    {\n      "id": "sch_def456",\n      "name": "Standard Invoice"\n    }\n  ],\n  "fields": [\n    {\n      "id": "f_total",\n      "canonicalName": "total_amount",\n      "displayName": "Total Amount",\n      "documentCount": 14,\n      "filterable": true,\n      "dataType": "number"\n    }\n  ]\n}'
       },
       {
         type: "callout",
         text: "Only `fields[]` entries with `filterable: true` can be used with `talonic_filter`. These have extracted data in the workspace. Fields with `filterable: false` exist in a schema definition but have no extracted data yet \u2014 they become filterable after documents are processed against their schema."
       },
-      { type: "heading", level: 3, text: "Cost" },
+      {
+        type: "callout",
+        text: 'Every `fieldMatches[]` and `fields[]` entry carries a `dataType` (`"string"`, `"number"`, `"array"`, etc.). Use it to pick the right `talonic_filter` operator on the first call \u2014 numeric operators (`gt`, `gte`, `lt`, `lte`, `between`) only resolve correctly when `dataType === "number"`. See the *Schema typing* section under `talonic_filter` for the full preventive / reactive pattern.'
+      },
+      {
+        type: "heading",
+        level: 3,
+        text: "Cost"
+      },
       {
         type: "paragraph",
         text: "Search calls are **free** \u2014 they do not consume extraction credits. Use search liberally to explore before extracting."
       },
-      { type: "heading", level: 3, text: "Errors" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Errors"
+      },
       {
         type: "param-table",
         title: "Common errors",
         params: [
-          { name: "unauthorized", type: "401", description: "Invalid or missing API key." },
+          {
+            name: "unauthorized",
+            type: "401",
+            description: "Invalid or missing API key."
+          },
           {
             name: "validation_error",
             type: "422",
@@ -26828,8 +28561,14 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "talonic_filter", slug: "mcp-talonic-filter" },
-      { label: "Omnisearch", slug: "omnisearch" }
+      {
+        label: "talonic_filter",
+        slug: "mcp-talonic-filter"
+      },
+      {
+        label: "Omnisearch",
+        slug: "omnisearch"
+      }
     ],
     faq: [
       {
@@ -26839,22 +28578,38 @@ var sections_default2 = [
       {
         question: "What entities does talonic_search return?",
         answer: "Documents, field matches (with canonical names and values), sources, schemas, and field definitions \u2014 all ranked by relevance score."
+      },
+      {
+        question: "How do I avoid a `talonic_filter` numeric query returning zero matches?",
+        answer: 'Check `dataType` on the field entry in the search response before constructing the filter. Numeric operators (`gt`, `gte`, `lt`, `lte`, `between`) only resolve against fields where `dataType === "number"`. If the type is `string` (common for monetary or formatted-number fields), suggest the user change the field\'s data type in the schema before filtering.'
       }
     ],
-    mentions: ["talonic_search", "omnisearch", "canonicalName", "field discovery"]
+    mentions: [
+      "canonicalName",
+      "dataType",
+      "field discovery",
+      "omnisearch",
+      "preventive guard",
+      "schema typing",
+      "talonic_search"
+    ]
   },
   {
     slug: "mcp-talonic-filter",
     parentSlug: "mcp-tools",
     title: "talonic_filter",
     seoTitle: "talonic_filter MCP Tool \u2014 Talonic Docs",
-    description: "Filter documents by extracted field values. Full operator reference, input/output schema, and composable condition examples.",
+    description: "Filter documents by extracted field values. Full operator reference, input/output schema, composable condition examples, and the preventive + reactive pattern for guarding numeric operators against string-typed fields.",
     content: [
       {
         type: "paragraph",
         text: "Filter documents by extracted field values using composable conditions. Conditions accept canonical field names (e.g. `vendor.name`, `policy.0_coverage_type`) or field UUIDs. The Talonic API resolves names to IDs server-side."
       },
-      { type: "heading", level: 3, text: "When to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -26864,7 +28619,11 @@ var sections_default2 = [
           "You need a sortable, paginated list filtered by field conditions."
         ]
       },
-      { type: "heading", level: 3, text: "When NOT to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When NOT to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -26874,7 +28633,11 @@ var sections_default2 = [
           "The user wants to extract from a new document \u2192 use `talonic_extract`."
         ]
       },
-      { type: "heading", level: 3, text: "Input schema" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Input schema"
+      },
       {
         type: "param-table",
         title: "Top-level parameters",
@@ -26900,7 +28663,11 @@ var sections_default2 = [
             type: "number",
             description: "Page number for pagination (1-based)."
           },
-          { name: "limit", type: "number", description: "Results per page. Default: 50." },
+          {
+            name: "limit",
+            type: "number",
+            description: "Results per page. Default: 50."
+          },
           {
             name: "source_connection_id",
             type: "string",
@@ -26940,13 +28707,25 @@ var sections_default2 = [
           }
         ]
       },
-      { type: "heading", level: 3, text: "Operator reference" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Operator reference"
+      },
       {
         type: "param-table",
         title: "Operators",
         params: [
-          { name: "eq", type: "=", description: "Exact equality." },
-          { name: "neq", type: "!=", description: "Not equal." },
+          {
+            name: "eq",
+            type: "=",
+            description: "Exact equality."
+          },
+          {
+            name: "neq",
+            type: "!=",
+            description: "Not equal."
+          },
           {
             name: "gt / gte",
             type: "> / >=",
@@ -26967,7 +28746,11 @@ var sections_default2 = [
             type: "substring",
             description: "Case-insensitive substring match on string fields."
           },
-          { name: "is_empty", type: "null check", description: "Field has no value." },
+          {
+            name: "is_empty",
+            type: "null check",
+            description: "Field has no value."
+          },
           {
             name: "is_not_empty",
             type: "presence",
@@ -26975,14 +28758,22 @@ var sections_default2 = [
           }
         ]
       },
-      { type: "heading", level: 3, text: "Example" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Example"
+      },
       {
         type: "code",
         language: "json",
         title: "Find invoices over 1000 from Acme",
         code: '{\n  "conditions": [\n    { "field": "vendor.name", "operator": "contains", "value": "Acme" },\n    { "field": "total_amount", "operator": "gt", "value": 1000 }\n  ],\n  "sort": { "field": "total_amount", "direction": "desc" },\n  "limit": 10\n}'
       },
-      { type: "heading", level: 3, text: "Response shape" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Response shape"
+      },
       {
         type: "code",
         language: "json",
@@ -26990,20 +28781,61 @@ var sections_default2 = [
         code: '{\n  "documents": [\n    {\n      "id": "d_abc123",\n      "filename": "acme-invoice-q4.pdf",\n      "documentType": "invoice",\n      "extractedFields": {\n        "vendor.name": "Acme Corp",\n        "total_amount": 14250.00\n      }\n    }\n  ],\n  "total": 1,\n  "page": 1,\n  "perPage": 10\n}'
       },
       {
-        type: "callout",
-        text: "Numeric operators (`gt`, `gte`, `lt`, `lte`, `between`) only resolve correctly when the schema field is typed as `number`. A field typed as `string` that holds numeric content (e.g. `\u20AC1,500.00`) will silently return zero matches. The API now returns a `warnings[]` array on the filter response when a numeric operator is applied to a string-typed field, explaining the lexicographic-comparison issue and suggesting a `data_type` change."
+        type: "heading",
+        level: 3,
+        text: "Schema typing (preventive + reactive)"
+      },
+      {
+        type: "paragraph",
+        text: "Numeric operators (`gt`, `gte`, `lt`, `lte`, `between`) only resolve correctly when the schema field is typed as `number`. A field typed as `string` that holds numeric content (e.g. `\u20AC1,500.00`) will silently return zero matches even after extraction. There are two ways to handle this \u2014 pick the right one before constructing the call."
+      },
+      {
+        type: "heading",
+        level: 4,
+        text: "Preventive \u2014 gate on `dataType`"
+      },
+      {
+        type: "paragraph",
+        text: 'Call `talonic_search` first and read `dataType` on the field entry. If `dataType !== "number"`, do **not** issue a numeric operator on that field. Pick a string-friendly operator (`eq`, `contains`) or warn the user that the field needs a `data_type` change in the schema before the query can succeed. This avoids the silent-zero-matches outcome entirely.'
+      },
+      {
+        type: "heading",
+        level: 4,
+        text: "Reactive \u2014 handle `warnings[]`"
+      },
+      {
+        type: "paragraph",
+        text: "When a numeric operator is applied to a string-typed field, the API attaches a `warnings[]` array to the filter response. Each entry has `code`, `message`, `field`/`field_id`, and a `suggestion`. The MCP tool surfaces this in `structuredContent` \u2014 agents should relay the `message` (and `suggestion`, when present) to the user rather than silently retrying."
+      },
+      {
+        type: "code",
+        language: "json",
+        title: "Response with a warning",
+        code: '{\n  "data": [],\n  "total": 0,\n  "warnings": [\n    {\n      "code": "numeric_operator_on_string_field",\n      "message": "Operator `gt` was applied to field `invoice_total` typed as string. Numeric comparisons against string-typed fields use lexicographic ordering and may return zero matches.",\n      "field": "invoice_total",\n      "field_id": "fld_inv_total",\n      "suggestion": "Change the field\'s data_type to `number` in the schema definition."\n    }\n  ]\n}'
+      },
+      {
+        type: "heading",
+        level: 3,
+        text: "Cost"
       },
-      { type: "heading", level: 3, text: "Cost" },
       {
         type: "paragraph",
         text: "Filter calls are **free** \u2014 they query already-extracted data and do not consume extraction credits."
       },
-      { type: "heading", level: 3, text: "Errors" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Errors"
+      },
       {
         type: "param-table",
         title: "Common errors",
         params: [
-          { name: "unauthorized", type: "401", description: "Invalid or missing API key." },
+          {
+            name: "unauthorized",
+            type: "401",
+            description: "Invalid or missing API key."
+          },
           {
             name: "no_field_match",
             type: "422",
@@ -27018,8 +28850,14 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "talonic_search", slug: "mcp-talonic-search" },
-      { label: "Filter & Search API", slug: "field-autocomplete" }
+      {
+        label: "talonic_search",
+        slug: "mcp-talonic-search"
+      },
+      {
+        label: "Filter & Search API",
+        slug: "field-autocomplete"
+      }
     ],
     faq: [
       {
@@ -27029,16 +28867,23 @@ var sections_default2 = [
       {
         question: "How do I find field names for filtering?",
         answer: "Call talonic_search first. Use fields[] entries where filterable is true \u2014 their canonicalName values are what you pass as the field parameter in filter conditions. Fields with filterable: false have no extracted data yet and cannot be filtered."
+      },
+      {
+        question: "Why does my `talonic_filter` query with `gt` return zero matches on a numeric-looking field?",
+        answer: "The schema field is almost certainly typed as `string`, not `number`. Numeric operators against string-typed fields fall back to lexicographic comparison and silently return zero. Prevention: call `talonic_search` first and check `dataType` before issuing the filter. Recovery: the response's `warnings[]` array explains the issue and suggests a `data_type` change in the schema definition."
       }
     ],
     mentions: [
-      "talonic_filter",
-      "filter",
+      "canonical field name",
       "conditions",
+      "dataType",
+      "filter",
+      "is_not_empty",
       "operators",
-      "canonical field name",
-      "warnings",
-      "is_not_empty"
+      "preventive guard",
+      "schema typing",
+      "talonic_filter",
+      "warnings"
     ]
   },
   {
@@ -27052,7 +28897,11 @@ var sections_default2 = [
         type: "paragraph",
         text: "Fetch full metadata for a single document by ID. Returns filename, page count, detected document type, language, processing log, and link URLs."
       },
-      { type: "heading", level: 3, text: "When to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -27062,7 +28911,11 @@ var sections_default2 = [
           "The user asks 'tell me about document X'."
         ]
       },
-      { type: "heading", level: 3, text: "When NOT to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When NOT to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -27072,7 +28925,11 @@ var sections_default2 = [
           "The user has a file but no `document_id` yet \u2192 call `talonic_extract` first."
         ]
       },
-      { type: "heading", level: 3, text: "Input schema" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Input schema"
+      },
       {
         type: "param-table",
         title: "Parameters",
@@ -27085,22 +28942,43 @@ var sections_default2 = [
           }
         ]
       },
-      { type: "heading", level: 3, text: "Response shape" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Response shape"
+      },
       {
         type: "code",
         language: "json",
         title: "Example response",
         code: '{\n  "id": "d_abc123",\n  "filename": "invoice.pdf",\n  "documentType": "invoice",\n  "language": "en",\n  "pageCount": 2,\n  "processingLog": [...],\n  "links": {\n    "self": "https://api.talonic.com/v1/documents/d_abc123",\n    "extractions": "https://api.talonic.com/v1/documents/d_abc123/extractions",\n    "dashboard": "https://app.talonic.com/documents/d_abc123"\n  }\n}'
       },
-      { type: "heading", level: 3, text: "Cost" },
-      { type: "paragraph", text: "Free \u2014 metadata lookups do not consume extraction credits." }
+      {
+        type: "heading",
+        level: 3,
+        text: "Cost"
+      },
+      {
+        type: "paragraph",
+        text: "Free \u2014 metadata lookups do not consume extraction credits."
+      }
     ],
     related: [
-      { label: "SDK Documents", slug: "sdk-documents" },
-      { label: "Get Document", slug: "get-document" }
+      {
+        label: "SDK Documents",
+        slug: "sdk-documents"
+      },
+      {
+        label: "Get Document",
+        slug: "get-document"
+      }
     ],
     faq: [],
-    mentions: ["talonic_get_document", "metadata", "document_id"]
+    mentions: [
+      "talonic_get_document",
+      "metadata",
+      "document_id"
+    ]
   },
   {
     slug: "mcp-talonic-to-markdown",
@@ -27113,7 +28991,11 @@ var sections_default2 = [
         type: "paragraph",
         text: "Get OCR-converted markdown for a document. Accepts an existing `document_id` (cheapest \u2014 one API call, no re-processing), or raw file bytes, a local path, or a URL."
       },
-      { type: "heading", level: 3, text: "When to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -27124,7 +29006,11 @@ var sections_default2 = [
           "The user has a raw PDF / scan / image and wants markdown directly without designing a schema."
         ]
       },
-      { type: "heading", level: 3, text: "When NOT to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When NOT to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -27132,8 +29018,15 @@ var sections_default2 = [
           "The user wants specific structured fields \u2192 use `talonic_extract` with a schema."
         ]
       },
-      { type: "heading", level: 3, text: "Input schema" },
-      { type: "paragraph", text: "Provide **exactly one** of the following:" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Input schema"
+      },
+      {
+        type: "paragraph",
+        text: "Provide **exactly one** of the following:"
+      },
       {
         type: "param-table",
         title: "Parameters",
@@ -27165,19 +29058,31 @@ var sections_default2 = [
           }
         ]
       },
-      { type: "heading", level: 3, text: "Response shape" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Response shape"
+      },
       {
         type: "code",
         language: "json",
         title: "Example response",
         code: '{\n  "documentId": "d_abc123",\n  "markdown": "# Invoice INV-2024-0847\\n\\n**Vendor:** Acme Corp\\n**Date:** 2024-01-15\\n\\n| Item | Qty | Unit Price | Total |\\n|------|-----|------------|-------|\\n| Widget A | 100 | 42.50 | 4,250.00 |\\n| Widget B | 200 | 50.00 | 10,000.00 |\\n\\n**Total: 14,250.00 EUR**"\n}'
       },
-      { type: "heading", level: 3, text: "Cost" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Cost"
+      },
       {
         type: "paragraph",
         text: "**Free when using `document_id`** \u2014 the document is already ingested. When passing a raw file (`file_data`, `file_path`, `file_url`), the tool auto-ingests via extract first, consuming **one extraction credit**. To avoid unnecessary cost: if you've already extracted a document, reuse the `document_id` from that response."
       },
-      { type: "heading", level: 3, text: "Errors" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Errors"
+      },
       {
         type: "param-table",
         title: "Common errors",
@@ -27206,8 +29111,14 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "talonic_extract", slug: "mcp-talonic-extract" },
-      { label: "SDK getMarkdown", slug: "sdk-documents" }
+      {
+        label: "talonic_extract",
+        slug: "mcp-talonic-extract"
+      },
+      {
+        label: "SDK getMarkdown",
+        slug: "sdk-documents"
+      }
     ],
     faq: [
       {
@@ -27219,7 +29130,12 @@ var sections_default2 = [
         answer: "If you already called talonic_extract, reuse the document_id from that response to call talonic_to_markdown for free."
       }
     ],
-    mentions: ["talonic_to_markdown", "OCR", "markdown", "document_id"]
+    mentions: [
+      "talonic_to_markdown",
+      "OCR",
+      "markdown",
+      "document_id"
+    ]
   },
   {
     slug: "mcp-talonic-list-schemas",
@@ -27232,7 +29148,11 @@ var sections_default2 = [
         type: "paragraph",
         text: "List all saved schemas in the workspace. Returns each schema with its ID, name, description, version, field count, and full JSON Schema definition."
       },
-      { type: "heading", level: 3, text: "When to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -27243,7 +29163,11 @@ var sections_default2 = [
           "You need a `schema_id` for `talonic_extract`."
         ]
       },
-      { type: "heading", level: 3, text: "When NOT to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When NOT to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -27251,17 +29175,41 @@ var sections_default2 = [
           "The user wants to extract data and provides an inline schema \u2192 call `talonic_extract` directly."
         ]
       },
-      { type: "heading", level: 3, text: "Input schema" },
-      { type: "paragraph", text: "No parameters required." },
-      { type: "heading", level: 3, text: "Cost" },
-      { type: "paragraph", text: "Free \u2014 listing schemas does not consume extraction credits." }
+      {
+        type: "heading",
+        level: 3,
+        text: "Input schema"
+      },
+      {
+        type: "paragraph",
+        text: "No parameters required."
+      },
+      {
+        type: "heading",
+        level: 3,
+        text: "Cost"
+      },
+      {
+        type: "paragraph",
+        text: "Free \u2014 listing schemas does not consume extraction credits."
+      }
     ],
     related: [
-      { label: "talonic_save_schema", slug: "mcp-talonic-save-schema" },
-      { label: "SDK Schemas", slug: "sdk-schemas" }
+      {
+        label: "talonic_save_schema",
+        slug: "mcp-talonic-save-schema"
+      },
+      {
+        label: "SDK Schemas",
+        slug: "sdk-schemas"
+      }
     ],
     faq: [],
-    mentions: ["talonic_list_schemas", "schemas", "schema_id"]
+    mentions: [
+      "talonic_list_schemas",
+      "schemas",
+      "schema_id"
+    ]
   },
   {
     slug: "mcp-talonic-save-schema",
@@ -27274,7 +29222,11 @@ var sections_default2 = [
         type: "paragraph",
         text: "Save a schema definition to the workspace for reuse across future extractions. Returns the saved schema with its assigned `id` (UUID) and `short_id` (`SCH-XXXXXXXX`)."
       },
-      { type: "heading", level: 3, text: "When to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -27284,7 +29236,11 @@ var sections_default2 = [
           "The user wants to standardise extraction across many documents of the same type."
         ]
       },
-      { type: "heading", level: 3, text: "When NOT to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When NOT to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -27293,7 +29249,11 @@ var sections_default2 = [
           "The user has not confirmed the schema design \u2014 avoid creating clutter."
         ]
       },
-      { type: "heading", level: 3, text: "Input schema" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Input schema"
+      },
       {
         type: "param-table",
         title: "Parameters",
@@ -27317,8 +29277,15 @@ var sections_default2 = [
           }
         ]
       },
-      { type: "heading", level: 3, text: "Schema format guidance" },
-      { type: "paragraph", text: "**Full JSON Schema (recommended):**" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Schema format guidance"
+      },
+      {
+        type: "paragraph",
+        text: "**Full JSON Schema (recommended):**"
+      },
       {
         type: "code",
         language: "json",
@@ -27344,12 +29311,25 @@ var sections_default2 = [
         type: "callout",
         text: "When you call `talonic_save_schema` (or update an existing schema), the API samples the field's prior extracted values. If 80% or more of a string-typed field's values parse as numbers (with at least 5 samples), the response includes a `warnings[]` suggesting `data_type: \"number\"`. Heed the warning if you plan to filter on that field with numeric operators."
       },
-      { type: "heading", level: 3, text: "Cost" },
-      { type: "paragraph", text: "Free \u2014 saving a schema does not consume extraction credits." }
+      {
+        type: "heading",
+        level: 3,
+        text: "Cost"
+      },
+      {
+        type: "paragraph",
+        text: "Free \u2014 saving a schema does not consume extraction credits."
+      }
     ],
     related: [
-      { label: "talonic_list_schemas", slug: "mcp-talonic-list-schemas" },
-      { label: "Schemas API", slug: "create-schema" }
+      {
+        label: "talonic_list_schemas",
+        slug: "mcp-talonic-list-schemas"
+      },
+      {
+        label: "Schemas API",
+        slug: "create-schema"
+      }
     ],
     faq: [
       {
@@ -27357,7 +29337,12 @@ var sections_default2 = [
         answer: "Full JSON Schema ({type: 'object', properties: {...}}) is most reliable. Flat key-type maps ({field: 'type'}) work for simple schemas but are normalized server-side and may produce errors with complex structures."
       }
     ],
-    mentions: ["talonic_save_schema", "schema", "JSON Schema", "flat key-type"]
+    mentions: [
+      "talonic_save_schema",
+      "schema",
+      "JSON Schema",
+      "flat key-type"
+    ]
   },
   {
     slug: "mcp-talonic-get-balance",
@@ -27370,7 +29355,11 @@ var sections_default2 = [
         type: "paragraph",
         text: "Read the user's current Talonic credit balance, EUR value, 30-day burn rate, projected runway, tier, and next-tier-reset timestamp. Use this to make budget-aware decisions before kicking off large batches or re-extractions."
       },
-      { type: "heading", level: 3, text: "When to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -27380,7 +29369,11 @@ var sections_default2 = [
           "The user asks how long their balance will last at the current rate."
         ]
       },
-      { type: "heading", level: 3, text: "When NOT to use" },
+      {
+        type: "heading",
+        level: 3,
+        text: "When NOT to use"
+      },
       {
         type: "list",
         ordered: false,
@@ -27389,14 +29382,29 @@ var sections_default2 = [
           "The user wants to top up credits \u2014 route them to the dashboard at `https://app.talonic.com`."
         ]
       },
-      { type: "heading", level: 3, text: "Input schema" },
-      { type: "paragraph", text: "No parameters required." },
-      { type: "heading", level: 3, text: "Response shape" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Input schema"
+      },
+      {
+        type: "paragraph",
+        text: "No parameters required."
+      },
+      {
+        type: "heading",
+        level: 3,
+        text: "Response shape"
+      },
       {
         type: "param-table",
         title: "Fields",
         params: [
-          { name: "balance_credits", type: "number", description: "Current credit balance." },
+          {
+            name: "balance_credits",
+            type: "number",
+            description: "Current credit balance."
+          },
           {
             name: "balance_eur",
             type: "number",
@@ -27430,12 +29438,25 @@ var sections_default2 = [
         title: "Example response",
         code: '{\n  "balance_credits": 1000,\n  "balance_eur": 50.00,\n  "burn_rate_30d_credits": 240,\n  "projected_runway_days": 125,\n  "tier": "pro",\n  "tier_resets_at": "2026-06-01T00:00:00.000Z"\n}'
       },
-      { type: "heading", level: 3, text: "Cost" },
-      { type: "paragraph", text: "Free \u2014 balance lookups do not consume extraction credits." }
+      {
+        type: "heading",
+        level: 3,
+        text: "Cost"
+      },
+      {
+        type: "paragraph",
+        text: "Free \u2014 balance lookups do not consume extraction credits."
+      }
     ],
     related: [
-      { label: "talonic_extract", slug: "mcp-talonic-extract" },
-      { label: "Cost & Rate Limits", slug: "mcp-cost-and-limits" }
+      {
+        label: "talonic_extract",
+        slug: "mcp-talonic-extract"
+      },
+      {
+        label: "Cost & Rate Limits",
+        slug: "mcp-cost-and-limits"
+      }
     ],
     faq: [
       {
@@ -27447,7 +29468,157 @@ var sections_default2 = [
         answer: "Days of runway at the trailing 30-day average burn rate. The value -1 means no consumption in the trailing window, so runway cannot be computed."
       }
     ],
-    mentions: ["talonic_get_balance", "credits", "balance", "tier", "burn rate", "runway"]
+    mentions: [
+      "talonic_get_balance",
+      "credits",
+      "balance",
+      "tier",
+      "burn rate",
+      "runway"
+    ]
+  },
+  {
+    slug: "mcp-talonic-request-upload",
+    parentSlug: "mcp-tools",
+    title: "talonic_request_upload",
+    seoTitle: "talonic_request_upload MCP Tool \u2014 Talonic Docs",
+    description: "Request a browser upload link for files too large for tool-call arguments or when running in a sandboxed hosted environment.",
+    content: [
+      {
+        type: "paragraph",
+        text: "Request a browser upload link for the user. Use this when the user wants to extract a file but you cannot deliver it directly \u2014 the file is too large for tool-call arguments (~32 KB cap on hosted connectors), or you're running in a sandboxed environment (Claude.ai, ChatGPT) that blocks outbound file transfers."
+      },
+      {
+        type: "heading",
+        level: 3,
+        text: "When to use"
+      },
+      {
+        type: "list",
+        ordered: false,
+        items: [
+          "The user has a file to extract but you cannot send it via `file_data` (file larger than ~32 KB, or the environment blocks outbound data).",
+          "You are running in a hosted/sandboxed environment (Claude.ai, ChatGPT) where `file_data` cannot be used reliably.",
+          "The user explicitly asks for an upload link."
+        ]
+      },
+      {
+        type: "heading",
+        level: 3,
+        text: "When NOT to use"
+      },
+      {
+        type: "list",
+        ordered: false,
+        items: [
+          "You can deliver the file directly via `file_data` (local stdio installs with small files).",
+          "The file is already accessible via a public URL \u2192 use `file_url` on `talonic_extract`.",
+          "The document is already in the workspace \u2192 use `document_id` on `talonic_extract`."
+        ]
+      },
+      {
+        type: "heading",
+        level: 3,
+        text: "How the flow works"
+      },
+      {
+        type: "list",
+        ordered: true,
+        items: [
+          "Call `talonic_request_upload` with the filename. You receive a `document_id`, an `upload_url`, and an `expires_at` timestamp.",
+          "Show the `upload_url` to the user and ask them to open it in their browser.",
+          "The user drops the file on the upload page. The browser uploads directly to Talonic \u2014 no tool-call size cap, no sandbox restriction.",
+          "Poll with `talonic_get_document` using the `document_id` until `status` is `uploaded`.",
+          "Call `talonic_extract` with the `document_id` and a schema to extract structured data."
+        ]
+      },
+      {
+        type: "heading",
+        level: 3,
+        text: "Input schema"
+      },
+      {
+        type: "param-table",
+        title: "Parameters",
+        params: [
+          {
+            name: "filename",
+            type: "string",
+            required: true,
+            description: "The name of the file being uploaded, including extension (e.g. `invoice.pdf`). Used to pre-allocate the document and infer MIME type."
+          }
+        ]
+      },
+      {
+        type: "heading",
+        level: 3,
+        text: "Response shape"
+      },
+      {
+        type: "code",
+        language: "json",
+        code: '{\n  "document_id": "d8f3a1b2-...",\n  "upload_url": "https://app.talonic.com/u/abc12345-...",\n  "expires_at": "2026-05-27T22:15:00.000Z"\n}'
+      },
+      {
+        type: "param-table",
+        title: "Response fields",
+        params: [
+          {
+            name: "document_id",
+            type: "string",
+            description: "The pre-allocated document ID. Use with `talonic_get_document` to poll status, and with `talonic_extract` once uploaded."
+          },
+          {
+            name: "upload_url",
+            type: "string",
+            description: "URL the user should open in their browser to drop the file. Expires after 15 minutes."
+          },
+          {
+            name: "expires_at",
+            type: "string",
+            description: "ISO 8601 timestamp when the upload link expires."
+          }
+        ]
+      },
+      {
+        type: "callout",
+        text: "Upload links are single-use and expire after 15 minutes. If the user doesn't upload in time, call `talonic_request_upload` again to get a fresh link."
+      }
+    ],
+    related: [
+      {
+        label: "talonic_extract",
+        slug: "mcp-talonic-extract"
+      },
+      {
+        label: "talonic_get_document",
+        slug: "mcp-talonic-get-document"
+      },
+      {
+        label: "Drag & Drop in Chat",
+        slug: "mcp-drag-drop"
+      }
+    ],
+    faq: [
+      {
+        question: "How do I upload a file through Claude.ai to Talonic?",
+        answer: "Call talonic_request_upload with the filename. Show the returned upload_url to the user. They open it in their browser and drop the file. Poll talonic_get_document until status is 'uploaded', then call talonic_extract with the document_id."
+      },
+      {
+        question: "Why can't I just send the file through file_data on Claude.ai?",
+        answer: "Claude.ai's hosted connector caps tool-call arguments at ~32 KB (decoded). Real documents are typically 100 KB to several MB. The browser-handoff upload bypasses this limit entirely by moving the file transfer to the user's browser."
+      }
+    ],
+    mentions: [
+      "upload",
+      "browser handoff",
+      "hosted connector",
+      "Claude.ai",
+      "ChatGPT",
+      "file size limit",
+      "sandbox",
+      "upload link"
+    ]
   },
   {
     slug: "mcp-schemas-resource",
@@ -27456,25 +29627,43 @@ var sections_default2 = [
     seoTitle: "MCP Resources \u2014 Talonic Docs",
     description: "Two MCP resources exposed by the Talonic server: talonic://schemas (saved schemas) and talonic://webhooks/reference (webhook event reference).",
     content: [
-      { type: "heading", level: 3, text: "talonic://schemas" },
+      {
+        type: "heading",
+        level: 3,
+        text: "talonic://schemas"
+      },
       {
         type: "paragraph",
         text: "Exposes the saved-schemas list to clients that browse MCP resources separately. Claude Desktop and Cowork render these in the UI. The contents mirror `talonic_list_schemas` but in a browseable form."
       },
-      { type: "heading", level: 3, text: "talonic://webhooks/reference" },
+      {
+        type: "heading",
+        level: 3,
+        text: "talonic://webhooks/reference"
+      },
       {
         type: "paragraph",
         text: "Static reference documenting the webhook events the Talonic API can fire (extraction lifecycle, document classification, etc.), their payload shapes, and how to subscribe. Useful when an agent is helping the user wire Talonic into a backend that needs to react to extraction events."
       }
     ],
-    related: [{ label: "talonic_list_schemas", slug: "mcp-talonic-list-schemas" }],
+    related: [
+      {
+        label: "talonic_list_schemas",
+        slug: "mcp-talonic-list-schemas"
+      }
+    ],
     faq: [
       {
         question: "What resources does the Talonic MCP server expose?",
         answer: "Two resources: talonic://schemas (browseable list of saved schemas) and talonic://webhooks/reference (static reference for the API's webhook events and payloads)."
       }
     ],
-    mentions: ["MCP resource", "talonic://schemas", "talonic://webhooks/reference", "webhooks"]
+    mentions: [
+      "MCP resource",
+      "talonic://schemas",
+      "talonic://webhooks/reference",
+      "webhooks"
+    ]
   },
   {
     slug: "mcp-cost-and-limits",
@@ -27483,7 +29672,11 @@ var sections_default2 = [
     seoTitle: "MCP Cost and Rate Limits \u2014 Talonic Docs",
     description: "Which MCP tool calls cost extraction credits, rate limit behavior, insufficient-credit handling, and how to avoid re-extraction.",
     content: [
-      { type: "heading", level: 3, text: "What costs credits" },
+      {
+        type: "heading",
+        level: 3,
+        text: "What costs credits"
+      },
       {
         type: "paragraph",
         text: "Only extraction operations consume credits. Everything else is free:"
@@ -27507,14 +29700,26 @@ var sections_default2 = [
             type: "free",
             description: "Document already ingested \u2014 just fetches stored markdown."
           },
-          { name: "talonic_search", type: "free", description: "Queries indexed data." },
+          {
+            name: "talonic_search",
+            type: "free",
+            description: "Queries indexed data."
+          },
           {
             name: "talonic_filter",
             type: "free",
             description: "Queries extracted field values."
           },
-          { name: "talonic_get_document", type: "free", description: "Metadata lookup." },
-          { name: "talonic_list_schemas", type: "free", description: "Lists saved schemas." },
+          {
+            name: "talonic_get_document",
+            type: "free",
+            description: "Metadata lookup."
+          },
+          {
+            name: "talonic_list_schemas",
+            type: "free",
+            description: "Lists saved schemas."
+          },
           {
             name: "talonic_save_schema",
             type: "free",
@@ -27531,7 +29736,11 @@ var sections_default2 = [
         type: "callout",
         text: "The per-call cost of `talonic_extract` (and `talonic_to_markdown` with a raw file) is also surfaced on the response under `cost` (`costCredits`, `costEur`, `balanceCredits`, plus a breakdown of `cellsResolvedRegistry` and `cellsResolvedAi`). Agents can read this immediately after the call rather than calling `talonic_get_balance` again."
       },
-      { type: "heading", level: 3, text: "Avoiding re-extraction" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Avoiding re-extraction"
+      },
       {
         type: "paragraph",
         text: "Agents should avoid extracting the same document twice. Best practices:"
@@ -27545,26 +29754,47 @@ var sections_default2 = [
           "Use `talonic_filter` to query already-extracted data instead of re-extracting with a different schema when the fields you need are already captured."
         ]
       },
-      { type: "heading", level: 3, text: "Rate limits" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Rate limits"
+      },
       {
         type: "paragraph",
         text: "The Talonic API enforces per-key rate limits. When exceeded, the server returns `429 Too Many Requests` with a `X-RateLimit-Reset` header. The MCP server (and the underlying Node SDK) retries automatically with exponential backoff up to `maxRetries` (default: 3). If retries are exhausted, the tool returns an error with the reset timestamp."
       },
-      { type: "heading", level: 3, text: "Insufficient credits" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Insufficient credits"
+      },
       {
         type: "paragraph",
         text: "When extraction credits are exhausted, the API returns `402 Payment Required`. The tool surfaces this as an error. The agent should inform the user that their credit balance is depleted and suggest upgrading their plan or waiting for the daily reset (free tier: 50 extractions/day, resets at midnight UTC)."
       },
-      { type: "heading", level: 3, text: "Free tier limits" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Free tier limits"
+      },
       {
         type: "paragraph",
         text: "The free tier includes 50 extractions per day (resets at midnight UTC). Search, filter, metadata, and schema operations are unlimited. No credit card required."
       }
     ],
     related: [
-      { label: "talonic_extract", slug: "mcp-talonic-extract" },
-      { label: "Authentication", slug: "mcp-authentication" },
-      { label: "SDK Retries", slug: "sdk-retries" }
+      {
+        label: "talonic_extract",
+        slug: "mcp-talonic-extract"
+      },
+      {
+        label: "Authentication",
+        slug: "mcp-authentication"
+      },
+      {
+        label: "SDK Retries",
+        slug: "sdk-retries"
+      }
     ],
     faq: [
       {
@@ -27580,7 +29810,14 @@ var sections_default2 = [
         answer: "The API returns 429 with a reset timestamp. The SDK retries automatically with exponential backoff (up to 3 retries by default)."
       }
     ],
-    mentions: ["credits", "rate limits", "429", "402", "free tier", "cost"]
+    mentions: [
+      "credits",
+      "rate limits",
+      "429",
+      "402",
+      "free tier",
+      "cost"
+    ]
   },
   {
     slug: "mcp-drag-drop",
@@ -27602,30 +29839,49 @@ var sections_default2 = [
         type: "paragraph",
         text: "From `@talonic/mcp@0.1.4`, agents can pass **`file_data`** (base64-encoded file bytes) and **`filename`** on `talonic_extract` and `talonic_to_markdown`. The agent reads the file bytes from the conversation, base64-encodes them, and passes them through the MCP tool call. The MCP server decodes, infers MIME type from the filename, and uploads to the Talonic API as a normal multipart request. Tool descriptions advertise `file_data` as the recommended input here, so well-trained agents reach for it automatically. No client-side configuration required."
       },
-      { type: "heading", level: 3, text: "Claude.ai hosted connector" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Claude.ai hosted connector"
+      },
       {
         type: "callout",
         variant: "warning",
-        text: "Claude.ai's hosted-connector pipeline imposes a hard size limit on tool-call arguments (effectively under ~1 KB). A base64-encoded real PDF (typically hundreds of KB at minimum) gets truncated before reaching the MCP server. The Talonic API receives a few hundred bytes, registers an empty document, and the response comes back with `null` extracted fields. This is a Claude.ai platform limit on connectors, not a Talonic MCP server bug."
+        text: "Claude.ai's hosted-connector pipeline imposes a hard size limit on tool-call arguments (effectively ~32 KB decoded). A base64-encoded real PDF (typically hundreds of KB at minimum) gets truncated before reaching the MCP server. The Talonic API receives ~32 KB, registers an empty document, and the response comes back with `null` extracted fields. This is a Claude.ai platform limit on connectors, not a Talonic MCP server bug."
+      },
+      {
+        type: "paragraph",
+        text: "**Workarounds for Claude.ai users:**"
       },
-      { type: "paragraph", text: "**Workarounds for Claude.ai users:**" },
       {
         type: "list",
         ordered: false,
         items: [
+          "`talonic_request_upload`: the recommended path. The agent gets an upload link, the user drops the file in their browser, and the agent continues with the `document_id`. Works with any file size the API accepts.",
           "`file_url`: pass a publicly reachable URL; the Talonic API fetches it server-side. Best for files already on the public web.",
           "`document_id`: upload the file once via `app.talonic.com`, then reference the returned id. Best for sensitive files you don't want to expose publicly.",
           "Switch to a local stdio install (`npx -y @talonic/mcp@latest` in Claude Desktop, Cursor, Cline, etc.) \u2014 local stdio has no parameter-size cap and `file_data` works for any file size the API accepts."
-        ]
+        ],
+        text: ""
       },
       {
         type: "paragraph",
-        text: "The architectural fix that would unblock drag-and-drop through the Claude.ai connector is a pre-signed upload URL flow: a new MCP tool returns a one-time URL plus a reserved `document_id`, the user uploads from their browser directly to Talonic's storage, and the agent then calls `talonic_extract` with the `document_id`. This bypasses the connector's argument-size pipe entirely."
+        text: "From `@talonic/mcp@0.1.7`, this is solved by the **`talonic_request_upload`** tool. The agent calls it to get a one-time upload URL plus a reserved `document_id`. The user opens the link in their browser and drops the file \u2014 no tool-call size cap, no sandbox restriction. The agent then polls `talonic_get_document` until the file is ready and proceeds with `talonic_extract` using the `document_id`."
       }
     ],
     related: [
-      { label: "talonic_extract", slug: "mcp-talonic-extract" },
-      { label: "Installation", slug: "mcp-installation" }
+      {
+        label: "talonic_extract",
+        slug: "mcp-talonic-extract"
+      },
+      {
+        label: "Installation",
+        slug: "mcp-installation"
+      },
+      {
+        label: "talonic_request_upload",
+        slug: "mcp-talonic-request-upload"
+      }
     ],
     faq: [
       {
@@ -27634,7 +29890,7 @@ var sections_default2 = [
       },
       {
         question: "Why does drag-and-drop fail on the Claude.ai connector?",
-        answer: "Claude.ai imposes a hard ~1 KB cap on tool-call argument values. A base64-encoded real PDF cannot fit, so file_data is truncated to a few hundred bytes before the MCP server receives it. Workarounds: file_url (public URL), document_id (upload via app.talonic.com first), or switch to a local stdio install."
+        answer: "Claude.ai imposes a ~32 KB cap on tool-call argument values. A base64-encoded real PDF cannot fit. Use talonic_request_upload to get a browser upload link \u2014 the user drops the file in their browser, bypassing the cap entirely. Alternatives: file_url (public URL), document_id (upload via app.talonic.com first), or switch to a local stdio install."
       }
     ],
     mentions: [
@@ -27680,11 +29936,23 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "Configuration", slug: "mcp-configuration" },
-      { label: "Introduction", slug: "mcp-introduction" }
+      {
+        label: "Configuration",
+        slug: "mcp-configuration"
+      },
+      {
+        label: "Introduction",
+        slug: "mcp-introduction"
+      }
     ],
     faq: [],
-    mentions: ["architecture", "stdio", "HTTP", "Streamable HTTP", "session"]
+    mentions: [
+      "architecture",
+      "stdio",
+      "HTTP",
+      "Streamable HTTP",
+      "session"
+    ]
   },
   {
     slug: "mcp-configuration",
@@ -27693,8 +29961,15 @@ var sections_default2 = [
     seoTitle: "MCP Server Configuration \u2014 Talonic Docs",
     description: "Environment variables for the local MCP server and header options for the hosted server.",
     content: [
-      { type: "heading", level: 3, text: "Local server (env vars)" },
-      { type: "paragraph", text: "Set via the `env` block in your MCP client config:" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Local server (env vars)"
+      },
+      {
+        type: "paragraph",
+        text: "Set via the `env` block in your MCP client config:"
+      },
       {
         type: "param-table",
         title: "Environment variables",
@@ -27712,7 +29987,11 @@ var sections_default2 = [
           }
         ]
       },
-      { type: "heading", level: 3, text: "Hosted server (headers)" },
+      {
+        type: "heading",
+        level: 3,
+        text: "Hosted server (headers)"
+      },
       {
         type: "paragraph",
         text: "The hosted server at `mcp.talonic.com` is configured entirely via the MCP client config:"
@@ -27728,11 +30007,22 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "Installation", slug: "mcp-installation" },
-      { label: "Authentication", slug: "mcp-authentication" }
+      {
+        label: "Installation",
+        slug: "mcp-installation"
+      },
+      {
+        label: "Authentication",
+        slug: "mcp-authentication"
+      }
     ],
     faq: [],
-    mentions: ["TALONIC_API_KEY", "TALONIC_BASE_URL", "configuration", "headers"]
+    mentions: [
+      "TALONIC_API_KEY",
+      "TALONIC_BASE_URL",
+      "configuration",
+      "headers"
+    ]
   },
   {
     slug: "mcp-troubleshooting",
@@ -27821,7 +30111,12 @@ var sections_default2 = [
         type: "paragraph",
         text: "Extraction credit balance is exhausted. Free tier: 50 extractions/day, resets at midnight UTC. Upgrade plan or wait for reset."
       },
-      { type: "heading", level: 3, id: "ts-cached", text: "Tool descriptions look wrong" },
+      {
+        type: "heading",
+        level: 3,
+        id: "ts-cached",
+        text: "Tool descriptions look wrong"
+      },
       {
         type: "paragraph",
         text: "Some MCP clients cache tool descriptions. Restart the client after a server update."
@@ -27838,17 +30133,33 @@ var sections_default2 = [
       }
     ],
     related: [
-      { label: "Installation", slug: "mcp-installation" },
-      { label: "Configuration", slug: "mcp-configuration" },
-      { label: "Cost & Rate Limits", slug: "mcp-cost-and-limits" }
+      {
+        label: "Installation",
+        slug: "mcp-installation"
+      },
+      {
+        label: "Configuration",
+        slug: "mcp-configuration"
+      },
+      {
+        label: "Cost & Rate Limits",
+        slug: "mcp-cost-and-limits"
+      }
     ],
     faq: [],
-    mentions: ["troubleshooting", "debugging", "errors", "401", "402", "500"]
+    mentions: [
+      "troubleshooting",
+      "debugging",
+      "errors",
+      "401",
+      "402",
+      "500"
+    ]
   }
 ];
 
 // src/content/mcp/index.ts
-var sections51 = sections_default2;
+var sections54 = sections_default2;
 
 // src/content/index.ts
 var ALL_PLATFORM_RAW = [
@@ -27902,10 +30213,13 @@ var ALL_API_RAW = [
   ...sections47,
   ...sections48,
   ...sections49,
+  ...sections50,
+  ...sections51,
+  ...sections52,
   ...sections45
 ];
-var ALL_SDK_RAW = [...sections50];
-var ALL_MCP_RAW = [...sections51];
+var ALL_SDK_RAW = [...sections53];
+var ALL_MCP_RAW = [...sections54];
 function enrich(raw, navSections, domain) {
   return raw.map((r) => {
     const { prev, next } = derivePrevNext(navSections, r.slug);