@pdfvector/instance-contract 0.0.27 → 0.0.29

package/.tsc/lib/router/academic/index.d.ts

@@ -1,3 +1,5 @@
 export * from "./fetch";
 export * from "./find-citations";
+export * from "./paper-graph";
 export * from "./search";
+export * from "./similar-papers";

package/.tsc/lib/router/academic/index.js

@@ -1,3 +1,5 @@
 export * from "./fetch";
 export * from "./find-citations";
+export * from "./paper-graph";
 export * from "./search";
+export * from "./similar-papers";

package/.tsc/lib/router/academic/paper-graph.d.ts

@@ -0,0 +1,85 @@
+import { z } from "zod";
+export declare const paperGraph: import("@orpc/contract").ContractProcedureBuilderWithInputOutput<z.ZodObject<{
+    id: z.ZodString;
+    citationsLimit: z.ZodDefault<z.ZodNumber>;
+    referencesLimit: z.ZodDefault<z.ZodNumber>;
+    citationsOffset: z.ZodDefault<z.ZodNumber>;
+    referencesOffset: z.ZodDefault<z.ZodNumber>;
+    fields: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodEnum<{
+        url: "url";
+        doi: "doi";
+        title: "title";
+        providerURL: "providerURL";
+        date: "date";
+        authors: "authors";
+        year: "year";
+        totalCitations: "totalCitations";
+        totalReferences: "totalReferences";
+        abstract: "abstract";
+        pdfURL: "pdfURL";
+        provider: "provider";
+        providerData: "providerData";
+    }>>>>;
+    callback: z.ZodOptional<z.ZodObject<{
+        url: z.ZodURL;
+        type: z.ZodOptional<z.ZodString>;
+    }, z.core.$strip>>;
+}, z.core.$strip>, z.ZodObject<{
+    paper: z.ZodObject<{
+        doi: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        title: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        url: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        providerURL: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        authors: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodObject<{
+            name: z.ZodString;
+            url: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        }, z.core.$strip>>>>;
+        date: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        year: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        totalCitations: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        totalReferences: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        abstract: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        pdfURL: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        provider: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        providerData: z.ZodOptional<z.ZodNullable<z.ZodRecord<z.ZodString, z.ZodUnknown>>>;
+    }, z.core.$strip>;
+    citations: z.ZodArray<z.ZodObject<{
+        doi: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        title: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        url: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        providerURL: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        authors: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodObject<{
+            name: z.ZodString;
+            url: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        }, z.core.$strip>>>>;
+        date: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        year: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        totalCitations: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        totalReferences: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        abstract: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        pdfURL: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        provider: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        providerData: z.ZodOptional<z.ZodNullable<z.ZodRecord<z.ZodString, z.ZodUnknown>>>;
+    }, z.core.$strip>>;
+    references: z.ZodArray<z.ZodObject<{
+        doi: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        title: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        url: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        providerURL: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        authors: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodObject<{
+            name: z.ZodString;
+            url: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        }, z.core.$strip>>>>;
+        date: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        year: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        totalCitations: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        totalReferences: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        abstract: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        pdfURL: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        provider: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        providerData: z.ZodOptional<z.ZodNullable<z.ZodRecord<z.ZodString, z.ZodUnknown>>>;
+    }, z.core.$strip>>;
+    totalCitations: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+    totalReferences: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+    requestId: z.ZodNumber;
+}, z.core.$strip>, Record<never, never>, Record<never, never>>;

package/.tsc/lib/router/academic/paper-graph.js

@@ -0,0 +1,140 @@
+import { oc } from "@orpc/contract";
+import { z } from "zod";
+const authorSchema = z.object({
+    name: z.string(),
+    url: z.string().nullish(),
+});
+const publicationSchema = z.object({
+    doi: z.string().nullish(),
+    title: z.string().nullish(),
+    url: z.string().nullish(),
+    providerURL: z.string().nullish(),
+    authors: z.array(authorSchema).nullish(),
+    date: z.string().nullish(),
+    year: z.number().nullish(),
+    totalCitations: z.number().nullish(),
+    totalReferences: z.number().nullish(),
+    abstract: z.string().nullish(),
+    pdfURL: z.string().nullish(),
+    provider: z.string().nullish(),
+    providerData: z.record(z.string(), z.unknown()).nullish(),
+});
+const requestExamples = {
+    "Get citations by DOI": {
+        summary: "Get citations and references for a paper by DOI",
+        value: {
+            id: "10.1038/nature12373",
+            citationsLimit: 20,
+            referencesLimit: 20,
+        },
+    },
+    "Get citations by ArXiv ID": {
+        summary: "Get citations and references for an ArXiv paper",
+        value: {
+            id: "2303.08774",
+            citationsLimit: 50,
+            referencesLimit: 50,
+        },
+    },
+    "Paginated citations": {
+        summary: "Page through a paper's citations",
+        value: {
+            id: "10.1038/nature12373",
+            citationsLimit: 20,
+            citationsOffset: 20,
+            referencesLimit: 0,
+        },
+    },
+};
+export const paperGraph = oc
+    .input(z.object({
+    id: z
+        .string({ message: "id is required and must be a string" })
+        .min(1, "id must not be empty")
+        .max(500, "id must be at most 500 characters")
+        .describe("Paper identifier. Supports DOI, Semantic Scholar ID, ArXiv ID, PubMed ID, OpenAlex ID, or URL."),
+    citationsLimit: z
+        .number({ message: "citationsLimit must be a number" })
+        .int("citationsLimit must be a whole number")
+        .min(0, "citationsLimit must be 0 or greater")
+        .max(1000, "citationsLimit must be at most 1000")
+        .default(100)
+        .describe("Maximum number of citing papers to return"),
+    referencesLimit: z
+        .number({ message: "referencesLimit must be a number" })
+        .int("referencesLimit must be a whole number")
+        .min(0, "referencesLimit must be 0 or greater")
+        .max(1000, "referencesLimit must be at most 1000")
+        .default(100)
+        .describe("Maximum number of referenced papers to return"),
+    citationsOffset: z
+        .number({ message: "citationsOffset must be a number" })
+        .int("citationsOffset must be a whole number")
+        .min(0, "citationsOffset must be 0 or greater")
+        .default(0)
+        .describe("Pagination offset for citations"),
+    referencesOffset: z
+        .number({ message: "referencesOffset must be a number" })
+        .int("referencesOffset must be a whole number")
+        .min(0, "referencesOffset must be 0 or greater")
+        .default(0)
+        .describe("Pagination offset for references"),
+    fields: z
+        .array(z.enum([
+        "doi",
+        "title",
+        "url",
+        "providerURL",
+        "authors",
+        "date",
+        "year",
+        "totalCitations",
+        "totalReferences",
+        "abstract",
+        "pdfURL",
+        "provider",
+        "providerData",
+    ]), {
+        message: "fields must be an array of valid publication field names",
+    })
+        .nullish()
+        .describe("Specific publication fields to return. If omitted, returns all fields except providerData."),
+    callback: z
+        .object({
+        url: z
+            .url()
+            .describe("Webhook URL where results will be POSTed when processing completes"),
+        type: z
+            .string()
+            .optional()
+            .describe("Callback type identifier (e.g. 'zapier')"),
+    })
+        .optional()
+        .describe("Optional webhook callback for async processing. " +
+        "When provided, the server returns 202 immediately and POSTs the full response payload to the callback URL when processing completes. " +
+        "On error, the callback receives a POST with X-Pdfvector-Callback-Failed: true header and error details in the body. " +
+        "Useful for long-running operations that may exceed client timeout limits."),
+}))
+    .output(z.object({
+    paper: publicationSchema,
+    citations: z.array(publicationSchema),
+    references: z.array(publicationSchema),
+    totalCitations: z.number().int().nullish(),
+    totalReferences: z.number().int().nullish(),
+    requestId: z.number().int(),
+}))
+    .route({
+    summary: "Get paper citations and references",
+    description: "Fetch a paper's citing papers and referenced papers as full publication objects with automatic provider detection and pagination support.",
+    tags: ["Academic"],
+    spec: (op) => {
+        op.security = [{ bearerAuth: [] }];
+        const reqBody = op.requestBody;
+        if (reqBody?.content) {
+            for (const mediaType of Object.values(reqBody.content)) {
+                mediaType.examples = requestExamples;
+            }
+        }
+        return op;
+    },
+});

package/.tsc/lib/router/academic/similar-papers.d.ts

@@ -0,0 +1,70 @@
+import { z } from "zod";
+export declare const similarPapers: import("@orpc/contract").ContractProcedureBuilderWithInputOutput<z.ZodObject<{
+    id: z.ZodString;
+    limit: z.ZodDefault<z.ZodNumber>;
+    includeEdges: z.ZodDefault<z.ZodBoolean>;
+    fields: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodEnum<{
+        url: "url";
+        doi: "doi";
+        title: "title";
+        providerURL: "providerURL";
+        date: "date";
+        authors: "authors";
+        year: "year";
+        totalCitations: "totalCitations";
+        totalReferences: "totalReferences";
+        abstract: "abstract";
+        pdfURL: "pdfURL";
+        provider: "provider";
+        providerData: "providerData";
+    }>>>>;
+    callback: z.ZodOptional<z.ZodObject<{
+        url: z.ZodURL;
+        type: z.ZodOptional<z.ZodString>;
+    }, z.core.$strip>>;
+}, z.core.$strip>, z.ZodObject<{
+    seed: z.ZodObject<{
+        doi: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        title: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        url: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        providerURL: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        authors: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodObject<{
+            name: z.ZodString;
+            url: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        }, z.core.$strip>>>>;
+        date: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        year: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        totalCitations: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        totalReferences: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        abstract: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        pdfURL: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        provider: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        providerData: z.ZodOptional<z.ZodNullable<z.ZodRecord<z.ZodString, z.ZodUnknown>>>;
+    }, z.core.$strip>;
+    results: z.ZodArray<z.ZodObject<{
+        publication: z.ZodObject<{
+            doi: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+            title: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+            url: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+            providerURL: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+            authors: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodObject<{
+                name: z.ZodString;
+                url: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+            }, z.core.$strip>>>>;
+            date: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+            year: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+            totalCitations: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+            totalReferences: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+            abstract: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+            pdfURL: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+            provider: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+            providerData: z.ZodOptional<z.ZodNullable<z.ZodRecord<z.ZodString, z.ZodUnknown>>>;
+        }, z.core.$strip>;
+        similarity: z.ZodNumber;
+        pageRank: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        distance: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        citingIds: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodString>>>;
+        citedByIds: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodString>>>;
+    }, z.core.$strip>>;
+    requestId: z.ZodNumber;
+}, z.core.$strip>, Record<never, never>, Record<never, never>>;

package/.tsc/lib/router/academic/similar-papers.js

@@ -0,0 +1,139 @@
+import { oc } from "@orpc/contract";
+import { z } from "zod";
+const authorSchema = z.object({
+    name: z.string(),
+    url: z.string().nullish(),
+});
+const publicationSchema = z.object({
+    doi: z.string().nullish(),
+    title: z.string().nullish(),
+    url: z.string().nullish(),
+    providerURL: z.string().nullish(),
+    authors: z.array(authorSchema).nullish(),
+    date: z.string().nullish(),
+    year: z.number().nullish(),
+    totalCitations: z.number().nullish(),
+    totalReferences: z.number().nullish(),
+    abstract: z.string().nullish(),
+    pdfURL: z.string().nullish(),
+    provider: z.string().nullish(),
+    providerData: z.record(z.string(), z.unknown()).nullish(),
+});
+const similarPaperSchema = z.object({
+    publication: publicationSchema,
+    similarity: z
+        .number()
+        .describe("Combined similarity score from citation network analysis"),
+    pageRank: z.number().nullish().describe("Citation network importance score"),
+    distance: z
+        .number()
+        .int()
+        .nullish()
+        .describe("Graph distance from seed paper"),
+    citingIds: z
+        .array(z.string())
+        .nullish()
+        .describe("OpenAlex IDs of papers in result set that cite this paper (only if includeEdges=true)"),
+    citedByIds: z
+        .array(z.string())
+        .nullish()
+        .describe("OpenAlex IDs of papers in result set cited by this paper (only if includeEdges=true)"),
+});
+const requestExamples = {
+    "Find similar papers by DOI": {
+        summary: "Find papers similar to a given paper",
+        value: {
+            id: "10.1038/nature12373",
+            limit: 20,
+        },
+    },
+    "Find similar with edges": {
+        summary: "Find similar papers with citation graph edges",
+        value: {
+            id: "10.1038/nature12373",
+            limit: 20,
+            includeEdges: true,
+        },
+    },
+    "Find similar by OpenAlex ID": {
+        summary: "Find similar papers using an OpenAlex ID",
+        value: {
+            id: "W2963403868",
+            limit: 10,
+        },
+    },
+};
+export const similarPapers = oc
+    .input(z.object({
+    id: z
+        .string({ message: "id is required and must be a string" })
+        .min(1, "id must not be empty")
+        .max(500, "id must be at most 500 characters")
+        .describe("Seed paper identifier. Supports DOI, Semantic Scholar ID, ArXiv ID, PubMed ID, OpenAlex ID, or URL."),
+    limit: z
+        .number({ message: "limit must be a number" })
+        .int("limit must be a whole number")
+        .min(1, "limit must be at least 1")
+        .max(100, "limit must be at most 100")
+        .default(30)
+        .describe("Maximum number of similar papers to return"),
+    includeEdges: z
+        .boolean({ message: "includeEdges must be a boolean" })
+        .default(false)
+        .describe("Whether to include citing/cited_by edge data for graph construction"),
+    fields: z
+        .array(z.enum([
+        "doi",
+        "title",
+        "url",
+        "providerURL",
+        "authors",
+        "date",
+        "year",
+        "totalCitations",
+        "totalReferences",
+        "abstract",
+        "pdfURL",
+        "provider",
+        "providerData",
+    ]), {
+        message: "fields must be an array of valid publication field names",
+    })
+        .nullish()
+        .describe("Specific publication fields to return. If omitted, returns all fields except providerData."),
+    callback: z
+        .object({
+        url: z
+            .url()
+            .describe("Webhook URL where results will be POSTed when processing completes"),
+        type: z
+            .string()
+            .optional()
+            .describe("Callback type identifier (e.g. 'zapier')"),
+    })
+        .optional()
+        .describe("Optional webhook callback for async processing. " +
+        "When provided, the server returns 202 immediately and POSTs the full response payload to the callback URL when processing completes. " +
+        "On error, the callback receives a POST with X-Pdfvector-Callback-Failed: true header and error details in the body. " +
+        "Useful for long-running operations that may exceed client timeout limits."),
+}))
+    .output(z.object({
+    seed: publicationSchema,
+    results: z.array(similarPaperSchema),
+    requestId: z.number().int(),
+}))
+    .route({
+    summary: "Find similar papers",
+    description: "Find papers similar to a seed paper using citation network analysis. Returns papers ranked by similarity with optional graph edge data for visualization.",
+    tags: ["Academic"],
+    spec: (op) => {
+        op.security = [{ bearerAuth: [] }];
+        const reqBody = op.requestBody;
+        if (reqBody?.content) {
+            for (const mediaType of Object.values(reqBody.content)) {
+                mediaType.examples = requestExamples;
+            }
+        }
+        return op;
+    },
+});

package/.tsc/lib/router/bankStatement/ask.js

@@ -1,6 +1,7 @@
 import { oc } from "@orpc/contract";
 import { z } from "zod";
 import { pdfvectorModelSchema } from "../../pdfvector-model-schema";
+import { outputAskModelDescription, specializedAskModelDescription, supportedFileFormatsDescription, supportedFileMimes, supportedFileTypesLong, } from "../../supported-mimes";
 import { getDefaultSpec } from "./get-default-spec";
 const specializedModelSchema = z
     .enum(["auto", ...pdfvectorModelSchema.options])
@@ -12,17 +13,9 @@ const askInputSchema = z.object({
         .describe("URL of the bank statement file to fetch and parse"),
     file: z
         .file()
-        .mime([
-        "application/pdf",
-        "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
-        "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
-        "text/csv",
-        "application/csv",
-        "image/png",
-        "image/jpeg",
-    ])
+        .mime([...supportedFileMimes])
         .optional()
-        .describe("Bank statement file upload via multipart form-data (PDF, DOCX, XLSX, CSV, PNG, JPG)"),
+        .describe(`Bank statement file upload via multipart form-data (${supportedFileFormatsDescription})`),
     base64: z
         .string()
         .optional()
@@ -31,12 +24,7 @@ const askInputSchema = z.object({
         .string()
         .min(4, "question must be at least 4 characters")
         .describe("The question to answer about the bank statement"),
-    model: specializedModelSchema.describe("Model tier for answering the question. " +
-        "'auto' (default): Automatically selects the best tier based on document complexity. " +
-        "'nano': 6 credits/page. Fastest and cheapest. Best for simple questions about straightforward documents. Supports PDF, Word, Excel, CSV. " +
-        "'mini': 10 credits/page. Balanced speed and accuracy. Handles moderately complex questions. Supports PDF, Word, Excel, CSV. " +
-        "'pro': 14 credits/page. High accuracy for nuanced questions about complex documents. Supports PDF, Word, Excel, CSV, Image. " +
-        "'max': 18 credits/page. Maximum accuracy with the most capable models. Best for difficult questions requiring deep reasoning. Supports PDF, Word, Excel, CSV, Image."),
+    model: specializedModelSchema.describe(specializedAskModelDescription),
     callback: z
         .object({
         url: z
@@ -56,11 +44,7 @@ const askInputSchema = z.object({
 const askOutputSchema = z.object({
     markdown: z.string().describe("The answer to the question"),
     pageCount: z.number().int().describe("Total number of pages in the document"),
-    model: pdfvectorModelSchema.describe("Model tier used to answer the question. " +
-        "'nano': Fastest, best for simple questions. Supports PDF, Word, Excel, CSV. " +
-        "'mini': Balanced speed and accuracy. Supports PDF, Word, Excel, CSV. " +
-        "'pro': High accuracy for complex questions. Supports PDF, Word, Excel, CSV, Image. " +
-        "'max': Maximum accuracy with deep reasoning. Supports PDF, Word, Excel, CSV, Image."),
+    model: pdfvectorModelSchema.describe(outputAskModelDescription),
     credits: z
         .number()
         .int()
@@ -86,7 +70,7 @@ const requestExamples = {
 export const ask = oc
     .route({
     summary: "Ask a question about a bank statement",
-    description: "Parse a bank statement and answer a question about its content using AI. Supports PDF, Word (.docx), Excel (.xlsx), CSV, and Image (.png, .jpg) files. Provide the document via file upload, a public URL, or a base64-encoded string.",
+    description: `Parse a bank statement and answer a question about its content using AI. Supports ${supportedFileTypesLong}. Provide the document via file upload, a public URL, or a base64-encoded string.`,
     tags: ["Bank Statement"],
     spec: (op) => getDefaultSpec(op, requestExamples),
 })

package/.tsc/lib/router/bankStatement/extract.js

@@ -1,6 +1,7 @@
 import { oc } from "@orpc/contract";
 import { z } from "zod";
 import { pdfvectorModelSchema } from "../../pdfvector-model-schema";
+import { outputExtractModelDescription, specializedExtractModelDescription, supportedFileFormatsDescription, supportedFileMimes, supportedFileTypesLong, } from "../../supported-mimes";
 import { getDefaultSpec } from "./get-default-spec";
 const specializedModelSchema = z
     .enum(["auto", ...pdfvectorModelSchema.options])
@@ -12,17 +13,9 @@ const extractInputSchema = z.object({
         .describe("URL of the bank statement file to fetch and parse"),
     file: z
         .file()
-        .mime([
-        "application/pdf",
-        "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
-        "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
-        "text/csv",
-        "application/csv",
-        "image/png",
-        "image/jpeg",
-    ])
+        .mime([...supportedFileMimes])
         .optional()
-        .describe("Bank statement file upload via multipart form-data (PDF, DOCX, XLSX, CSV, PNG, JPG)"),
+        .describe(`Bank statement file upload via multipart form-data (${supportedFileFormatsDescription})`),
     base64: z
         .string()
         .optional()
@@ -57,12 +50,7 @@ const extractInputSchema = z.object({
         }),
     ])
         .describe("JSON Schema describing the structure of the data to extract from the bank statement. Can be a JSON object or a JSON string."),
-    model: specializedModelSchema.describe("Model tier for extracting structured data. " +
-        "'auto' (default): Automatically selects the best tier based on document complexity. " +
-        "'nano': 6 credits/page. Fastest and cheapest. Best for simple documents with straightforward schemas. Supports PDF, Word, Excel, CSV. " +
-        "'mini': 10 credits/page. Balanced speed and accuracy. Handles moderately complex documents and schemas. Supports PDF, Word, Excel, CSV. " +
-        "'pro': 14 credits/page. High accuracy for complex documents with large or nested schemas. Supports PDF, Word, Excel, CSV, Image. " +
-        "'max': 18 credits/page. Maximum accuracy with the most capable models. Best for difficult extractions requiring deep reasoning. Supports PDF, Word, Excel, CSV, Image."),
+    model: specializedModelSchema.describe(specializedExtractModelDescription),
     callback: z
         .object({
         url: z
@@ -86,11 +74,7 @@ const extractOutputSchema = z.object({
         (typeof val !== "object" || Object.keys(val).length > 0), { message: "Extracted data must not be empty" })
         .describe("Extracted structured data matching the provided JSON Schema"),
     pageCount: z.number().int().describe("Total number of pages in the document"),
-    model: pdfvectorModelSchema.describe("Model tier used to extract the data. " +
-        "'nano': Fastest, best for simple documents. Supports PDF, Word, Excel, CSV. " +
-        "'mini': Balanced speed and accuracy. Supports PDF, Word, Excel, CSV. " +
-        "'pro': High accuracy for complex documents. Supports PDF, Word, Excel, CSV, Image. " +
-        "'max': Maximum accuracy with deep reasoning. Supports PDF, Word, Excel, CSV, Image."),
+    model: pdfvectorModelSchema.describe(outputExtractModelDescription),
     credits: z
         .number()
         .int()
@@ -135,7 +119,7 @@ const requestExamples = {
 export const extract = oc
     .route({
     summary: "Extract structured data from a bank statement",
-    description: "Parse a bank statement and extract structured data matching a provided JSON Schema using AI. Supports PDF, Word (.docx), Excel (.xlsx), CSV, and Image (.png, .jpg) files. Provide the document via file upload, a public URL, or a base64-encoded string.",
+    description: `Parse a bank statement and extract structured data matching a provided JSON Schema using AI. Supports ${supportedFileTypesLong}. Provide the document via file upload, a public URL, or a base64-encoded string.`,
     tags: ["Bank Statement"],
     spec: (op) => getDefaultSpec(op, requestExamples),
 })

package/.tsc/lib/router/bankStatement/parse.js

@@ -1,5 +1,6 @@
 import { oc } from "@orpc/contract";
 import { z } from "zod";
+import { specializedParseModelDescription, supportedFileFormatsDescription, supportedFileMimes, supportedFileTypesLong, } from "../../supported-mimes";
 import { getDefaultSpec } from "./get-default-spec";
 const specializedParseModelSchema = z
     .enum(["pro", "max", "auto"], {
@@ -13,25 +14,14 @@ const parseInputSchema = z.object({
         .describe("URL of the bank statement file to fetch and parse"),
     file: z
         .file()
-        .mime([
-        "application/pdf",
-        "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
-        "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
-        "text/csv",
-        "application/csv",
-        "image/png",
-        "image/jpeg",
-    ])
+        .mime([...supportedFileMimes])
         .optional()
-        .describe("Bank statement file upload via multipart form-data (PDF, DOCX, XLSX, CSV, PNG, JPG)"),
+        .describe(`Bank statement file upload via multipart form-data (${supportedFileFormatsDescription})`),
     base64: z
         .string()
         .optional()
         .describe("Base64-encoded bank statement file content"),
-    model: specializedParseModelSchema.describe("Model tier for parsing. " +
-        "'auto' (default): Automatically selects the best parsing strategy with intelligent fallback. " +
-        "'pro': 6 credits/page. Extracts structured bank statement fields with standard accuracy. " +
-        "'max': 10 credits/page. Extracts structured bank statement fields with highest accuracy and fallback."),
+    model: specializedParseModelSchema.describe(specializedParseModelDescription("bank statement")),
     callback: z
         .object({
         url: z
@@ -97,7 +87,7 @@ const requestExamples = {
 export const parse = oc
     .route({
     summary: "Parse a bank statement",
-    description: "Extract text and structured data from a bank statement. Supports PDF, Word (.docx), Excel (.xlsx), CSV, and Image (.png, .jpg) files. Provide the document via file upload, a public URL, or a base64-encoded string.",
+    description: `Extract text and structured data from a bank statement. Supports ${supportedFileTypesLong}. Provide the document via file upload, a public URL, or a base64-encoded string.`,
     tags: ["Bank Statement"],
     spec: (op) => getDefaultSpec(op, requestExamples),
 })