@juspay/neurolink 9.95.1 → 9.95.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +6 -0
- package/dist/browser/neurolink.min.js +176 -176
- package/dist/cli/factories/commandFactory.d.ts +10 -0
- package/dist/cli/factories/commandFactory.js +30 -0
- package/dist/cli/loop/optionsSchema.d.ts +1 -1
- package/dist/core/baseProvider.js +1 -0
- package/dist/core/constants.d.ts +2 -0
- package/dist/core/constants.js +9 -0
- package/dist/core/modules/MessageBuilder.js +2 -0
- package/dist/lib/core/baseProvider.js +1 -0
- package/dist/lib/core/constants.d.ts +2 -0
- package/dist/lib/core/constants.js +9 -0
- package/dist/lib/core/modules/MessageBuilder.js +2 -0
- package/dist/lib/neurolink.js +8 -4
- package/dist/lib/types/file.d.ts +10 -0
- package/dist/lib/types/generate.d.ts +14 -0
- package/dist/lib/types/stream.d.ts +7 -0
- package/dist/lib/utils/errorHandling.d.ts +10 -0
- package/dist/lib/utils/errorHandling.js +28 -0
- package/dist/lib/utils/messageBuilder.js +36 -5
- package/dist/lib/utils/multimodalOptionsBuilder.d.ts +4 -0
- package/dist/lib/utils/multimodalOptionsBuilder.js +1 -0
- package/dist/lib/utils/pdfProcessor.d.ts +15 -0
- package/dist/lib/utils/pdfProcessor.js +83 -3
- package/dist/neurolink.js +8 -4
- package/dist/types/file.d.ts +10 -0
- package/dist/types/generate.d.ts +14 -0
- package/dist/types/stream.d.ts +7 -0
- package/dist/utils/errorHandling.d.ts +10 -0
- package/dist/utils/errorHandling.js +28 -0
- package/dist/utils/messageBuilder.js +36 -5
- package/dist/utils/multimodalOptionsBuilder.d.ts +4 -0
- package/dist/utils/multimodalOptionsBuilder.js +1 -0
- package/dist/utils/pdfProcessor.d.ts +15 -0
- package/dist/utils/pdfProcessor.js +83 -3
- package/package.json +1 -1
|
@@ -22,6 +22,16 @@ export declare class CLICommandFactory {
|
|
|
22
22
|
private static processCliImages;
|
|
23
23
|
private static processCliCSVFiles;
|
|
24
24
|
private static processCliPDFFiles;
|
|
25
|
+
/**
|
|
26
|
+
* Resolve the PDF decryption password, preferring the NEUROLINK_PDF_PASSWORD
|
|
27
|
+
* env var over the `--pdf-password` flag. A plaintext CLI flag leaks into
|
|
28
|
+
* shell history, `ps`/process listings, and CI logs — the env var avoids
|
|
29
|
+
* that, matching the project's existing "credentials via env vars, not
|
|
30
|
+
* flags" stance for the loop session. The flag stays supported (dropping it
|
|
31
|
+
* would be a breaking CLI change), but using it prints a one-line stderr
|
|
32
|
+
* warning recommending the env var instead.
|
|
33
|
+
*/
|
|
34
|
+
private static resolvePdfPassword;
|
|
25
35
|
private static processCliFiles;
|
|
26
36
|
private static processCliVideoFiles;
|
|
27
37
|
private static processOptions;
|
|
@@ -156,6 +156,12 @@ export class CLICommandFactory {
|
|
|
156
156
|
type: "string",
|
|
157
157
|
description: "Add PDF file for analysis (can be used multiple times)",
|
|
158
158
|
},
|
|
159
|
+
"pdf-password": {
|
|
160
|
+
type: "string",
|
|
161
|
+
description: "Password for an encrypted PDF (used on the image-conversion path). " +
|
|
162
|
+
"Visible in shell history/process listings — prefer the " +
|
|
163
|
+
"NEUROLINK_PDF_PASSWORD env var instead.",
|
|
164
|
+
},
|
|
159
165
|
video: {
|
|
160
166
|
type: "string",
|
|
161
167
|
description: "Add video file for analysis (can be used multiple times) (MP4, WebM, MOV, AVI, MKV)",
|
|
@@ -793,6 +799,24 @@ export class CLICommandFactory {
|
|
|
793
799
|
// URLs are preserved as-is by resolveFilePaths
|
|
794
800
|
return resolveFilePaths(paths);
|
|
795
801
|
}
|
|
802
|
+
/**
|
|
803
|
+
* Resolve the PDF decryption password, preferring the NEUROLINK_PDF_PASSWORD
|
|
804
|
+
* env var over the `--pdf-password` flag. A plaintext CLI flag leaks into
|
|
805
|
+
* shell history, `ps`/process listings, and CI logs — the env var avoids
|
|
806
|
+
* that, matching the project's existing "credentials via env vars, not
|
|
807
|
+
* flags" stance for the loop session. The flag stays supported (dropping it
|
|
808
|
+
* would be a breaking CLI change), but using it prints a one-line stderr
|
|
809
|
+
* warning recommending the env var instead.
|
|
810
|
+
*/
|
|
811
|
+
static resolvePdfPassword(argv) {
|
|
812
|
+
const flagValue = argv.pdfPassword;
|
|
813
|
+
const envValue = process.env.NEUROLINK_PDF_PASSWORD;
|
|
814
|
+
if (flagValue) {
|
|
815
|
+
process.stderr.write(chalk.yellow("⚠️ --pdf-password is visible in shell history and process listings. " +
|
|
816
|
+
"Prefer the NEUROLINK_PDF_PASSWORD environment variable instead.\n"));
|
|
817
|
+
}
|
|
818
|
+
return flagValue || envValue;
|
|
819
|
+
}
|
|
796
820
|
// Helper method to process CLI files with auto-detection
|
|
797
821
|
static processCliFiles(files) {
|
|
798
822
|
if (!files) {
|
|
@@ -2649,6 +2673,9 @@ export class CLICommandFactory {
|
|
|
2649
2673
|
const inputAudioFormat = inferAudioFormatFromPath(inputAudioPath);
|
|
2650
2674
|
const runGenerate = () => sdk.generate({
|
|
2651
2675
|
input: generateInput,
|
|
2676
|
+
pdfOptions: {
|
|
2677
|
+
password: CLICommandFactory.resolvePdfPassword(argv),
|
|
2678
|
+
},
|
|
2652
2679
|
csvOptions: CLICommandFactory.buildCsvOptionsFromArgv(argv),
|
|
2653
2680
|
videoOptions: {
|
|
2654
2681
|
frames: argv.videoFrames,
|
|
@@ -2901,6 +2928,9 @@ export class CLICommandFactory {
|
|
|
2901
2928
|
...(videoFiles && { videoFiles }),
|
|
2902
2929
|
...(files && { files }),
|
|
2903
2930
|
},
|
|
2931
|
+
pdfOptions: {
|
|
2932
|
+
password: CLICommandFactory.resolvePdfPassword(argv),
|
|
2933
|
+
},
|
|
2904
2934
|
csvOptions: CLICommandFactory.buildCsvOptionsFromArgv(argv),
|
|
2905
2935
|
videoOptions: {
|
|
2906
2936
|
frames: argv.videoFrames,
|
|
@@ -4,4 +4,4 @@ import type { OptionSchema, TextGenerationOptions } from "../../lib/types/index.
|
|
|
4
4
|
* This object provides metadata for validation and help text in the CLI loop.
|
|
5
5
|
* It is derived from the main TextGenerationOptions interface to ensure consistency.
|
|
6
6
|
*/
|
|
7
|
-
export declare const textGenerationOptionsSchema: Record<keyof Omit<TextGenerationOptions, "prompt" | "input" | "schema" | "tools" | "context" | "conversationHistory" | "conversationMessages" | "conversationMemoryConfig" | "originalPrompt" | "middleware" | "expectedOutcome" | "evaluationCriteria" | "region" | "csvOptions" | "tts" | "stt" | "thinkingConfig" | "requestId" | "fileRegistry" | "abortSignal" | "toolFilter" | "excludeTools" | "toolChoice" | "prepareStep" | "credentials" | "onFinish" | "onError" | "processors" | "piiDetection" | "responseValidation" | "inputValidation">, OptionSchema>;
|
|
7
|
+
export declare const textGenerationOptionsSchema: Record<keyof Omit<TextGenerationOptions, "prompt" | "input" | "schema" | "tools" | "context" | "conversationHistory" | "conversationMessages" | "conversationMemoryConfig" | "originalPrompt" | "middleware" | "expectedOutcome" | "evaluationCriteria" | "region" | "csvOptions" | "pdfOptions" | "tts" | "stt" | "thinkingConfig" | "requestId" | "fileRegistry" | "abortSignal" | "toolFilter" | "excludeTools" | "toolChoice" | "prepareStep" | "credentials" | "onFinish" | "onError" | "processors" | "piiDetection" | "responseValidation" | "inputValidation">, OptionSchema>;
|
|
@@ -420,6 +420,7 @@ export class BaseProvider {
|
|
|
420
420
|
toolUsageContext: options.toolUsageContext,
|
|
421
421
|
context: options.context,
|
|
422
422
|
csvOptions: options.csvOptions,
|
|
423
|
+
pdfOptions: options.pdfOptions,
|
|
423
424
|
// Forward abort, tool filtering, and timeout options to prevent
|
|
424
425
|
// silent bypass when falling back from real streaming to fake streaming
|
|
425
426
|
abortSignal: options.abortSignal,
|
package/dist/core/constants.d.ts
CHANGED
|
@@ -122,6 +122,8 @@ export declare const PDF_LIMITS: {
|
|
|
122
122
|
MAX_SIZE_MB: number;
|
|
123
123
|
DEFAULT_MAX_PAGES: number;
|
|
124
124
|
MAX_SCALE: number;
|
|
125
|
+
DEFAULT_MAX_CANVAS_PIXELS: number;
|
|
126
|
+
MIN_EFFECTIVE_SCALE: number;
|
|
125
127
|
};
|
|
126
128
|
export declare const SYSTEM_LIMITS: {
|
|
127
129
|
MAX_PROMPT_LENGTH: number;
|
package/dist/core/constants.js
CHANGED
|
@@ -220,6 +220,15 @@ export const PDF_LIMITS = {
|
|
|
220
220
|
// Upper bound for the render scale factor. Above this, a single page can
|
|
221
221
|
// allocate hundreds of MB; scale <= 0 produces a degenerate viewport.
|
|
222
222
|
MAX_SCALE: 10,
|
|
223
|
+
// Default per-page pixel ceiling for image conversion (#260). 16.7M px ×
|
|
224
|
+
// 4 bytes RGBA ≈ 64 MB per page — safely bounded. A larger page is
|
|
225
|
+
// uniformly downscaled to stay under this instead of allocating gigabytes.
|
|
226
|
+
DEFAULT_MAX_CANVAS_PIXELS: 16_777_216,
|
|
227
|
+
// Floor for the downscaled render scale (#260 follow-up). A crafted/malformed
|
|
228
|
+
// MediaBox can drive the estimated pixel count toward Infinity, which would
|
|
229
|
+
// otherwise collapse `effectiveScale` to 0 and hand `pdf-to-img` a degenerate
|
|
230
|
+
// viewport. Never let the downscale branch go below this.
|
|
231
|
+
MIN_EFFECTIVE_SCALE: 0.1,
|
|
223
232
|
};
|
|
224
233
|
// Performance and System Limits
|
|
225
234
|
export const SYSTEM_LIMITS = {
|
|
@@ -103,6 +103,7 @@ export class MessageBuilder {
|
|
|
103
103
|
files: input?.files,
|
|
104
104
|
},
|
|
105
105
|
csvOptions: options.csvOptions,
|
|
106
|
+
pdfOptions: options.pdfOptions,
|
|
106
107
|
provider: options.provider,
|
|
107
108
|
model: options.model,
|
|
108
109
|
temperature: options.temperature,
|
|
@@ -224,6 +225,7 @@ export class MessageBuilder {
|
|
|
224
225
|
files: input?.files,
|
|
225
226
|
},
|
|
226
227
|
csvOptions: options.csvOptions,
|
|
228
|
+
pdfOptions: options.pdfOptions,
|
|
227
229
|
provider: options.provider,
|
|
228
230
|
model: options.model,
|
|
229
231
|
temperature: options.temperature,
|
|
@@ -420,6 +420,7 @@ export class BaseProvider {
|
|
|
420
420
|
toolUsageContext: options.toolUsageContext,
|
|
421
421
|
context: options.context,
|
|
422
422
|
csvOptions: options.csvOptions,
|
|
423
|
+
pdfOptions: options.pdfOptions,
|
|
423
424
|
// Forward abort, tool filtering, and timeout options to prevent
|
|
424
425
|
// silent bypass when falling back from real streaming to fake streaming
|
|
425
426
|
abortSignal: options.abortSignal,
|
|
@@ -122,6 +122,8 @@ export declare const PDF_LIMITS: {
|
|
|
122
122
|
MAX_SIZE_MB: number;
|
|
123
123
|
DEFAULT_MAX_PAGES: number;
|
|
124
124
|
MAX_SCALE: number;
|
|
125
|
+
DEFAULT_MAX_CANVAS_PIXELS: number;
|
|
126
|
+
MIN_EFFECTIVE_SCALE: number;
|
|
125
127
|
};
|
|
126
128
|
export declare const SYSTEM_LIMITS: {
|
|
127
129
|
MAX_PROMPT_LENGTH: number;
|
|
@@ -220,6 +220,15 @@ export const PDF_LIMITS = {
|
|
|
220
220
|
// Upper bound for the render scale factor. Above this, a single page can
|
|
221
221
|
// allocate hundreds of MB; scale <= 0 produces a degenerate viewport.
|
|
222
222
|
MAX_SCALE: 10,
|
|
223
|
+
// Default per-page pixel ceiling for image conversion (#260). 16.7M px ×
|
|
224
|
+
// 4 bytes RGBA ≈ 64 MB per page — safely bounded. A larger page is
|
|
225
|
+
// uniformly downscaled to stay under this instead of allocating gigabytes.
|
|
226
|
+
DEFAULT_MAX_CANVAS_PIXELS: 16_777_216,
|
|
227
|
+
// Floor for the downscaled render scale (#260 follow-up). A crafted/malformed
|
|
228
|
+
// MediaBox can drive the estimated pixel count toward Infinity, which would
|
|
229
|
+
// otherwise collapse `effectiveScale` to 0 and hand `pdf-to-img` a degenerate
|
|
230
|
+
// viewport. Never let the downscale branch go below this.
|
|
231
|
+
MIN_EFFECTIVE_SCALE: 0.1,
|
|
223
232
|
};
|
|
224
233
|
// Performance and System Limits
|
|
225
234
|
export const SYSTEM_LIMITS = {
|
|
@@ -103,6 +103,7 @@ export class MessageBuilder {
|
|
|
103
103
|
files: input?.files,
|
|
104
104
|
},
|
|
105
105
|
csvOptions: options.csvOptions,
|
|
106
|
+
pdfOptions: options.pdfOptions,
|
|
106
107
|
provider: options.provider,
|
|
107
108
|
model: options.model,
|
|
108
109
|
temperature: options.temperature,
|
|
@@ -224,6 +225,7 @@ export class MessageBuilder {
|
|
|
224
225
|
files: input?.files,
|
|
225
226
|
},
|
|
226
227
|
csvOptions: options.csvOptions,
|
|
228
|
+
pdfOptions: options.pdfOptions,
|
|
227
229
|
provider: options.provider,
|
|
228
230
|
model: options.model,
|
|
229
231
|
temperature: options.temperature,
|
package/dist/lib/neurolink.js
CHANGED
|
@@ -3899,11 +3899,15 @@ Current user's request: ${currentInput}`;
|
|
|
3899
3899
|
evaluationDomain: options.evaluationDomain,
|
|
3900
3900
|
toolUsageContext: options.toolUsageContext,
|
|
3901
3901
|
input: options.input,
|
|
3902
|
-
//
|
|
3903
|
-
// TextGenerationOptions — the message builder reads them
|
|
3904
|
-
// formatting
|
|
3905
|
-
//
|
|
3902
|
+
// Multimodal file-processing options must survive the reconstruction
|
|
3903
|
+
// into TextGenerationOptions — the message builder reads them downstream
|
|
3904
|
+
// (csv encoding/formatting/sanitization/parse timeout, PDF
|
|
3905
|
+
// password/decryption #258). Omitting them here silently dropped e.g.
|
|
3906
|
+
// every csvOptions field (incl. the CLI --csv-* flags) or
|
|
3907
|
+
// pdfOptions.password so the CLI --pdf-password / SDK pdfOptions never
|
|
3908
|
+
// reached convertToImages.
|
|
3906
3909
|
csvOptions: options.csvOptions,
|
|
3910
|
+
pdfOptions: options.pdfOptions,
|
|
3907
3911
|
region: options.region,
|
|
3908
3912
|
tts: options.tts,
|
|
3909
3913
|
stt: options.stt,
|
package/dist/lib/types/file.d.ts
CHANGED
|
@@ -226,6 +226,8 @@ export type PDFProcessorOptions = {
|
|
|
226
226
|
* Set to false to bypass limit enforcement (logs warning instead)
|
|
227
227
|
*/
|
|
228
228
|
enforceLimits?: boolean;
|
|
229
|
+
/** Password for an encrypted PDF (used on the image-conversion path) (#258). */
|
|
230
|
+
password?: string;
|
|
229
231
|
};
|
|
230
232
|
/**
|
|
231
233
|
* Audio provider configuration for transcription services
|
|
@@ -395,6 +397,14 @@ export type PDFImageConversionOptions = {
|
|
|
395
397
|
maxPages?: number;
|
|
396
398
|
/** Output format (default: png). Only PNG is currently implemented by PDFProcessor. */
|
|
397
399
|
format?: "png";
|
|
400
|
+
/**
|
|
401
|
+
* Per-page pixel ceiling (#260). Any page whose width×height×scale² would
|
|
402
|
+
* exceed this is uniformly downscaled to stay under it, preventing a huge
|
|
403
|
+
* page from allocating gigabytes of canvas. Default: PDF_LIMITS.DEFAULT_MAX_CANVAS_PIXELS.
|
|
404
|
+
*/
|
|
405
|
+
maxCanvasPixels?: number;
|
|
406
|
+
/** Password for an encrypted PDF (passed to the underlying renderer) (#258). */
|
|
407
|
+
password?: string;
|
|
398
408
|
};
|
|
399
409
|
/** Result of PDF to image conversion. */
|
|
400
410
|
export type PDFImageConversionResult = {
|
|
@@ -125,6 +125,13 @@ export type GenerateOptions = {
|
|
|
125
125
|
music?: MusicOptions;
|
|
126
126
|
};
|
|
127
127
|
csvOptions?: CSVProcessorOptions;
|
|
128
|
+
/** PDF processing options (#258). */
|
|
129
|
+
pdfOptions?: {
|
|
130
|
+
/** Password for an encrypted PDF (image-conversion fallback path). */
|
|
131
|
+
password?: string;
|
|
132
|
+
/** Max rendered-canvas pixels per page (#260 memory guard); oversized pages auto-downscale. */
|
|
133
|
+
maxCanvasPixels?: number;
|
|
134
|
+
};
|
|
128
135
|
videoOptions?: {
|
|
129
136
|
frames?: number;
|
|
130
137
|
quality?: number;
|
|
@@ -1131,6 +1138,13 @@ export type TextGenerationOptions = {
|
|
|
1131
1138
|
expectedOutcome?: string;
|
|
1132
1139
|
evaluationCriteria?: string[];
|
|
1133
1140
|
csvOptions?: CSVProcessorOptions;
|
|
1141
|
+
/** PDF processing options (#258). */
|
|
1142
|
+
pdfOptions?: {
|
|
1143
|
+
/** Password for an encrypted PDF (image-conversion fallback path). */
|
|
1144
|
+
password?: string;
|
|
1145
|
+
/** Max rendered-canvas pixels per page (#260 memory guard); oversized pages auto-downscale. */
|
|
1146
|
+
maxCanvasPixels?: number;
|
|
1147
|
+
};
|
|
1134
1148
|
enableSummarization?: boolean;
|
|
1135
1149
|
/**
|
|
1136
1150
|
* Skip injecting tool schemas into the system prompt.
|
|
@@ -207,6 +207,13 @@ export type StreamOptions = {
|
|
|
207
207
|
};
|
|
208
208
|
};
|
|
209
209
|
csvOptions?: CSVProcessorOptions;
|
|
210
|
+
/** PDF processing options (#258). */
|
|
211
|
+
pdfOptions?: {
|
|
212
|
+
/** Password for an encrypted PDF (image-conversion fallback path). */
|
|
213
|
+
password?: string;
|
|
214
|
+
/** Max rendered-canvas pixels per page (#260 memory guard); oversized pages auto-downscale. */
|
|
215
|
+
maxCanvasPixels?: number;
|
|
216
|
+
};
|
|
210
217
|
videoOptions?: {
|
|
211
218
|
frames?: number;
|
|
212
219
|
quality?: number;
|
|
@@ -39,6 +39,8 @@ export declare const ERROR_CODES: {
|
|
|
39
39
|
readonly FILE_PROCESSING_FAILED: "FILE_PROCESSING_FAILED";
|
|
40
40
|
readonly CSV_PROCESSING_FAILED: "CSV_PROCESSING_FAILED";
|
|
41
41
|
readonly PDF_PAGE_LIMIT_EXCEEDED: "PDF_PAGE_LIMIT_EXCEEDED";
|
|
42
|
+
readonly PDF_PASSWORD_REQUIRED: "PDF_PASSWORD_REQUIRED";
|
|
43
|
+
readonly PDF_INCORRECT_PASSWORD: "PDF_INCORRECT_PASSWORD";
|
|
42
44
|
readonly RATE_LIMITER_QUEUE_FULL: "RATE_LIMITER_QUEUE_FULL";
|
|
43
45
|
readonly RATE_LIMITER_QUEUE_TIMEOUT: "RATE_LIMITER_QUEUE_TIMEOUT";
|
|
44
46
|
readonly RATE_LIMITER_RESET: "RATE_LIMITER_RESET";
|
|
@@ -199,6 +201,14 @@ export declare class ErrorFactory {
|
|
|
199
201
|
* Create a PDF page limit exceeded error
|
|
200
202
|
*/
|
|
201
203
|
static pdfPageLimitExceeded(estimatedPages: number, maxPages: number, provider: string): NeuroLinkError;
|
|
204
|
+
/**
|
|
205
|
+
* The PDF is encrypted and no password was supplied (#258).
|
|
206
|
+
*/
|
|
207
|
+
static pdfPasswordRequired(): NeuroLinkError;
|
|
208
|
+
/**
|
|
209
|
+
* A password was supplied for an encrypted PDF but it was incorrect (#258).
|
|
210
|
+
*/
|
|
211
|
+
static pdfIncorrectPassword(): NeuroLinkError;
|
|
202
212
|
/**
|
|
203
213
|
* Create an image too large error
|
|
204
214
|
*/
|
|
@@ -52,6 +52,8 @@ export const ERROR_CODES = {
|
|
|
52
52
|
CSV_PROCESSING_FAILED: "CSV_PROCESSING_FAILED",
|
|
53
53
|
// PDF validation errors
|
|
54
54
|
PDF_PAGE_LIMIT_EXCEEDED: "PDF_PAGE_LIMIT_EXCEEDED",
|
|
55
|
+
PDF_PASSWORD_REQUIRED: "PDF_PASSWORD_REQUIRED",
|
|
56
|
+
PDF_INCORRECT_PASSWORD: "PDF_INCORRECT_PASSWORD",
|
|
55
57
|
// Rate limiter errors
|
|
56
58
|
RATE_LIMITER_QUEUE_FULL: "RATE_LIMITER_QUEUE_FULL",
|
|
57
59
|
RATE_LIMITER_QUEUE_TIMEOUT: "RATE_LIMITER_QUEUE_TIMEOUT",
|
|
@@ -553,6 +555,32 @@ export class ErrorFactory {
|
|
|
553
555
|
},
|
|
554
556
|
});
|
|
555
557
|
}
|
|
558
|
+
/**
|
|
559
|
+
* The PDF is encrypted and no password was supplied (#258).
|
|
560
|
+
*/
|
|
561
|
+
static pdfPasswordRequired() {
|
|
562
|
+
return new NeuroLinkError({
|
|
563
|
+
code: ERROR_CODES.PDF_PASSWORD_REQUIRED,
|
|
564
|
+
message: "This PDF is password-protected. Supply the password via " +
|
|
565
|
+
"`pdfOptions: { password: '…' }` (SDK) or `--pdf-password` (CLI).",
|
|
566
|
+
category: ErrorCategory.VALIDATION,
|
|
567
|
+
severity: ErrorSeverity.MEDIUM,
|
|
568
|
+
retriable: false,
|
|
569
|
+
});
|
|
570
|
+
}
|
|
571
|
+
/**
|
|
572
|
+
* A password was supplied for an encrypted PDF but it was incorrect (#258).
|
|
573
|
+
*/
|
|
574
|
+
static pdfIncorrectPassword() {
|
|
575
|
+
return new NeuroLinkError({
|
|
576
|
+
code: ERROR_CODES.PDF_INCORRECT_PASSWORD,
|
|
577
|
+
message: "The password supplied for this PDF is incorrect. Check the " +
|
|
578
|
+
"`pdfOptions.password` / `--pdf-password` value and try again.",
|
|
579
|
+
category: ErrorCategory.VALIDATION,
|
|
580
|
+
severity: ErrorSeverity.MEDIUM,
|
|
581
|
+
retriable: false,
|
|
582
|
+
});
|
|
583
|
+
}
|
|
556
584
|
/**
|
|
557
585
|
* Create an image too large error
|
|
558
586
|
*/
|
|
@@ -1011,6 +1011,12 @@ async function processExplicitPdfFiles(options, maxSize, provider) {
|
|
|
1011
1011
|
buffer: result.content,
|
|
1012
1012
|
filename,
|
|
1013
1013
|
pageCount: result.metadata?.estimatedPages ?? null,
|
|
1014
|
+
// #258: carry the password so the image-fallback conversion can
|
|
1015
|
+
// decrypt an encrypted PDF for providers without native PDF support.
|
|
1016
|
+
password: options.pdfOptions?.password,
|
|
1017
|
+
// #260: carry the per-page canvas-pixel ceiling so the caller can
|
|
1018
|
+
// raise (or lower) the memory guard for the image-fallback render.
|
|
1019
|
+
maxCanvasPixels: options.pdfOptions?.maxCanvasPixels,
|
|
1014
1020
|
});
|
|
1015
1021
|
logger.info(`[PDF] ✅ Queued for multimodal: ${filename} (${result.metadata?.estimatedPages ?? "unknown"} pages)`);
|
|
1016
1022
|
}
|
|
@@ -1103,7 +1109,13 @@ export async function buildMultimodalMessagesArray(options, provider, model) {
|
|
|
1103
1109
|
// Check if this is a multimodal request
|
|
1104
1110
|
const hasImages = (inp.images && inp.images.length > 0) ||
|
|
1105
1111
|
(inp.content && inp.content.some((c) => c.type === "image"));
|
|
1106
|
-
|
|
1112
|
+
// A PDF supplied only via input.content (type: "pdf", no explicit
|
|
1113
|
+
// input.pdfFiles and no image alongside it) must still route through the
|
|
1114
|
+
// multimodal path below — otherwise it silently falls through to the
|
|
1115
|
+
// text-only branch and the PDF (and any pdfOptions) never reaches
|
|
1116
|
+
// convertContentToProviderFormat at all.
|
|
1117
|
+
const hasPDFs = pdfFiles.length > 0 ||
|
|
1118
|
+
!!(inp.content && inp.content.some((c) => c.type === "pdf"));
|
|
1107
1119
|
// If no images or PDFs, use standard message building and convert to MultimodalChatMessage[]
|
|
1108
1120
|
if (!hasImages && !hasPDFs) {
|
|
1109
1121
|
// #289: CSV content[] items don't need vision, so they never reach the
|
|
@@ -1139,8 +1151,11 @@ export async function buildMultimodalMessagesArray(options, provider, model) {
|
|
|
1139
1151
|
`Supported providers: ${ProviderImageAdapter.getVisionProviders().join(", ")}`);
|
|
1140
1152
|
}
|
|
1141
1153
|
const messages = [];
|
|
1142
|
-
// Build enhanced system prompt
|
|
1143
|
-
|
|
1154
|
+
// Build enhanced system prompt. Gate on the same `hasPDFs` predicate used
|
|
1155
|
+
// for routing above — a PDF supplied only via input.content (no explicit
|
|
1156
|
+
// input.pdfFiles) must still get the "treat inlined content as an
|
|
1157
|
+
// attachment" instruction, or the model can claim no files were attached.
|
|
1158
|
+
const systemPrompt = buildMultimodalSystemPrompt(options, hasPDFs);
|
|
1144
1159
|
if (systemPrompt.trim()) {
|
|
1145
1160
|
messages.push({
|
|
1146
1161
|
role: "system",
|
|
@@ -1206,7 +1221,7 @@ export async function buildMultimodalMessagesArray(options, provider, model) {
|
|
|
1206
1221
|
try {
|
|
1207
1222
|
let userContent;
|
|
1208
1223
|
if (inp.content && inp.content.length > 0) {
|
|
1209
|
-
userContent = await convertContentToProviderFormat(inp.content, provider, model);
|
|
1224
|
+
userContent = await convertContentToProviderFormat(inp.content, provider, model, options.pdfOptions);
|
|
1210
1225
|
}
|
|
1211
1226
|
else if ((inp.images && inp.images.length > 0) || pdfFiles.length > 0) {
|
|
1212
1227
|
userContent = await convertMultimodalToProviderFormat(inp.text ?? "", inp.images || [], pdfFiles, provider, model);
|
|
@@ -1285,7 +1300,7 @@ async function appendCsvContentToText(csvItems, baseText) {
|
|
|
1285
1300
|
/**
|
|
1286
1301
|
* Convert advanced content format to provider-specific format
|
|
1287
1302
|
*/
|
|
1288
|
-
async function convertContentToProviderFormat(content, provider, _model) {
|
|
1303
|
+
async function convertContentToProviderFormat(content, provider, _model, pdfOptions) {
|
|
1289
1304
|
const textContent = content.find((c) => c.type === "text");
|
|
1290
1305
|
const imageContent = content.filter((c) => c.type === "image");
|
|
1291
1306
|
const pdfContent = content.filter((c) => c.type === "pdf");
|
|
@@ -1312,6 +1327,11 @@ async function convertContentToProviderFormat(content, provider, _model) {
|
|
|
1312
1327
|
buffer: typeof pdf.data === "string" ? Buffer.from(pdf.data, "base64") : pdf.data,
|
|
1313
1328
|
filename: pdf.metadata?.filename || "document.pdf",
|
|
1314
1329
|
pageCount: pdf.metadata?.pages ?? null,
|
|
1330
|
+
// #258/#260: carry password + canvas-pixel ceiling so a PDF supplied via
|
|
1331
|
+
// the advanced `input.content` array gets the same decryption/memory
|
|
1332
|
+
// guard as the `input.pdfFiles` path (see `processExplicitPdfFiles`).
|
|
1333
|
+
password: pdfOptions?.password,
|
|
1334
|
+
maxCanvasPixels: pdfOptions?.maxCanvasPixels,
|
|
1315
1335
|
}));
|
|
1316
1336
|
return await convertMultimodalToProviderFormat(text, images, pdfFiles, provider, _model);
|
|
1317
1337
|
}
|
|
@@ -1694,6 +1714,10 @@ async function convertMultimodalToProviderFormat(text, images, pdfFiles, provide
|
|
|
1694
1714
|
const conversionResult = await PDFImageConverter.convertToImages(pdf.buffer, {
|
|
1695
1715
|
scale: 2.0, // High quality for OCR/analysis
|
|
1696
1716
|
maxPages: 20, // Limit pages to prevent token overflow
|
|
1717
|
+
...(pdf.password ? { password: pdf.password } : {}), // #258
|
|
1718
|
+
...(pdf.maxCanvasPixels
|
|
1719
|
+
? { maxCanvasPixels: pdf.maxCanvasPixels }
|
|
1720
|
+
: {}), // #260
|
|
1697
1721
|
});
|
|
1698
1722
|
logger.info(`[PDF→Image] ✅ Converted ${pdf.filename}: ${conversionResult.pageCount} page(s) → images`);
|
|
1699
1723
|
// Add each page as an ImagePart (raw base64, not data: URI — see SSRF note above)
|
|
@@ -1715,6 +1739,13 @@ async function convertMultimodalToProviderFormat(text, images, pdfFiles, provide
|
|
|
1715
1739
|
catch (error) {
|
|
1716
1740
|
const errorMessage = error instanceof Error ? error.message : String(error);
|
|
1717
1741
|
logger.error(`[PDF→Image] ❌ Failed to convert ${pdf.filename}: ${errorMessage}`);
|
|
1742
|
+
// #258: password errors are already actionable typed errors — re-throw
|
|
1743
|
+
// them unwrapped so the "supply a password" guidance isn't buried.
|
|
1744
|
+
const code = error?.code;
|
|
1745
|
+
if (code === "PDF_PASSWORD_REQUIRED" ||
|
|
1746
|
+
code === "PDF_INCORRECT_PASSWORD") {
|
|
1747
|
+
throw error;
|
|
1748
|
+
}
|
|
1718
1749
|
// Re-throw so the user knows PDF processing failed
|
|
1719
1750
|
throw new Error(`PDF to image conversion failed for ${pdf.filename}: ${errorMessage}. ` +
|
|
1720
1751
|
`Provider ${provider} doesn't support native PDFs and image conversion failed.`, { cause: error });
|
|
@@ -51,6 +51,10 @@ export declare function buildMultimodalOptions(options: StreamOptions, providerN
|
|
|
51
51
|
pdfFiles: (string | Buffer<ArrayBufferLike>)[] | undefined;
|
|
52
52
|
};
|
|
53
53
|
csvOptions: import("../types/file.js").CSVProcessorOptions | undefined;
|
|
54
|
+
pdfOptions: {
|
|
55
|
+
password?: string;
|
|
56
|
+
maxCanvasPixels?: number;
|
|
57
|
+
} | undefined;
|
|
54
58
|
systemPrompt: string | undefined;
|
|
55
59
|
conversationHistory: import("../types/conversation.js").ChatMessage[] | undefined;
|
|
56
60
|
provider: string;
|
|
@@ -51,6 +51,7 @@ export function buildMultimodalOptions(options, providerName, modelName) {
|
|
|
51
51
|
pdfFiles: options.input?.pdfFiles,
|
|
52
52
|
},
|
|
53
53
|
csvOptions: options.csvOptions,
|
|
54
|
+
pdfOptions: options.pdfOptions,
|
|
54
55
|
systemPrompt: options.systemPrompt,
|
|
55
56
|
conversationHistory: options.conversationMessages,
|
|
56
57
|
provider: providerName,
|
|
@@ -21,6 +21,21 @@ export declare class PDFProcessor {
|
|
|
21
21
|
private static isValidPDF;
|
|
22
22
|
private static extractBasicMetadata;
|
|
23
23
|
static estimateTokens(pageCount: number, mode?: "text-only" | "visual"): number;
|
|
24
|
+
/**
|
|
25
|
+
* Estimate the largest page's rendered pixel count from the PDF's MediaBox
|
|
26
|
+
* entries, WITHOUT rendering (#260). Parses `/MediaBox [llx lly urx ury]`
|
|
27
|
+
* (dimensions in points); rendered pixels ≈ (width·scale)·(height·scale).
|
|
28
|
+
* Returns 0 when no plaintext MediaBox is found (e.g. compressed object
|
|
29
|
+
* streams) — the caller then skips downscaling, matching prior behavior.
|
|
30
|
+
*
|
|
31
|
+
* A crafted/malformed MediaBox can have finite but astronomically large
|
|
32
|
+
* width/height; `width * height * scale * scale` can then overflow past
|
|
33
|
+
* `Number.MAX_VALUE` to `Infinity`. Clamp the product to
|
|
34
|
+
* `Number.MAX_SAFE_INTEGER` so it stays finite — a large-but-finite pixel
|
|
35
|
+
* estimate still triggers downscaling, whereas `Infinity` would collapse
|
|
36
|
+
* the computed downscale factor to 0 (see `convertToImages`).
|
|
37
|
+
*/
|
|
38
|
+
private static largestPagePixels;
|
|
24
39
|
/**
|
|
25
40
|
* Convert a PDF buffer to an array of base64 PNG images
|
|
26
41
|
*
|
|
@@ -231,6 +231,40 @@ export class PDFProcessor {
|
|
|
231
231
|
// ============================================================================
|
|
232
232
|
// PDF → Image Conversion (for providers without native PDF support)
|
|
233
233
|
// ============================================================================
|
|
234
|
+
/**
|
|
235
|
+
* Estimate the largest page's rendered pixel count from the PDF's MediaBox
|
|
236
|
+
* entries, WITHOUT rendering (#260). Parses `/MediaBox [llx lly urx ury]`
|
|
237
|
+
* (dimensions in points); rendered pixels ≈ (width·scale)·(height·scale).
|
|
238
|
+
* Returns 0 when no plaintext MediaBox is found (e.g. compressed object
|
|
239
|
+
* streams) — the caller then skips downscaling, matching prior behavior.
|
|
240
|
+
*
|
|
241
|
+
* A crafted/malformed MediaBox can have finite but astronomically large
|
|
242
|
+
* width/height; `width * height * scale * scale` can then overflow past
|
|
243
|
+
* `Number.MAX_VALUE` to `Infinity`. Clamp the product to
|
|
244
|
+
* `Number.MAX_SAFE_INTEGER` so it stays finite — a large-but-finite pixel
|
|
245
|
+
* estimate still triggers downscaling, whereas `Infinity` would collapse
|
|
246
|
+
* the computed downscale factor to 0 (see `convertToImages`).
|
|
247
|
+
*/
|
|
248
|
+
static largestPagePixels(pdfBuffer, scale) {
|
|
249
|
+
const text = pdfBuffer.toString("latin1");
|
|
250
|
+
const re = /\/MediaBox\s*\[\s*(-?[\d.]+)\s+(-?[\d.]+)\s+(-?[\d.]+)\s+(-?[\d.]+)\s*\]/g;
|
|
251
|
+
let max = 0;
|
|
252
|
+
let match;
|
|
253
|
+
while ((match = re.exec(text)) !== null) {
|
|
254
|
+
const width = Math.abs(parseFloat(match[3]) - parseFloat(match[1]));
|
|
255
|
+
const height = Math.abs(parseFloat(match[4]) - parseFloat(match[2]));
|
|
256
|
+
if (Number.isFinite(width) && Number.isFinite(height)) {
|
|
257
|
+
const pixels = width * height * scale * scale;
|
|
258
|
+
const boundedPixels = Number.isFinite(pixels)
|
|
259
|
+
? Math.min(pixels, Number.MAX_SAFE_INTEGER)
|
|
260
|
+
: Number.MAX_SAFE_INTEGER;
|
|
261
|
+
if (boundedPixels > max) {
|
|
262
|
+
max = boundedPixels;
|
|
263
|
+
}
|
|
264
|
+
}
|
|
265
|
+
}
|
|
266
|
+
return max;
|
|
267
|
+
}
|
|
234
268
|
/**
|
|
235
269
|
* Convert a PDF buffer to an array of base64 PNG images
|
|
236
270
|
*
|
|
@@ -257,7 +291,7 @@ export class PDFProcessor {
|
|
|
257
291
|
*/
|
|
258
292
|
static async convertToImages(pdfBuffer, options) {
|
|
259
293
|
const startTime = Date.now();
|
|
260
|
-
const { scale = 2, maxPages = PDF_LIMITS.DEFAULT_MAX_PAGES, format = "png", } = options || {};
|
|
294
|
+
const { scale = 2, maxPages = PDF_LIMITS.DEFAULT_MAX_PAGES, format = "png", maxCanvasPixels = PDF_LIMITS.DEFAULT_MAX_CANVAS_PIXELS, password, } = options || {};
|
|
261
295
|
const images = [];
|
|
262
296
|
const warnings = [];
|
|
263
297
|
// ============================================================================
|
|
@@ -273,6 +307,11 @@ export class PDFProcessor {
|
|
|
273
307
|
if (!Number.isFinite(scale) || scale <= 0 || scale > PDF_LIMITS.MAX_SCALE) {
|
|
274
308
|
throw new Error(`Invalid scale: ${scale}. Scale must be a finite number greater than 0 and at most ${PDF_LIMITS.MAX_SCALE}.`);
|
|
275
309
|
}
|
|
310
|
+
// 0c. Validate the per-page pixel ceiling (#260). A non-finite or
|
|
311
|
+
// non-positive value would disable the guard or yield a NaN downscale.
|
|
312
|
+
if (!Number.isFinite(maxCanvasPixels) || maxCanvasPixels <= 0) {
|
|
313
|
+
throw new Error(`Invalid maxCanvasPixels: ${maxCanvasPixels}. Must be a finite number greater than 0.`);
|
|
314
|
+
}
|
|
276
315
|
// 1. Validate buffer is not empty or too small
|
|
277
316
|
if (!pdfBuffer || pdfBuffer.length < 5) {
|
|
278
317
|
throw new Error("Invalid PDF: Buffer is too small or empty. " +
|
|
@@ -303,8 +342,38 @@ export class PDFProcessor {
|
|
|
303
342
|
scale,
|
|
304
343
|
maxPages: maxPages || "all",
|
|
305
344
|
});
|
|
306
|
-
//
|
|
307
|
-
|
|
345
|
+
// #260: pre-flight page-size check WITHOUT rendering. pdf-to-img applies
|
|
346
|
+
// `scale` uniformly with no per-page hook and no pixel guard, so a very
|
|
347
|
+
// large page (e.g. an architectural drawing with a huge MediaBox) can
|
|
348
|
+
// allocate gigabytes of canvas. Read the largest MediaBox from the PDF
|
|
349
|
+
// bytes and, if that page at the requested scale would exceed
|
|
350
|
+
// maxCanvasPixels, downscale the whole render uniformly to stay under it.
|
|
351
|
+
let effectiveScale = scale;
|
|
352
|
+
const largestPixels = PDFProcessor.largestPagePixels(pdfBuffer, scale);
|
|
353
|
+
if (largestPixels > maxCanvasPixels) {
|
|
354
|
+
const downscale = Math.sqrt(maxCanvasPixels / largestPixels);
|
|
355
|
+
// Floor the result: an astronomically large (but now finite, see
|
|
356
|
+
// `largestPagePixels`) pixel estimate would otherwise push `downscale`
|
|
357
|
+
// — and therefore `effectiveScale` — toward 0, handing `pdf-to-img` a
|
|
358
|
+
// degenerate viewport instead of a small-but-renderable page.
|
|
359
|
+
effectiveScale = Math.max(PDF_LIMITS.MIN_EFFECTIVE_SCALE, scale * downscale);
|
|
360
|
+
// Recompute the ratio actually applied (may differ from `downscale`
|
|
361
|
+
// when the floor above kicks in) so the logged estimate stays honest.
|
|
362
|
+
const actualDownscale = effectiveScale / scale;
|
|
363
|
+
const beforeMB = (largestPixels * 4) / (1024 * 1024);
|
|
364
|
+
const afterMB = (largestPixels * actualDownscale * actualDownscale * 4) /
|
|
365
|
+
(1024 * 1024);
|
|
366
|
+
const msg = `Downscaled render (scale ${scale} → ${effectiveScale.toFixed(3)}): ` +
|
|
367
|
+
`the largest page would allocate ~${beforeMB.toFixed(0)}MB, above the ` +
|
|
368
|
+
`maxCanvasPixels ceiling; reduced to ~${afterMB.toFixed(0)}MB per page.`;
|
|
369
|
+
logger.warn(`[PDF→Image] ⚠️ ${msg}`);
|
|
370
|
+
warnings.push(msg);
|
|
371
|
+
}
|
|
372
|
+
// Create PDF document iterator (password forwarded for encrypted PDFs, #258)
|
|
373
|
+
const document = await pdf(pdfBuffer, {
|
|
374
|
+
scale: effectiveScale,
|
|
375
|
+
...(password ? { password } : {}),
|
|
376
|
+
});
|
|
308
377
|
let pageIndex = 0;
|
|
309
378
|
// Iterate through pages and convert to base64
|
|
310
379
|
for await (const page of document) {
|
|
@@ -346,6 +415,17 @@ export class PDFProcessor {
|
|
|
346
415
|
error: errorMessage,
|
|
347
416
|
conversionTimeMs,
|
|
348
417
|
});
|
|
418
|
+
// #258: map pdfjs's PasswordException to an actionable typed error so a
|
|
419
|
+
// caller learns to supply (or correct) the password instead of seeing a
|
|
420
|
+
// generic "conversion failed". pdfjs code 1 = NEED_PASSWORD, 2 = INCORRECT.
|
|
421
|
+
const pdfErr = error;
|
|
422
|
+
if (pdfErr?.name === "PasswordException" ||
|
|
423
|
+
/password/i.test(errorMessage)) {
|
|
424
|
+
const incorrect = pdfErr.code === 2 || /incorrect|invalid/i.test(errorMessage);
|
|
425
|
+
throw incorrect
|
|
426
|
+
? ErrorFactory.pdfIncorrectPassword()
|
|
427
|
+
: ErrorFactory.pdfPasswordRequired();
|
|
428
|
+
}
|
|
349
429
|
throw new Error(`PDF to image conversion failed: ${errorMessage}`, {
|
|
350
430
|
cause: error,
|
|
351
431
|
});
|
package/dist/neurolink.js
CHANGED
|
@@ -3899,11 +3899,15 @@ Current user's request: ${currentInput}`;
|
|
|
3899
3899
|
evaluationDomain: options.evaluationDomain,
|
|
3900
3900
|
toolUsageContext: options.toolUsageContext,
|
|
3901
3901
|
input: options.input,
|
|
3902
|
-
//
|
|
3903
|
-
// TextGenerationOptions — the message builder reads them
|
|
3904
|
-
// formatting
|
|
3905
|
-
//
|
|
3902
|
+
// Multimodal file-processing options must survive the reconstruction
|
|
3903
|
+
// into TextGenerationOptions — the message builder reads them downstream
|
|
3904
|
+
// (csv encoding/formatting/sanitization/parse timeout, PDF
|
|
3905
|
+
// password/decryption #258). Omitting them here silently dropped e.g.
|
|
3906
|
+
// every csvOptions field (incl. the CLI --csv-* flags) or
|
|
3907
|
+
// pdfOptions.password so the CLI --pdf-password / SDK pdfOptions never
|
|
3908
|
+
// reached convertToImages.
|
|
3906
3909
|
csvOptions: options.csvOptions,
|
|
3910
|
+
pdfOptions: options.pdfOptions,
|
|
3907
3911
|
region: options.region,
|
|
3908
3912
|
tts: options.tts,
|
|
3909
3913
|
stt: options.stt,
|