@juspay/neurolink 7.50.0 → 7.51.1

package/CHANGELOG.md

@@ -1,3 +1,12 @@
+## [7.51.1](https://github.com/juspay/neurolink/compare/v7.51.0...v7.51.1) (2025-10-13)
+
+## [7.51.0](https://github.com/juspay/neurolink/compare/v7.50.0...v7.51.0) (2025-10-12)
+
+### Features
+
+- **(multimodal):** add comprehensive PDF file support with native document processing ([52abf1a](https://github.com/juspay/neurolink/commit/52abf1a85d3a1d9de819919eec1d0ffea3007702))
+- **(multimodal):** add comprehensive PDF file support with native document processing ([020e15a](https://github.com/juspay/neurolink/commit/020e15af0bb46a2b827bb4853baafba5da03cad2))
+
 ## [7.50.0](https://github.com/juspay/neurolink/compare/v7.49.0...v7.50.0) (2025-10-08)
 
 ### Features

package/README.md

@@ -26,6 +26,7 @@ Extracted from production systems at Juspay and battle-tested at enterprise scal
 ## What's New (Q4 2025)
 
 - **CSV File Support** – Attach CSV files to prompts for AI-powered data analysis with auto-detection. → [CSV Guide](docs/features/multimodal-chat.md#csv-file-support)
+- **PDF File Support** – Process PDF documents with native visual analysis for Vertex AI, Anthropic, Bedrock, AI Studio. → [PDF Guide](docs/features/pdf-support.md)
 - **LiteLLM Integration** – Access 100+ AI models from all major providers through unified interface. → [Setup Guide](docs/LITELLM-INTEGRATION.md)
 - **SageMaker Integration** – Deploy and use custom trained models on AWS infrastructure. → [Setup Guide](docs/SAGEMAKER-INTEGRATION.md)
 - **Human-in-the-loop workflows** – Pause generation for user approval/input before tool execution. → [HITL Guide](docs/features/hitl.md)
@@ -266,9 +267,11 @@ const result = await neurolink.generate({
     text: "Create a comprehensive analysis",
     files: [
       "./sales_data.csv", // Auto-detected as CSV
+      "examples/data/invoice.pdf", // Auto-detected as PDF
       "./diagrams/architecture.png", // Auto-detected as image
     ],
   },
+  provider: "vertex", // PDF-capable provider (see docs/features/pdf-support.md)
   enableEvaluation: true,
   region: "us-east-1",
 });
@@ -281,15 +284,15 @@ Full command and API breakdown lives in [`docs/cli/commands.md`](docs/cli/comman
 
 ## Platform Capabilities at a Glance
 
-| Capability               | Highlights                                                                                               |
-| ------------------------ | -------------------------------------------------------------------------------------------------------- |
-| **Provider unification** | 12+ providers with automatic fallback, cost-aware routing, provider orchestration (Q3).                  |
-| **Multimodal pipeline**  | Stream images + CSV data across providers with local/remote assets. Auto-detection for mixed file types. |
-| **Quality & governance** | Auto-evaluation engine (Q3), guardrails middleware (Q4), HITL workflows (Q4), audit logging.             |
-| **Memory & context**     | Conversation memory, Mem0 integration, Redis history export (Q4), context summarization (Q4).            |
-| **CLI tooling**          | Loop sessions (Q3), setup wizard, config validation, Redis auto-detect, JSON output.                     |
-| **Enterprise ops**       | Proxy support, regional routing (Q3), telemetry hooks, configuration management.                         |
-| **Tool ecosystem**       | MCP auto discovery, LiteLLM hub access, SageMaker custom deployment, web search.                         |
+| Capability               | Highlights                                                                                                               |
+| ------------------------ | ------------------------------------------------------------------------------------------------------------------------ |
+| **Provider unification** | 12+ providers with automatic fallback, cost-aware routing, provider orchestration (Q3).                                  |
+| **Multimodal pipeline**  | Stream images + CSV data + PDF documents across providers with local/remote assets. Auto-detection for mixed file types. |
+| **Quality & governance** | Auto-evaluation engine (Q3), guardrails middleware (Q4), HITL workflows (Q4), audit logging.                             |
+| **Memory & context**     | Conversation memory, Mem0 integration, Redis history export (Q4), context summarization (Q4).                            |
+| **CLI tooling**          | Loop sessions (Q3), setup wizard, config validation, Redis auto-detect, JSON output.                                     |
+| **Enterprise ops**       | Proxy support, regional routing (Q3), telemetry hooks, configuration management.                                         |
+| **Tool ecosystem**       | MCP auto discovery, LiteLLM hub access, SageMaker custom deployment, web search.                                         |
 
 ## Documentation Map
 

package/dist/adapters/providerImageAdapter.js

@@ -20,7 +20,29 @@ export class MultimodalLogger {
  * Vision capability definitions for each provider
  */
 const VISION_CAPABILITIES = {
-    openai: ["gpt-4o", "gpt-4o-mini", "gpt-4-turbo", "gpt-4-vision-preview"],
+    openai: [
+        // GPT-5 family (released Aug 2025)
+        "gpt-5",
+        "gpt-5-2025-08-07",
+        "gpt-5-pro",
+        "gpt-5-mini",
+        "gpt-5-nano",
+        // GPT-4.1 family (released Apr 2025)
+        "gpt-4.1",
+        "gpt-4.1-mini",
+        "gpt-4.1-nano",
+        // o-series reasoning models (released Apr 2025)
+        "o3",
+        "o3-mini",
+        "o4",
+        "o4-mini",
+        "o4-mini-deep-research",
+        // Existing GPT-4 models
+        "gpt-4o",
+        "gpt-4o-mini",
+        "gpt-4-turbo",
+        "gpt-4-vision-preview",
+    ],
     "google-ai": [
         "gemini-2.5-pro",
         "gemini-2.5-flash",
@@ -29,44 +51,91 @@ const VISION_CAPABILITIES = {
         "gemini-pro-vision",
     ],
     anthropic: [
+        "claude-3-7-sonnet",
         "claude-3-5-sonnet",
         "claude-3-opus",
         "claude-3-sonnet",
         "claude-3-haiku",
     ],
     azure: [
+        // GPT-5 family
+        "gpt-5",
+        "gpt-5-pro",
+        "gpt-5-mini",
+        // GPT-4.1 family
+        "gpt-4.1",
+        "gpt-4.1-mini",
+        "gpt-4.1-nano",
+        // Existing GPT-4
         "gpt-4o",
         "gpt-4o-mini",
         "gpt-4-turbo",
         "gpt-4-vision-preview",
-        "gpt-4.1",
         "gpt-4",
     ],
     vertex: [
         // Gemini models on Vertex AI
         "gemini-2.5-pro",
         "gemini-2.5-flash",
+        "gemini-2.0-flash",
         "gemini-1.5-pro",
         "gemini-1.5-flash",
-        // Claude models on Vertex AI (with actual Vertex naming patterns)
+        // Claude 4.x models (versioned format)
+        "claude-sonnet-4-5@",
+        "claude-sonnet-4@",
+        "claude-opus-4-1@",
+        "claude-opus-4@",
+        // Claude 3.x models (versioned format)
+        "claude-3-7-sonnet@",
+        "claude-3-5-sonnet@",
+        "claude-opus-3@",
+        "claude-haiku-3@",
+        // Claude models (non-versioned format)
+        "claude-3-7-sonnet",
         "claude-3-5-sonnet",
         "claude-3-opus",
         "claude-3-sonnet",
         "claude-3-haiku",
-        "claude-sonnet-3",
         "claude-sonnet-4",
+        "claude-sonnet-3",
         "claude-opus-3",
         "claude-haiku-3",
-        // Additional Vertex AI Claude model patterns
+        // Additional patterns for compatibility
         "claude-3.5-sonnet",
         "claude-3.5-haiku",
         "claude-3.0-sonnet",
         "claude-3.0-opus",
-        // Versioned model names (e.g., claude-sonnet-4@20250514)
-        "claude-sonnet-4@",
-        "claude-opus-3@",
-        "claude-haiku-3@",
-        "claude-3-5-sonnet@",
+    ],
+    litellm: [
+        // LiteLLM proxies to underlying providers
+        // List models that support vision when going through the proxy
+        "gemini-2.5-pro",
+        "gemini-2.5-flash",
+        "claude-sonnet-4",
+        "claude-sonnet-4-5",
+        "claude-opus-4-1",
+        "gpt-4o",
+        "gpt-4.1",
+        "gpt-5",
+    ],
+    ollama: [
+        // Llama 4 family (May 2025 - Best vision + tool calling)
+        "llama4:scout",
+        "llama4:maverick",
+        // Llama 3.2 vision
+        "llama3.2-vision",
+        // Gemma 3 family (SigLIP vision encoder - supports tool calling + vision)
+        "gemma3:4b",
+        "gemma3:12b",
+        "gemma3:27b",
+        "gemma3:latest",
+        // Mistral Small family (vision + tool calling)
+        "mistral-small3.1",
+        "mistral-small3.1:large",
+        "mistral-small3.1:medium",
+        "mistral-small3.1:small",
+        // LLaVA (vision-focused)
+        "llava",
     ],
 };
 /**
@@ -100,6 +169,9 @@ export class ProviderImageAdapter {
                 case "vertex":
                     adaptedPayload = this.formatForVertex(text, images, model);
                     break;
+                case "ollama":
+                    adaptedPayload = this.formatForOpenAI(text, images);
+                    break;
                 default:
                     throw new Error(`Vision not supported for provider: ${provider}`);
             }

package/dist/agent/directTools.d.ts

@@ -349,13 +349,13 @@ export declare const directAgentTools: {
     analyzeCSV: import("ai").Tool<z.ZodObject<{
         filePath: z.ZodEffects<z.ZodString, string, string>;
         operation: z.ZodEnum<["count_by_column", "sum_by_column", "average_by_column", "min_max_by_column", "describe"]>;
-        column: z.ZodOptional<z.ZodString>;
-        maxRows: z.ZodOptional<z.ZodNumber>;
+        column: z.ZodDefault<z.ZodOptional<z.ZodString>>;
+        maxRows: z.ZodDefault<z.ZodOptional<z.ZodNumber>>;
     }, "strip", z.ZodTypeAny, {
         operation: "count_by_column" | "sum_by_column" | "average_by_column" | "min_max_by_column" | "describe";
+        maxRows: number;
         filePath: string;
-        maxRows?: number | undefined;
-        column?: string | undefined;
+        column: string;
     }, {
         operation: "count_by_column" | "sum_by_column" | "average_by_column" | "min_max_by_column" | "describe";
         filePath: string;
@@ -364,7 +364,7 @@ export declare const directAgentTools: {
     }>, {
         success: boolean;
         operation: "count_by_column" | "sum_by_column" | "average_by_column" | "min_max_by_column" | "describe";
-        column: string | undefined;
+        column: string;
         result: string;
         rowCount: number;
     } | {
@@ -376,17 +376,17 @@ export declare const directAgentTools: {
         success: boolean;
         error: string;
         operation: "count_by_column" | "sum_by_column" | "average_by_column" | "min_max_by_column" | "describe";
-        column: string | undefined;
+        column: string;
     }> & {
         execute: (args: {
             operation: "count_by_column" | "sum_by_column" | "average_by_column" | "min_max_by_column" | "describe";
+            maxRows: number;
             filePath: string;
-            maxRows?: number | undefined;
-            column?: string | undefined;
+            column: string;
         }, options: import("ai").ToolExecutionOptions) => PromiseLike<{
             success: boolean;
             operation: "count_by_column" | "sum_by_column" | "average_by_column" | "min_max_by_column" | "describe";
-            column: string | undefined;
+            column: string;
             result: string;
             rowCount: number;
         } | {
@@ -398,7 +398,7 @@ export declare const directAgentTools: {
             success: boolean;
             error: string;
             operation: "count_by_column" | "sum_by_column" | "average_by_column" | "min_max_by_column" | "describe";
-            column: string | undefined;
+            column: string;
         }>;
     };
     websearchGrounding: import("ai").Tool<z.ZodObject<{

package/dist/agent/directTools.js

@@ -31,7 +31,7 @@ export const directAgentTools = {
             timezone: z
                 .string()
                 .optional()
-                .describe('Timezone (e.g., "America/New_York", "Asia/Kolkata"). Defaults to local time.'),
+                .describe('Timezone (e.g., "America/New_York", "Asia/Kolkata"). Defaults to system local time.'),
         }),
         execute: async ({ timezone }) => {
             try {
@@ -103,8 +103,8 @@ export const directAgentTools = {
             includeHidden: z
                 .boolean()
                 .optional()
-                .describe("Include hidden files (starting with .)")
-                .default(false),
+                .default(false)
+                .describe("Include hidden files (starting with .)"),
         }),
         execute: async ({ path: dirPath, includeHidden }) => {
             try {
@@ -377,10 +377,12 @@ export const directAgentTools = {
             column: z
                 .string()
                 .optional()
+                .default("")
                 .describe("Column name for the operation (required for most operations)"),
             maxRows: z
                 .number()
                 .optional()
+                .default(1000)
                 .describe("Maximum rows to process (default: 1000)"),
         }),
         execute: async ({ filePath, operation, column, maxRows = 1000 }) => {

package/dist/cli/commands/config.js

@@ -811,3 +811,4 @@ export class ConfigManager {
 }
 // Export for use in other CLI commands
 export const configManager = new ConfigManager();
+//# sourceMappingURL=config.js.map

package/dist/cli/commands/mcp.js

@@ -928,3 +928,4 @@ export class MCPCommandFactory {
         return servers;
     }
 }
+//# sourceMappingURL=mcp.js.map

package/dist/cli/commands/models.js

@@ -655,3 +655,4 @@ export class ModelsCommandFactory {
         }
     }
 }
+//# sourceMappingURL=models.js.map

package/dist/cli/commands/ollama.js

@@ -286,3 +286,4 @@ async function setupHandler() {
     logger.always(chalk.gray("\nFor more information, see: https://docs.neurolink.ai/providers/ollama"));
 }
 export default addOllamaCommands;
+//# sourceMappingURL=ollama.js.map

package/dist/cli/commands/setup-anthropic.js

@@ -412,3 +412,4 @@ function showUsageExample() {
     logger.always(chalk.cyan("   pnpm cli generate 'Explain quantum computing' --provider anthropic"));
     logger.always(chalk.cyan("   pnpm cli generate 'Analyze this data' --provider anthropic --enable-analytics"));
 }
+//# sourceMappingURL=setup-anthropic.js.map

package/dist/cli/commands/setup-azure.js

@@ -413,3 +413,4 @@ function showUsageExample() {
     logger.always(chalk.cyan("   pnpm cli generate 'Explain quantum computing' --provider azure"));
     logger.always(chalk.cyan("   pnpm cli generate 'Analyze this data' --provider azure --enable-analytics"));
 }
+//# sourceMappingURL=setup-azure.js.map

package/dist/cli/commands/setup-bedrock.js

@@ -485,3 +485,4 @@ function maskCredential(credential) {
     const middle = "*".repeat(Math.max(4, credential.length - 8));
     return `${start}${middle}${end}`;
 }
+//# sourceMappingURL=setup-bedrock.js.map

package/dist/cli/commands/setup-gcp.js

@@ -567,3 +567,4 @@ async function updateEnvFile(method, config) {
         throw error;
     }
 }
+//# sourceMappingURL=setup-gcp.js.map

package/dist/cli/commands/setup-google-ai.js

@@ -367,3 +367,4 @@ function showUsageExample() {
     logger.always(chalk.cyan("   pnpm cli generate 'Explain quantum computing' --provider google-ai"));
     logger.always(chalk.cyan("   pnpm cli generate 'Analyze this data' --provider google-ai --enable-analytics"));
 }
+//# sourceMappingURL=setup-google-ai.js.map

package/dist/cli/commands/setup-huggingface.js

@@ -198,3 +198,4 @@ export const setupHuggingFaceBuilder = {
         default: false,
     },
 };
+//# sourceMappingURL=setup-huggingface.js.map

package/dist/cli/commands/setup-mistral.js

@@ -231,3 +231,4 @@ export const setupMistralBuilder = {
         default: false,
     },
 };
+//# sourceMappingURL=setup-mistral.js.map

package/dist/cli/commands/setup-openai.js

@@ -400,3 +400,4 @@ function showUsageExample() {
     logger.always(chalk.cyan("   pnpm cli generate 'Hello from OpenAI!' --provider openai"));
     logger.always(chalk.cyan("   pnpm cli generate 'Explain quantum computing' --provider openai"));
 }
+//# sourceMappingURL=setup-openai.js.map

package/dist/cli/commands/setup.js

@@ -537,3 +537,4 @@ async function showProviderStatus() {
         logger.error(chalk.red("Error checking provider status:"), error);
     }
 }
+//# sourceMappingURL=setup.js.map

package/dist/cli/errorHandler.js

@@ -26,3 +26,4 @@ export function handleError(_error, context) {
         process.exit(1);
     }
 }
+//# sourceMappingURL=errorHandler.js.map

package/dist/cli/factories/commandFactory.d.ts

@@ -7,6 +7,7 @@ export declare class CLICommandFactory {
     private static buildOptions;
     private static processCliImages;
     private static processCliCSVFiles;
+    private static processCliPDFFiles;
     private static processCliFiles;
     private static processOptions;
     private static handleOutput;

package/dist/cli/factories/commandFactory.js

@@ -54,6 +54,10 @@ export class CLICommandFactory {
             description: "Add CSV file for data analysis (can be used multiple times)",
             alias: "c",
         },
+        pdf: {
+            type: "string",
+            description: "Add PDF file for analysis (can be used multiple times)",
+        },
         file: {
             type: "string",
             description: "Add file with auto-detection (CSV, image, etc. - can be used multiple times)",
@@ -215,6 +219,13 @@ export class CLICommandFactory {
         }
         return Array.isArray(csvFiles) ? csvFiles : [csvFiles];
     }
+    // Helper method to process CLI PDF files
+    static processCliPDFFiles(pdfFiles) {
+        if (!pdfFiles) {
+            return undefined;
+        }
+        return Array.isArray(pdfFiles) ? pdfFiles : [pdfFiles];
+    }
     // Helper method to process CLI files with auto-detection
     static processCliFiles(files) {
         if (!files) {
@@ -1064,14 +1075,17 @@ export class CLICommandFactory {
             // Process CLI multimodal inputs
             const imageBuffers = CLICommandFactory.processCliImages(argv.image);
             const csvFiles = CLICommandFactory.processCliCSVFiles(argv.csv);
+            const pdfFiles = CLICommandFactory.processCliPDFFiles(argv.pdf);
             const files = CLICommandFactory.processCliFiles(argv.file);
+            const generateInput = {
+                text: inputText,
+                ...(imageBuffers && { images: imageBuffers }),
+                ...(csvFiles && { csvFiles }),
+                ...(pdfFiles && { pdfFiles }),
+                ...(files && { files }),
+            };
             const result = await sdk.generate({
-                input: {
-                    text: inputText,
-                    ...(imageBuffers && { images: imageBuffers }),
-                    ...(csvFiles && { csvFiles }),
-                    ...(files && { files }),
-                },
+                input: generateInput,
                 csvOptions: {
                     maxRows: argv.csvMaxRows,
                     formatStyle: argv.csvFormat,
@@ -1248,12 +1262,14 @@ export class CLICommandFactory {
         // Process CLI multimodal inputs
         const imageBuffers = CLICommandFactory.processCliImages(argv.image);
         const csvFiles = CLICommandFactory.processCliCSVFiles(argv.csv);
+        const pdfFiles = CLICommandFactory.processCliPDFFiles(argv.pdf);
         const files = CLICommandFactory.processCliFiles(argv.file);
         const stream = await sdk.stream({
             input: {
                 text: inputText,
                 ...(imageBuffers && { images: imageBuffers }),
                 ...(csvFiles && { csvFiles }),
+                ...(pdfFiles && { pdfFiles }),
                 ...(files && { files }),
             },
             csvOptions: {
@@ -2025,3 +2041,4 @@ export class CLICommandFactory {
         }
     }
 }
+//# sourceMappingURL=commandFactory.js.map

package/dist/cli/factories/ollamaCommandFactory.js

@@ -374,3 +374,4 @@ export class OllamaCommandFactory {
         logger.always(chalk.gray("\nFor more information, see: https://docs.neurolink.ai/providers/ollama"));
     }
 }
+//# sourceMappingURL=ollamaCommandFactory.js.map

package/dist/cli/factories/sagemakerCommandFactory.js

@@ -782,3 +782,4 @@ export class SageMakerCommandFactory {
         }
     }
 }
+//# sourceMappingURL=sagemakerCommandFactory.js.map

package/dist/cli/factories/setupCommandFactory.js

@@ -135,3 +135,4 @@ export class SetupCommandFactory {
         });
     }
 }
+//# sourceMappingURL=setupCommandFactory.js.map

package/dist/cli/index.js

@@ -57,3 +57,4 @@ async function cleanup() {
     const { flushOpenTelemetry } = await import("../lib/services/server/ai/observability/instrumentation.js");
     await flushOpenTelemetry();
 }
+//# sourceMappingURL=index.js.map

package/dist/cli/loop/conversationSelector.js

@@ -220,3 +220,4 @@ export class ConversationSelector {
         };
     }
 }
+//# sourceMappingURL=conversationSelector.js.map

package/dist/cli/loop/optionsSchema.js

@@ -57,3 +57,4 @@ export const textGenerationOptionsSchema = {
         description: "Context about tools/MCPs used in the interaction.",
     },
 };
+//# sourceMappingURL=optionsSchema.js.map

package/dist/cli/loop/session.js

@@ -507,3 +507,4 @@ export class LoopSession {
         }
     }
 }
+//# sourceMappingURL=session.js.map

package/dist/cli/parser.js

@@ -159,3 +159,4 @@ export function initializeCliParser() {
         // Setup Commands - Using SetupCommandFactory
         .command(SetupCommandFactory.createSetupCommands())); // Close the main return statement
 }
+//# sourceMappingURL=parser.js.map

package/dist/cli/utils/completeSetup.js

@@ -80,3 +80,4 @@ export async function configInit(quiet = false) {
     }
     await runCompleteSetup(quiet);
 }
+//# sourceMappingURL=completeSetup.js.map

package/dist/cli/utils/envManager.js

@@ -239,3 +239,4 @@ export function validateEnvFile(envPath = ".env", requiredVars = []) {
     }
     return result;
 }
+//# sourceMappingURL=envManager.js.map

package/dist/cli/utils/interactiveSetup.js

@@ -372,3 +372,4 @@ export function displaySetupSummary(result, quiet = false) {
     }
     logger.always(chalk.blue("\n📚 Documentation: https://github.com/juspay/neurolink#setup"));
 }
+//# sourceMappingURL=interactiveSetup.js.map

package/dist/cli/utils/ollamaUtils.js

@@ -181,3 +181,4 @@ export class OllamaUtils {
         }
     }
 }
+//# sourceMappingURL=ollamaUtils.js.map

package/dist/constants/index.js

@@ -42,7 +42,7 @@ export const OPERATION_TIMEOUTS = {
 };
 // Import the constants from the individual files for use in composite configurations
 import { PROVIDER_TIMEOUTS, MCP_TIMEOUTS, TimeoutUtils, TOOL_TIMEOUTS, } from "./timeouts.js";
-import { RETRY_ATTEMPTS, RETRY_DELAYS, PROVIDER_RETRY, RetryUtils, } from "./retry.js";
+import { RETRY_ATTEMPTS, RETRY_DELAYS, PROVIDER_RETRY, BACKOFF_CONFIG, RetryUtils, } from "./retry.js";
 import { CONCURRENCY_LIMITS, MEMORY_THRESHOLDS, BUFFER_SIZES, } from "./performance.js";
 import { TokenUtils } from "./tokens.js";
 /**

package/dist/core/baseProvider.d.ts

@@ -133,6 +133,11 @@ export declare abstract class BaseProvider implements AIProvider {
      * Process custom tools from setupToolExecutor
      */
     private processCustomTools;
+    /**
+     * Recursively fix JSON Schema for OpenAI strict mode compatibility
+     * OpenAI requires additionalProperties: false at ALL levels and preserves required array
+     */
+    private fixSchemaForOpenAIStrictMode;
     /**
      * Create an external MCP tool
      */