@vint.tri/report_gen_mcp 1.7.2 → 1.7.3

package/FIXED_NEURAL_NETWORK_INSTRUCTIONS_v1.6.5.md

@@ -500,9 +500,9 @@
 
 ### Изображения, сгенерированные с помощью generate-image (Chutes AI)
 
-При использовании инструмента `generate-image` для генерации изображений с помощью Chutes AI, важно правильно форматировать параметры. Все числовые параметры должны быть представлены как числа, а не как строки.
+При использовании инструмента `generate-image` для генерации изображений с помощью Chutes AI, числовые параметры могут быть представлены как числа или как строки. Инструмент автоматически преобразует строковые параметры в числа.
 
-✅ ПРАВИЛЬНО (числовые параметры как числа):
+✅ ПРИМЕР 1 (числовые параметры как числа):
 ```json
 {
   "method": "tools/call",
@@ -521,7 +521,7 @@
 }
 ```
 
-❌ НЕПРАВИЛЬНО (числовые параметры как строки):
+✅ ПРИМЕР 2 (числовые параметры как строки):
 ```json
 {
   "method": "tools/call",
@@ -540,7 +540,26 @@
 }
 ```
 
-**ВАЖНО:** Все числовые параметры (`width`, `height`, `guidance_scale`, `num_inference_steps`, `seed`) должны быть представлены как числа, а не как строки в кавычках. Это критически важно для корректной работы инструмента.
+✅ ПРИМЕР 3 (смешанные параметры):
+```json
+{
+  "method": "tools/call",
+  "params": {
+    "name": "generate-image",
+    "arguments": {
+      "prompt": "a dog made of US dollar bills",
+      "width": "512",
+      "height": 512,
+      "guidance_scale": "7.5",
+      "negative_prompt": "blurry, low quality",
+      "num_inference_steps": 50,
+      "seed": "0"
+    }
+  }
+}
+```
+
+**Гибкость параметров:** Инструмент поддерживает как числовые, так и строковые значения для числовых параметров (`width`, `height`, `guidance_scale`, `num_inference_steps`, `seed`). Это обеспечивает совместимость с различными способами формирования запросов.
 
 ### Изображения по URL
 

package/NEURAL_NETWORK_INSTRUCTIONS.md

@@ -479,6 +479,69 @@
 }
 ```
 
+### Изображения, сгенерированные с помощью generate-image (Chutes AI)
+
+При использовании инструмента `generate-image` для генерации изображений с помощью Chutes AI, числовые параметры могут быть представлены как числа или как строки. Инструмент автоматически преобразует строковые параметры в числа.
+
+✅ ПРИМЕР 1 (числовые параметры как числа):
+```json
+{
+  "method": "tools/call",
+  "params": {
+    "name": "generate-image",
+    "arguments": {
+      "prompt": "a dog made of US dollar bills",
+      "width": 512,
+      "height": 512,
+      "guidance_scale": 7.5,
+      "negative_prompt": "blurry, low quality",
+      "num_inference_steps": 50,
+      "seed": 0
+    }
+  }
+}
+```
+
+✅ ПРИМЕР 2 (числовые параметры как строки):
+```json
+{
+  "method": "tools/call",
+  "params": {
+    "name": "generate-image",
+    "arguments": {
+      "prompt": "a dog made of US dollar bills",
+      "width": "512",
+      "height": "512",
+      "guidance_scale": "7.5",
+      "negative_prompt": "blurry, low quality",
+      "num_inference_steps": "50",
+      "seed": "0"
+    }
+  }
+}
+```
+
+✅ ПРИМЕР 3 (смешанные параметры):
+```json
+{
+  "method": "tools/call",
+  "params": {
+    "name": "generate-image",
+    "arguments": {
+      "prompt": "a dog made of US dollar bills",
+      "width": "512",
+      "height": 512,
+      "guidance_scale": "7.5",
+      "negative_prompt": "blurry, low quality",
+      "num_inference_steps": 50,
+      "seed": "0"
+    }
+  }
+}
+```
+
+**Гибкость параметров:** Инструмент поддерживает как числовые, так и строковые значения для числовых параметров (`width`, `height`, `guidance_scale`, `num_inference_steps`, `seed`). Это обеспечивает совместимость с различными способами формирования запросов.
+
 ### Изображения по URL
 
 ```json

package/PUBLICATION_CONFIRMATION_V1.7.0.md

@@ -0,0 +1,35 @@
+# Publication Confirmation - Version 1.7.0
+
+## Release Details
+- **Version:** 1.7.0
+- **Release Date:** September 2, 2025
+- **Status:** Published
+
+## Summary
+Version 1.7.0 has been successfully published. This release focuses on improving the robustness of the image generation tools by fixing parameter validation issues that were causing errors when the neural network sent numeric parameters as strings.
+
+## Key Improvements
+1. **String Parameter Validation Fix** - Updated Zod schema validation to accept both numbers and strings for numeric parameters in image generation tools
+2. **Enhanced Neural Network Compatibility** - The neural network can now send parameters as strings without errors
+3. **Backward Compatibility Maintained** - Existing code that sends numeric parameters continues to work
+4. **Flexible Input Handling** - Supports mixed parameter types (some as numbers, some as strings)
+
+## Files Updated
+- `src/mcp/imageGenerationServer.ts` - Core fix implementation
+- `dist/mcp/imageGenerationServer.js` - Compiled version
+- `NEURAL_NETWORK_INSTRUCTIONS.md` - Documentation updates
+- `FIXED_NEURAL_NETWORK_INSTRUCTIONS_v1.6.5.md` - Documentation updates
+- `package.json` - Version bump to 1.7.0
+
+## Testing
+All comprehensive tests pass successfully, confirming the fix resolves the original issue while maintaining full backward compatibility:
+- String parameters test ✓
+- Mixed parameters test ✓
+- Numeric parameters test (backward compatibility) ✓
+- Schema validation test ✓
+
+## Impact
+This release eliminates frustrating validation errors that were preventing image generation, significantly improving the user experience while maintaining full backward compatibility.
+
+## Next Steps
+No immediate action required. The fix is backward compatible and will automatically benefit all users of the image generation tools.

package/STRING_PARAMETER_FIX_SUMMARY.md

@@ -0,0 +1,97 @@
+# String Parameter Validation Fix for Image Generation Tools
+
+## Problem Description
+
+The neural network was sending numeric parameters as strings to the `generate-image` and `generate-image-to-file` tools, causing validation errors:
+
+```
+Error calling tool generate-image: Error: Error invoking remote method 'mcp:call-tool': McpError: MCP error -32602: Invalid arguments for tool generate-image: [
+  {
+    "code": "invalid_type",
+    "expected": "number",
+    "received": "string",
+    "path": ["width"],
+    "message": "Expected number, received string"
+  },
+  // Similar errors for height, guidance_scale, num_inference_steps, seed
+]
+```
+
+## Root Cause
+
+The Zod schema validation in the MCP server expected actual numbers for numeric parameters, but the neural network was sending them as JSON strings:
+
+```json
+{
+  "params": {
+    "prompt": "a dog made of dollar bills",
+    "width": "512",
+    "height": "512",
+    "guidance_scale": "7.5",
+    "negative_prompt": "blurry, low quality",
+    "num_inference_steps": "50",
+    "seed": "0"
+  }
+}
+```
+
+The validation was happening at the MCP framework level before the conversion logic in our handler could run.
+
+## Solution Implemented
+
+### 1. Updated Zod Schema Validation
+
+Modified the input schema for both `generate_image` and `generate_image_to_file` tools in `src/mcp/imageGenerationServer.ts` to accept both numbers and strings:
+
+```typescript
+// Before (only accepted numbers)
+width: z.number().optional().describe("Ширина изображения (128-2048)"),
+
+// After (accepts both numbers and strings)
+width: z.union([z.number(), z.string()]).optional().describe("Ширина изображения (128-2048)"),
+```
+
+This change was applied to all numeric parameters:
+- `width`
+- `height`
+- `guidance_scale`
+- `num_inference_steps`
+- `seed`
+
+### 2. Maintained Existing Conversion Logic
+
+The existing parameter conversion logic in the handler functions was already correctly implemented and continues to work:
+
+```typescript
+// Преобразуем строковые значения в числа, если необходимо
+const generationParams: Partial<ImageGenerationSettings> = {};
+if (args.width !== undefined) {
+  generationParams.width = typeof args.width === 'string' ? parseFloat(args.width) : args.width;
+}
+// ... similar for other parameters
+```
+
+## Benefits
+
+1. **Neural Network Compatibility**: The neural network can now send parameters as strings without errors
+2. **Backward Compatibility**: Existing code that sends numeric parameters continues to work
+3. **Flexible Input Handling**: Supports mixed parameter types (some as numbers, some as strings)
+4. **Robust Error Handling**: Proper validation and conversion with appropriate error messages
+
+## Testing
+
+Created comprehensive tests to verify the fix works correctly:
+
+1. **String Parameters Test**: Verifies that string parameters are accepted and converted correctly
+2. **Mixed Parameters Test**: Ensures mixed parameter types (numbers and strings) work properly
+3. **Numeric Parameters Test**: Confirms backward compatibility with all numeric parameters
+4. **Validation Test**: Tests that the Zod schema correctly validates different parameter combinations
+
+## Files Modified
+
+- `src/mcp/imageGenerationServer.ts` - Updated Zod schema validation
+- `dist/mcp/imageGenerationServer.js` - Compiled version (automatically generated)
+
+## Impact
+
+This fix resolves the issue where neural networks sending string-formatted numeric parameters would receive validation errors, allowing seamless image generation with the Chutes AI service while maintaining full backward compatibility.

package/VERSION_1.7.0_RELEASE_NOTES.md

@@ -0,0 +1,126 @@
+# Version 1.7.0 Release Notes
+
+## Overview
+
+This release focuses on improving the robustness of the image generation tools by fixing parameter validation issues that were causing errors when the neural network sent numeric parameters as strings.
+
+## Key Changes
+
+### String Parameter Validation Fix for Image Generation Tools
+
+**Problem:** The neural network was sending numeric parameters as strings to the `generate-image` and `generate-image-to-file` tools, causing validation errors:
+
+```
+Error calling tool generate-image: Error: Error invoking remote method 'mcp:call-tool': McpError: MCP error -32602: Invalid arguments for tool generate-image: [
+  {
+    "code": "invalid_type",
+    "expected": "number",
+    "received": "string",
+    "path": ["width"],
+    "message": "Expected number, received string"
+  },
+  // Similar errors for height, guidance_scale, num_inference_steps, seed
+]
+```
+
+**Solution:** Updated the Zod schema validation in `src/mcp/imageGenerationServer.ts` to accept both numbers and strings for numeric parameters, with automatic conversion to numbers in the implementation layer.
+
+### Files Modified
+
+1. `src/mcp/imageGenerationServer.ts` - Updated Zod schema validation for both `generate_image` and `generate_image_to_file` tools
+2. `dist/mcp/imageGenerationServer.js` - Compiled version (automatically generated)
+3. `NEURAL_NETWORK_INSTRUCTIONS.md` - Updated documentation to reflect the flexibility in parameter formats
+4. `FIXED_NEURAL_NETWORK_INSTRUCTIONS_v1.6.5.md` - Updated documentation with examples of different parameter formats
+
+### Benefits
+
+1. **Neural Network Compatibility**: The neural network can now send parameters as strings without errors
+2. **Backward Compatibility**: Existing code that sends numeric parameters continues to work
+3. **Flexible Input Handling**: Supports mixed parameter types (some as numbers, some as strings)
+4. **Improved User Experience**: Eliminates frustrating validation errors that were preventing image generation
+
+### Testing
+
+Created comprehensive tests to verify the fix works correctly:
+- String parameters test
+- Mixed parameters test
+- Numeric parameters test (backward compatibility)
+- Schema validation test
+
+All tests pass successfully, confirming the fix resolves the original issue while maintaining full backward compatibility.
+
+## How to Upgrade
+
+To upgrade to version 1.7.0, simply update your package and rebuild:
+
+```bash
+npm install
+npm run build
+```
+
+The fix is backward compatible, so no code changes are required in existing implementations.
+
+## Documentation Updates
+
+Updated neural network instructions now clearly explain that numeric parameters can be sent as either numbers or strings:
+
+✅ PREFERRED FORMAT (numbers):
+```json
+{
+  "method": "tools/call",
+  "params": {
+    "name": "generate-image",
+    "arguments": {
+      "prompt": "a dog made of US dollar bills",
+      "width": 512,
+      "height": 512,
+      "guidance_scale": 7.5,
+      "negative_prompt": "blurry, low quality",
+      "num_inference_steps": 50,
+      "seed": 0
+    }
+  }
+}
+```
+
+✅ ALTERNATIVE FORMAT (strings):
+```json
+{
+  "method": "tools/call",
+  "params": {
+    "name": "generate-image",
+    "arguments": {
+      "prompt": "a dog made of US dollar bills",
+      "width": "512",
+      "height": "512",
+      "guidance_scale": "7.5",
+      "negative_prompt": "blurry, low quality",
+      "num_inference_steps": "50",
+      "seed": "0"
+    }
+  }
+}
+```
+
+✅ MIXED FORMAT:
+```json
+{
+  "method": "tools/call",
+  "params": {
+    "name": "generate-image",
+    "arguments": {
+      "prompt": "a dog made of US dollar bills",
+      "width": "512",
+      "height": 512,
+      "guidance_scale": "7.5",
+      "negative_prompt": "blurry, low quality",
+      "num_inference_steps": 50,
+      "seed": "0"
+    }
+  }
+}
+```
+
+## Impact
+
+This fix resolves the issue where neural networks sending string-formatted numeric parameters would receive validation errors, allowing seamless image generation with the Chutes AI service while maintaining full backward compatibility.

package/dist/index.js

@@ -103,7 +103,7 @@ if (process.argv.length === 2) {
     // No command specified, run in stdio mode using MCP SDK
     const mcpServer = new McpServer({
         name: "report_gen_mcp",
-        version: "1.7.2"
+        version: "1.7.3"
     }, {
         // Disable health check to prevent automatic calls
         capabilities: {

package/dist/mcp/imageGenerationServer.js

@@ -106,12 +106,12 @@ app.registerTool("generate_image", {
     description: "Генерирует изображения. ВАЖНО: промпт и негативный промпт должны быть СТРОГО на английском языке!",
     inputSchema: {
         prompt: z.string().describe("Текстовый промпт для генерации изображения НА АНГЛИЙСКОМ ЯЗЫКЕ"),
-        width: z.number().optional().describe("Ширина изображения (128-2048)"),
-        height: z.number().optional().describe("Высота изображения (128-2048)"),
-        guidance_scale: z.number().optional().describe("Сила следования промпту (1.0-20.0)"),
+        width: z.union([z.number(), z.string()]).optional().describe("Ширина изображения (128-2048)"),
+        height: z.union([z.number(), z.string()]).optional().describe("Высота изображения (128-2048)"),
+        guidance_scale: z.union([z.number(), z.string()]).optional().describe("Сила следования промпту (1.0-20.0)"),
         negative_prompt: z.string().optional().describe("Негативный промпт (что НЕ включать в изображение)"),
-        num_inference_steps: z.number().optional().describe("Количество шагов генерации (1-50)"),
-        seed: z.number().optional().describe("Seed для воспроизводимости (0 = случайный)")
+        num_inference_steps: z.union([z.number(), z.string()]).optional().describe("Количество шагов генерации (1-50)"),
+        seed: z.union([z.number(), z.string()]).optional().describe("Seed для воспроизводимости (0 = случайный)")
     },
 }, async (args) => {
     try {
@@ -199,12 +199,12 @@ app.registerTool("generate_image_to_file", {
         prompt: z.string().describe("Текстовый промпт для генерации изображения НА АНГЛИЙСКОМ ЯЗЫКЕ"),
         directory: z.string().describe("Каталог для сохранения изображения"),
         filename: z.string().describe("Имя файла для сохранения (с расширением)"),
-        width: z.number().optional().describe("Ширина изображения (128-2048)"),
-        height: z.number().optional().describe("Высота изображения (128-2048)"),
-        guidance_scale: z.number().optional().describe("Сила следования промпту (1.0-20.0)"),
+        width: z.union([z.number(), z.string()]).optional().describe("Ширина изображения (128-2048)"),
+        height: z.union([z.number(), z.string()]).optional().describe("Высота изображения (128-2048)"),
+        guidance_scale: z.union([z.number(), z.string()]).optional().describe("Сила следования промпту (1.0-20.0)"),
         negative_prompt: z.string().optional().describe("Негативный промпт (что НЕ включать в изображение)"),
-        num_inference_steps: z.number().optional().describe("Количество шагов генерации (1-50)"),
-        seed: z.number().optional().describe("Seed для воспроизводимости (0 = случайный)")
+        num_inference_steps: z.union([z.number(), z.string()]).optional().describe("Количество шагов генерации (1-50)"),
+        seed: z.union([z.number(), z.string()]).optional().describe("Seed для воспроизводимости (0 = случайный)")
     },
 }, async (args) => {
     try {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vint.tri/report_gen_mcp",
-  "version": "1.7.2",
+  "version": "1.7.3",
   "description": "CLI tool for generating HTML reports with embedded charts and images",
   "main": "dist/index.js",
   "bin": {

package/publish.sh

@@ -1,45 +1,12 @@
 #!/bin/bash
 
-# Script to publish @vint.tri/report_gen_mcp package version 1.6.9
+# Exit immediately if a command exits with a non-zero status.
+set -e
 
-echo "Publishing @vint.tri/report_gen_mcp version 1.6.9"
-
-# Check if we're logged into npm
-echo "Checking npm login status..."
-npm whoami
-
-if [ $? -ne 0 ]; then
-  echo "Error: Not logged into npm. Please run 'npm login' first."
-  exit 1
-fi
-
-# Check current version
-echo "Checking current package version..."
-CURRENT_VERSION=$(node -p "require('./package.json').version")
-echo "Current version: $CURRENT_VERSION"
-
-if [ "$CURRENT_VERSION" != "1.6.9" ]; then
-  echo "Error: Package version is not 1.6.9. Please update package.json first."
-  exit 1
-fi
-
-# Run build to ensure we have the latest compiled files
-echo "Building project..."
+echo "Building the project..."
 npm run build
 
-if [ $? -ne 0 ]; then
-  echo "Error: Build failed. Please fix build issues before publishing."
-  exit 1
-fi
-
-# Publish to npm
-echo "Publishing to npm..."
+echo "Publishing the package to npm..."
 npm publish --access public
 
-if [ $? -eq 0 ]; then
-  echo "Successfully published @vint.tri/report_gen_mcp version 1.6.9"
-  echo "You can now install it using: npm install @vint.tri/report_gen_mcp@1.6.9"
-else
-  echo "Error: Failed to publish package"
-  exit 1
-fi
+echo "Publication complete."