fix(mcp): handle non-JSON success response, update params_to_tool docs

- client.ts: wrap success response.json() in try-catch to handle
  empty/non-JSON responses gracefully
- README.md, docs/en-US/mcp-server.md, docs/zh-CN/mcp-server.md:
  update params_to_tool type from string to object, fix examples
  to use object literal instead of JSON.stringify

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: linfangw

docs/en-US/mcp-server.md

@@ -178,7 +178,7 @@ This is the **Call** action and costs **1-100 credits** per invocation, priced b
 |-----------|------|----------|-------------|
 | `tool_id` | string | Yes | Tool ID from discovery results |
 | `search_id` | string | Yes | Search ID from the discovery that found this tool |
-| `params_to_tool` | string | Yes | JSON-stringified parameters to pass to the tool |
+| `params_to_tool` | object | Yes | Dictionary of parameters to pass to the tool |
 | `session_id` | string | No | Session identifier for tracking |
 | `max_response_size` | number | No | Max response size in bytes (default `20480`) |
 
@@ -188,7 +188,7 @@ Example:
 {
   "tool_id": "openweathermap.weather.execute.v1",
   "search_id": "YOUR_SEARCH_ID",
-  "params_to_tool": "{\"city\":\"London\",\"units\":\"metric\"}"
+  "params_to_tool": {"city": "London", "units": "metric"}
 }
 ```
 
@@ -251,7 +251,7 @@ If `session_id` is omitted, the MCP server may generate one for the lifetime of
 - Verify the API key is valid
 - Verify the selected `tool_id` came from a prior discovery
 - Re-run `get_tools_by_ids` to inspect the tool before calling
-- Check that `params_to_tool` is valid JSON
+- Check that `params_to_tool` is a valid object
 
 ### Windows-specific issues
 

docs/zh-CN/mcp-server.md

@@ -178,7 +178,7 @@ QVERIS_API_KEY=your-api-key
 |-----|------|------|------|
 | `tool_id` | string | 是 | 来自发现结果的工具 ID |
 | `search_id` | string | 是 | 发现该工具的搜索 ID |
-| `params_to_tool` | string | 是 | JSON 字符串化的工具参数 |
+| `params_to_tool` | object | 是 | 传递给工具的参数字典 |
 | `session_id` | string | 否 | 用于追踪的会话标识符 |
 | `max_response_size` | number | 否 | 最大响应字节数（默认 `20480`） |
 
@@ -188,7 +188,7 @@ QVERIS_API_KEY=your-api-key
 {
   "tool_id": "openweathermap.weather.execute.v1",
   "search_id": "YOUR_SEARCH_ID",
-  "params_to_tool": "{\"city\":\"London\",\"units\":\"metric\"}"
+  "params_to_tool": {"city": "London", "units": "metric"}
 }
 ```
 
@@ -251,7 +251,7 @@ QVERIS_API_KEY=your-api-key
 - 验证 API 密钥是否有效
 - 验证所选 `tool_id` 来自此前的发现结果
 - 重新运行 `get_tools_by_ids` 检查工具后再调用
-- 检查 `params_to_tool` 是否为有效 JSON
+- 检查 `params_to_tool` 是否为有效对象
 
 ### Windows 特定问题
 

packages/mcp/README.md

@@ -99,7 +99,7 @@ Execute a discovered tool with specific parameters.
 |-----------|------|----------|-------------|
 | `tool_id` | string | ✓ | Tool ID from search results |
 | `search_id` | string | ✓ | Search ID from the search that found this tool |
-| `params_to_tool` | string | ✓ | JSON string of parameters to pass to the tool |
+| `params_to_tool` | object | ✓ | A dictionary of parameters to pass to the tool |
 | `session_id` | string | | Session identifier (auto-generated if omitted) |
 | `max_response_size` | number | | Max response size in bytes (default: 20480) |
 
@@ -109,7 +109,7 @@ Execute a discovered tool with specific parameters.
 {
   "tool_id": "openweathermap.weather.execute.v1",
   "search_id": "abcd1234-ab12-ab12-ab12-abcdef123456",
-  "params_to_tool": "{\"city\": \"London\", \"units\": \"metric\"}"
+  "params_to_tool": {"city": "London", "units": "metric"}
 }
 ```
 

packages/mcp/src/api/client.ts

@@ -118,7 +118,15 @@ export class QverisClient {
         throw error;
       }
 
-      return response.json() as Promise<T>;
+      try {
+        return await response.json() as T;
+      } catch {
+        const error: ApiError = {
+          status: response.status,
+          message: 'Invalid or empty JSON response from API',
+        };
+        throw error;
+      }
     } catch (err: unknown) {
       if (err && typeof err === 'object' && 'status' in err) {
         // ApiError — rethrow as-is