An intelligent multi-dimensional data table field processing method and system for integrated MCP tool invocation

Abstract

This invention discloses a method and system for processing intelligent multidimensional data table fields integrating MCP tool invocation, belonging to the technical field of intelligent multidimensional data table software and MCP tool invocation integration. The method includes: defining a tool invocation field type in the field system of the intelligent multidimensional data table, wherein the field type has field configuration attributes including tool identifier, parameter template, and execution strategy; storing the field configuration attributes in response to configuration operations; obtaining the field configuration attributes of the field instance in the target record in response to a triggered event; constructing tool invocation request parameters based on the parameter template and the field values ​​of the target record; sending a tool invocation request to an external tool server and receiving a response; extracting result data from the response; and writing the result data into the cell corresponding to the field instance in the target record. This invention integrates MCP tool invocation capability as a native attribute of the field type into the intelligent multidimensional data table field system, realizing field-level tool configuration, dynamic parameter construction, accurate result backfilling, and automated linkage processing, improving the system's scalability and interoperability.

Description

Technical Field

[0001] This invention relates to the field of intelligent multidimensional data table software and MCP tool integration technology, specifically to an intelligent multidimensional data table field processing method and system that integrates MCP tool calls. Background Technology

[0002] Intelligent multidimensional tables (such as Lark Multidimensional Tables, Airtable, SeaTable, etc.) are software tools that organize and manage data in tabular form. These tools combine the flexible operation characteristics of spreadsheets with the structured data management characteristics of databases and have been widely used in scenarios such as project management, data analysis, and rapid construction of business systems.

[0003] In practical applications of intelligent multidimensional tables, users often need to call external tools or services to process the data in the tables. For example, they might call translation services to translate the text content in the table into other languages, call AI inference services to perform intelligent analysis of the table content, or call web search services to obtain supplementary information and populate the table. However, the current integration methods between intelligent multidimensional tables and external tools have many shortcomings.

[0004] Patent application CN109255114A discloses a solution for automatically generating Excel spreadsheets by calling an external module via command line. This solution employs a three-layer architecture: an Excel template and JSON file configuration module, a command-line execution module, and a spreadsheet generation module. By configuring the Excel template and JSON file, the external module can be called via command line to automatically generate Excel spreadsheets and illustrations.

[0005] However, this solution uses batch processing for external tool calls, and command-line calls are executed as a whole. It cannot configure tool call logic at the field level, nor does it support dynamic parameter construction and result backfilling at the field level. The limitation of this solution is that external tool calls are independent of the table field system, and it cannot utilize the reference relationships and change event mechanisms between table fields to achieve automated, interconnected processing.

[0006] US Patent Application Publication No. US20240303421A1 discloses a method for integrating a large language model into an intelligent multidimensional data table environment. The system receives user natural language input and sample table data, generates prompts containing problem statements and solution requests, sends these prompts to a large language model service, and performs data cleaning and solution processing based on the large language model's response.

[0007] However, in this solution, the large language model service call is an application-level auxiliary function, not a native capability of the field type. Users need to input natural language each time to trigger the call; automated tool calls cannot be achieved through declarative configuration. This solution does not support field-level parameter templates and execution strategy configuration, nor does it support automatic triggering based on field changes.

[0008] In addition, the following technical problems are also common in the existing technology.

[0009] First, the parameter construction lacks dynamism. In the existing solution, external service call parameters are statically configured or manually entered, and it is impossible to dynamically obtain the values ​​of other fields in the same record as call parameters using field reference syntax, resulting in a lack of linkage between parameters and record data.

[0010] Second, the accuracy of the backfilled results is insufficient. In the existing solution, the results of external service calls need to be manually copied to the corresponding cells in the table by the user, or written back through complex mapping configurations, which is cumbersome and prone to errors.

[0011] Third, the execution strategy is too simplistic. The existing solution lacks a flexible execution strategy configuration mechanism, making it difficult to adapt to the differentiated needs of different business scenarios (such as initial data batch filling, single record verification, full data regeneration, etc.).

[0012] Fourth, there is a lack of field linkage mechanism. The existing solution does not support configuring a list of source fields, and cannot automatically trigger the re-execution of external tools when the source field value changes, thus failing to form an automated data processing chain.

[0013] Fifth, there is insufficient standardization of protocols. Existing solutions mostly rely on private APIs or specific platform interfaces to connect with external tools, lacking native support for open protocols such as MCP, which limits system scalability and interoperability. Summary of the Invention

[0014] The present invention aims to solve at least one of the problems existing in the prior art and provides a method and system for processing intelligent multidimensional data table fields that integrates MCP tool invocation. This solution integrates the MCP tool invocation capability as a native feature of the field type into the intelligent multidimensional data table field system to realize field-level tool configuration, dynamic parameter construction, accurate result backfilling and automated linkage processing.

[0015] To solve the above-mentioned technical problems, the present invention provides the following technical solution: The first step is to define a tool call field type in the field system of the intelligent multidimensional data table. The tool call field type has field configuration attributes, which include tool identifier, parameter template and execution strategy.

[0016] The second step is to store the field configuration attributes of the field instance in response to a configuration operation on a field instance of the field type invoked by the tool.

[0017] The third step is to retrieve the field configuration attributes of the field instance in the target record in response to the tool call trigger event.

[0018] The fourth step is to construct the tool call request parameters based on the parameter template in the field configuration attribute and the field value of the target record.

[0019] The fifth step is to send a tool invocation request containing the tool identifier and the tool invocation request parameters to an external tool server.

[0020] The sixth step is to receive the tool call response returned by the external tool server.

[0021] The seventh step is to extract the result data from the tool call response.

[0022] The eighth step is to write the result data into the cell corresponding to the field instance in the target record.

[0023] Preferably, in the fourth step, the method constructs tool call request parameters based on the parameter template in the field configuration attribute and the field value of the target record, specifically including: The first sub-step involves the method obtaining the value template string for each parameter in the parameter template, wherein the value template string contains field reference placeholders identified in a predetermined format; The second sub-step involves the method parsing the value template string, identifying the field reference placeholders, and extracting the source field identifiers referenced therein. The third sub-step involves the method reading the corresponding source field value from the target record based on the source field identifier. The fourth sub-step involves replacing the field reference placeholders in the value template string with the source field values ​​to generate the final parameter values.

[0024] Preferably, in the third step, the tool call trigger event is a batch execution trigger event; the method further includes: determining the range of records to be processed according to the execution strategy in the field configuration attribute; wherein, the execution strategy includes at least one of the following: a manual trigger strategy, wherein the method executes the tool call in response to a user's manual trigger operation on a single record; a null-value priority batch strategy, wherein the method filters out records where the source field is not null and the cell corresponding to the field instance is null, and selects the first N records in a predetermined order to execute the tool call; and a full column execution strategy, wherein the method executes the tool call on all records where the source field is not null.

[0025] Preferably, the field configuration attributes further include a source field list and an auto-update identifier; the method further includes: the method listens for field value change events in the intelligent multidimensional data table; when a change is detected in the value of any field involved in the source field list, the method determines whether it needs to be automatically re-executed based on the auto-update identifier; if automatic re-execution is required, the method automatically triggers a tool call to the field instance in the record where the change occurred.

[0026] Preferably, after the method automatically triggers a tool call to the field instance in the record where the change is located, it further includes: the method recording a set of field identifiers in the current execution chain; if the field to be triggered already exists in the set of field identifiers, the method skips the trigger and records a loop warning log.

[0027] Preferably, in the fifth step, the method sends a tool invocation request containing the tool identifier and the tool invocation request parameters to an external tool server, specifically including: the method constructing a request body according to the MCP protocol specification, the request body containing a tool name field and a parameter field; and the method sending the request body to the external tool server.

[0028] Preferably, in the seventh step, the method extracts result data from the tool call response, specifically including: the method parses the response body of the tool call response; the method extracts text-type entries from the output content array in the response body; and the method uses the extracted text content as the result data.

[0029] Solution 2: An intelligent multidimensional data table field processing system integrating MCP tool calls, the system comprising: The field definition module is used to define tool call field types in the field system of the intelligent multidimensional data table and manage the field configuration attributes of field instances of that type. The field configuration attributes include tool identifier, parameter template and execution strategy. A configuration storage module is configured to store the field configuration attributes of the field instance; The trigger management module is used to respond to tool call trigger events and obtain the field configuration attributes of the field instance in the target record; The parameter construction module is used to construct tool call request parameters based on the parameter template in the field configuration attribute and the field value of the target record. The protocol communication module is used to send a tool call request containing the tool identifier and the tool call request parameters to an external tool server, and to receive a tool call response returned by the external tool server; The response processing module is used to extract result data from the tool call response and write the result data into the cell corresponding to the field instance in the target record.

[0030] Preferably, the parameter construction module includes: The template parsing unit is used to obtain the value template string of each parameter in the parameter template, parse and identify the field reference placeholders and the source field identifiers referenced therein; A field reading unit, configured to read the corresponding source field value from the target record based on the source field identifier; An interpolation replacement unit is used to replace the field reference placeholder in the value template string with the source field value to generate the final parameter value.

[0031] Preferably, the trigger management module includes: A manual triggering unit, which is used to respond to a user's call to a tool to perform a manual triggering operation on a single record; A batch execution unit is used to determine the range of records to be processed according to an execution strategy and to batch execute tool calls. The execution strategy includes at least one of a null value priority batch strategy and a full column execution strategy. An automatic linkage unit is used to listen for field value change events. When a change in the value of any field in the source field list in the field configuration attribute is detected, the tool is automatically invoked.

[0032] Preferably, the automatic linkage unit is further configured to: record the set of field identifiers in the current execution chain; if the field to be triggered already exists in the set of field identifiers, then skip the trigger and record a loop warning log.

[0033] Preferably, the protocol communication module constructs a request body containing a tool name field and a parameter field according to the MCP protocol specification and sends it to the external tool server; the response processing module extracts the text-type entries from the output content array of the response body of the tool call response as the result data.

[0034] Option 3: The present invention also provides a computer device, the computer device including a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in any of the above embodiments.

[0035] Option 3: The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method described in any of the above embodiments.

[0036] Compared with the prior art, the present invention has the following advantages: (1) This invention integrates the MCP tool calling capability as a native attribute of the field type into the intelligent multidimensional data table field system, so that the way users configure and use external tools is consistent with other field types, without having to switch to independent plugins or external platforms, effectively reducing the learning cost for users and improving the consistency of system operation.

[0037] (2) This invention enables parameter templates to support field reference placeholder syntax through a field reference interpolation parameter dynamic construction mechanism, which can dynamically reference the values ​​of other fields in the same record as tool call parameters, realize data linkage between fields, and greatly improve the flexibility and automation of parameter construction.

[0038] (3) This invention supports multiple strategies such as manual triggering, batch execution with null values ​​as priority, and full column execution through a multi-strategy execution engine, which meets the differentiated needs of different business scenarios.

[0039] (4) This invention uses a source field linkage automatic triggering mechanism to automatically trigger the tool to re-execute when the source field value changes, forming an automated data processing link from source data change to automatic tool invocation and result backfilling.

[0040] (5) This invention uses a standardized tool calling mechanism based on the MCP open protocol to connect to any external tool server that conforms to the MCP specification. There is no need to develop an adapter for each tool, which greatly improves the scalability and interoperability of the system.

[0041] (6) This invention effectively prevents security risks such as malicious code injection, unauthorized tool calls and unauthorized data access through a multi-layer security protection system. Attached Figure Description

[0042] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0043] Figure 1 This is an architecture diagram of an intelligent multidimensional data table field processing system that integrates the MCP tool and is provided in an embodiment of the present invention.

[0044] Figure 2 This is a schematic diagram of the overall process of the intelligent multidimensional data table field processing method called by the integrated MCP tool provided in the embodiments of the present invention.

[0045] Figure 3 This is a flowchart illustrating the parameter interpolation mechanism in an embodiment of the present invention.

[0046] Figure 4This is a flowchart illustrating the batch execution strategy in an embodiment of the present invention.

[0047] Figure 5 This is a flowchart illustrating the source field linkage automatic triggering mechanism in an embodiment of the present invention. Detailed Implementation

[0048] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below with reference to specific illustrations.

[0049] To address the technical problems existing in the integration of intelligent multidimensional data tables with external tools, such as "decoupling of external tool calls from the field system", "lack of dynamic parameter construction", "insufficient accuracy of result backfilling", "single execution strategy", "lack of field linkage mechanism" and "insufficient protocol standardization", this invention proposes a core technical solution that integrates the MCP tool call capability as a native attribute of the field type into the field system.

[0050] The core innovation of this solution lies in three aspects: Firstly, a dedicated tool invocation field type is defined in the field system of the intelligent multidimensional data table, so that each field instance can be independently configured with attributes such as tool identifier, parameter template, execution strategy and source field list, transforming the MCP tool invocation capability from an external function to an inherent capability of the field.

[0051] Secondly, a field reference placeholder syntax is introduced into the parameter template. When the tool is executed, the system automatically replaces the placeholder with the actual value of the corresponding field in the current record, realizing the dynamic construction of parameter values ​​and the linkage between fields.

[0052] Thirdly, it natively supports the MCP (Model Context Protocol) open protocol, enabling it to connect to any external tool server that conforms to the MCP specification through standardized request / response formats.

[0053] These three aspects work together to form a complete automated closed loop, from field configuration to parameter construction, tool invocation, result backfilling, and linked updates.

[0054] like Figure 2 As shown, the intelligent multidimensional data table field processing method called by the integrated MCP tool provided by the present invention includes eight steps in total, and the steps form a complete execution loop through data flow and control flow.

[0055] The first step is the field type definition step. This step configures the registration of a new field type, namely the tool-invoking field type, in the field system of the intelligent multidimensional data table. The technical feature of this step is that it integrates the MCP tool-invoking capability into the native system of the field type, giving this field type the same status and configurability as other basic field types. This step provides the type foundation for all subsequent steps; without this step, users will not be able to create field instances with MCP tool-invoking capabilities in the table.

[0056] The second step is the field configuration storage step. This step is configured to persistently store the various configuration attributes of the field instance in response to user configuration operations. The key technical feature of this step is the establishment of a mapping relationship between the field instance and its MCP tool invocation logic, allowing each field instance to independently possess a different tool identifier, parameter template, and execution strategy. This step receives the field type framework defined in the first step, fills in the user's personalized configuration into this framework, forms a complete field definition, and saves it.

[0057] The third step is the trigger response and configuration acquisition step. This step is configured to respond to various trigger events, determine the target record and target field that need to be called by the MCP tool, and load the complete configuration of that field. The technical feature of this step is that it supports three triggering methods: manual triggering, batch execution triggering, and automatic linkage triggering, allowing the execution process to enter from different entry points. This step connects user operations or system events with the subsequent parameter construction process and is the entry point for the entire execution process.

[0058] The fourth step is the parameter construction step. This step is configured to dynamically construct the actual parameters required for the MCP tool invocation request based on the parameter template and the field values ​​of the target record. The technical feature of this step is the introduction of a field reference interpolation mechanism, which allows the parameter values ​​to automatically change as the record data changes. This step is the key bridge connecting tabular data and the MCP tool invocation, and its output (parameter object) is the input of the fifth step.

[0059] The fifth step is the MCP tool invocation request sending step. This step is configured to encapsulate the tool identifier and constructed parameters into a request according to the MCP protocol format and send it to the MCP server. The technical feature of this step is that it implements standardized tool invocation based on the MCP open protocol, enabling the system to interface with any external tool that conforms to the MCP specification. This step receives the parameter object output from the fourth step, assembles it with the tool identifier, and then sends it.

[0060] The sixth step is the response receiving step. This step is configured to receive the HTTP response returned by the MCP server. This step is the pairing step of the fifth step, completing the receiving end of the request-response interaction.

[0061] The seventh step is the result extraction step. This step is configured to parse the response body according to the MCP protocol format and extract the result data from it. The technical features of this step are standardized response parsing and security filtering mechanisms. This step receives the raw response obtained in the sixth step, extracts the usable result data, and then passes it to the eighth step.

[0062] The eighth step is the result backfilling step. This step is configured to write the extracted result data into the cells corresponding to the target fields in the target records. The technical feature of this step is precise backfilling at the field level, without requiring users to manually copy or configure mapping rules. This step is the end point of the execution process, completing a full closed loop from triggering to result writing back.

[0063] The intelligent multidimensional data table field processing method based on the above steps is as follows: the first step is a preparatory step at the system definition level; the second step is a preparatory step at the configuration level; and the third to eighth steps are the core steps at the execution level. The triggering method of the third step determines the entry direction of the execution process; the parameter construction in the fourth step determines the accuracy of the MCP tool call; the protocol interaction in the fifth to seventh steps determines the system's scalability and compatibility; and the result backfilling in the eighth step determines the integrity of the integration and the user experience. Each step forms a close logical relationship through data transmission (field configuration, target record, parameter object, response data, and result data).

[0064] Based on an understanding of the overall function and logical coordination of the above eight steps, the following section provides a more detailed explanation of the specific implementation plan for each step from a technical implementation perspective.

[0065] In the first step described above, this method registers a new field type in the field type system of the intelligent multidimensional data table and names the field type "tool call field".

[0066] The metadata definition for this field type includes the following: A field type identifier, used to uniquely identify this field type within the system. A list of configurable attributes, including: a tool identifier attribute (string type), a parameter template attribute (object type, key is the parameter name, value is a configuration object containing a value template string and a type identifier), an execution policy attribute (enumeration type), a source field list attribute (string array type), an auto-update identifier attribute (boolean type), and an optional MCP server address attribute (string type). Attribute validation rules, including: the tool identifier must be in the system-configured whitelist, the parameter template value template string must be in a valid string format, and the execution policy must be a predefined enumeration value, etc.

[0067] After the field type is defined, when creating or editing a table field, the user can select this type from the field type drop-down list. The system dynamically renders the corresponding configuration interface according to the metadata definition of the field type. The technical effect of this step is that through the native extension of the field type system, the MCP tool invocation capability becomes a built-in function of the intelligent multi-dimensional data table, which users can use without installing plugins or writing scripts, effectively lowering the usage threshold.

[0068] In the above second step, after the user completes the configuration of the tool invocation field on the field configuration interface, the present method serializes the configuration information into a structured data object, and stores the data object as an attribute of the field instance in the table metadata storage.

[0069] The stored content includes the tool name, the name of each parameter and its value template string and type, execution strategy type, source field identification list, automatic update flag, and an optional MCP server address, etc. The table metadata storage can be implemented by a relational database or a document database, and the field configuration information is stored together with the table structure definition. The technical effect of this step is that through the structured configuration storage mechanism, the MCP tool invocation configuration of the field instance can be persistently stored, version-managed and migrated, providing a data basis for subsequent execution and linkage.

[0070] In the above third step, the present method supports three triggering modes, which are specifically as follows.

[0071] The first one is the manual triggering mode. The user clicks the execution button of the tool invocation field on the record editing interface, and the method obtains the unique identifier of the current record and the field identifier of the field instance, and reads the complete configuration of the field instance from the table metadata storage.

[0072] The second one is the batch execution triggering mode. The user selects the "batch execution" operation on the field configuration interface or the record list interface, the method obtains the field identifier of the target field instance, reads the configuration from the table metadata storage, and then determines the to-be-processed record list according to the execution strategy.

[0073] The third one is the automatic linkage triggering mode. After the field change monitoring module detects that a certain field value changes, the method queries whether the source field list of all tool invocation fields contains the changed field. If it is contained and the automatic update flag is true, the execution of the field instance is automatically triggered, and the target record is the record where the change occurs.

[0074] All three triggering methods share the same subsequent execution flow (i.e., steps four through eight), differing only in how the trigger source and the range of records to be processed are determined. This step allows for the configuration of multiple triggering methods to meet the needs of different business scenarios. Manual triggering is suitable for verifying and debugging single records, batch execution is suitable for initializing data population, and automatic linkage triggering is suitable for establishing automated data processing pipelines.

[0075] In the fourth step described above, this method constructs the MCP tool call request parameters based on the parameter template in the field configuration attributes and the field values ​​of the target record. For example... Figure 3 As shown, this step specifically includes the following sub-steps.

[0076] The first sub-step involves retrieving all field definitions from the table's metadata storage and constructing a field mapping table. This table uses field identifiers as keys and metadata containing database field names and types as values. This mapping table is used for subsequent field reference resolution.

[0077] The second sub-step involves this method iterating through each parameter configuration in the parameter template of the field configuration and performing subsequent processing on each parameter.

[0078] The third sub-step involves retrieving the template string of the current parameter value and using predefined regular expression matching rules to match all placeholders that begin and end with a specific symbol and consist of a combination of alphanumeric underscores. The capture group in each match result is the identifier of the referenced field.

[0079] In the fourth sub-step, for each extracted field identifier, this method checks whether the field identifier exists in the field mapping table and whether it is in the whitelist of source fields configured for the field. If the field does not exist or is not in the whitelist, this method throws an exception and terminates execution to prevent unauthorized access to field data.

[0080] In the fifth sub-step, for each field identifier that passes the validity check, this method retrieves the corresponding database field name from the field mapping table and then reads the actual value corresponding to that database field name from the target record object. The record object's key is the database field name, not the field identifier; this design ensures that record data is not affected when the field identifier changes.

[0081] In the sixth sub-step, this method replaces all placeholders in the value template string with the corresponding source field values ​​(converted to string form), generating the replaced parameter string. Then, this method performs type conversion on the replaced value according to the type identifier in the parameter configuration.

[0082] When the type is string, this method preserves the string format. When the type is number, this method attempts to convert the string to a number; if the conversion fails, a type validation exception is thrown. When the type is boolean, this method converts the string to a boolean value according to predefined rules. When the type is JSON, this method attempts to parse it into an object; if parsing fails, an exception is thrown. After a successful type conversion, this method uses the converted value as the final parameter value.

[0083] In the seventh sub-step, after all parameters have been processed, this method assembles the final parameter object. The key of this parameter object is the parameter name, and the value is the parameter value after type conversion.

[0084] In this step, through the field reference interpolation mechanism, the MCP tool can call parameters that automatically follow changes in the recorded data, realizing dynamic parameter construction of "configure once, adapt automatically", which greatly reduces the workload of users manually entering parameters and avoids the problem of parameters and data being out of sync.

[0085] In the fifth step above, this method constructs and sends an MCP tool call request in accordance with the MCP protocol specification.

[0086] Specifically, it includes the following sub-steps: In the first sub-step, the method constructs a request body according to the MCP protocol specification. The request body includes a tool name field (whose value is the tool identifier in the field configuration) and a parameter field (whose value is the parameter object constructed in the previous step).

[0087] The second sub-step involves adding authentication information, including a signature token, to the request. The token contains a request timestamp and validity period information; upon receiving the request, the MCP server verifies the token's validity and expiration date.

[0088] The third sub-step involves sending the request body to the configured MCP server address via the HTTP protocol (or using the system default address if not specified in the field configuration), and setting a request timeout. If no response is received within the timeout period, the call is terminated and a timeout log is recorded.

[0089] This step utilizes a standardized tool system based on the MCP open protocol, enabling the system to interface with any external tool server that conforms to the MCP specification. This eliminates the need to develop a separate adapter for each tool, significantly improving the system's scalability.

[0090] In the sixth step described above, the method receives an HTTP response from the MCP server, with the response body in JSON format. This step is a complementary step to the fifth step, completing the receiving end of the request-response interaction.

[0091] In the seventh step above, the method extracts result data from the MCP tool invocation response. In a specific implementation, the seventh step includes the following sub-steps: The first sub-step: the method parses the response body according to the MCP protocol specification. The MCP response body includes an output object, the output object includes a content array, and each entry in the content array includes a type field and a text field.

[0092] The second sub-step: the method traverses the content array, filters out entries whose type is text, extracts the text field values of these entries, and concatenates all text content into a result string.

[0093] The third sub-step: the method performs security filtering on the extracted text content, detects and removes any potentially contained malicious scripts, and retains only safe plain text content.

[0094] In this step, through the standardized response parsing and security filtering mechanism, it is ensured that the MCP tool invocation result can be accurately extracted and safely used, avoiding attacks on the system by malicious response content.

[0095] In the eighth step above, the method writes the result data into the cell corresponding to the field instance in the target record. In a specific implementation, it specifically includes: the method determines the target storage position — the unique identifier of the target record and the field identifier of the target field have been determined in the third step. The method constructs a field value update object, taking the field identifier (or database field name) as the key and the extracted result text as the value. The method invokes the update interface for record data storage to write the result data into the target cell. The update operation triggers a field change event, which provides a basis for subsequent linkage.

[0096] In this step, the result is automatically and accurately written back to the cell where the specified field is located, without requiring the user to manually copy or configure complex mapping rules, which eliminates the problems of cumbersome operations and high error proneness.

[0097] In addition, the method further includes the following steps in a batch execution scenario, as Figure 4 shown in the accompanying drawings.

[0098] The method reads the execution strategy attribute value from the field configuration. If the strategy is a manual trigger strategy, the method is executed only for the current single record, and no batch processing is involved.

[0099] If the strategy is a null-first batch strategy, the method scans all records in the table and filters out records that simultaneously meet the following two conditions: at least one field in the source field list has a non-null value, and the target field (the field currently invoked by the tool) has a null value or an empty string. The filtered records are sorted in ascending order of creation time, and the first N records are taken (N is specified by the configuration, with a default value of 10). The method iterates through these N records, executing the MCP tool call for each one.

[0100] If the strategy is to execute all records in the column, the method scans all records in the table, filters out all records where at least one field in the source field list is not empty, with no upper limit on the number of rows. The method iterates through all records that meet the conditions and executes the MCP tool call one by one. The three execution strategies cover the needs of different business scenarios. The manual strategy is suitable for single validation, the null-value priority batch strategy is suitable for initialization and filling (avoiding overwriting existing data), and the all-column execution strategy is suitable for full updates.

[0101] In some implementations, the method also implements an automatic triggering mechanism for source field linkage, such as... Figure 5 As shown, the specific steps include:

[0102] The method registers a field change event listener in the record data storage layer. When the value of any field changes, the listener is triggered and obtains the change information. The change information includes the record identifier of the change, the field identifier of the change, the value before the change, and the value after the change.

[0103] When a field change event is detected, the method queries all tool call field instances and filters out a list of field instances whose source field list contains the changed field identifier and whose identifier is automatically updated to true.

[0104] For each field instance selected, the method automatically triggers the execution process (from the third to the eighth step) with the record where the change occurred as the target record.

[0105] To prevent circular dependencies between fields from causing infinite triggering, the method records a set of field identifiers in the current execution chain each time a linkage is triggered. If the field to be triggered already exists in this set, the method skips the trigger and logs a circular warning.

[0106] Based on these steps, this method achieves an automated data processing chain of "source data change → automatic invocation of MCP tool → result backfilling" through field change monitoring and automatic linkage triggering, realizing real-time data synchronization. A circular dependency protection mechanism ensures the stable operation of the system.

[0107] Based on the detailed explanation of the method and process above, the following section combines... Figure 1 A system that implements the above method is described.

[0108] like Figure 1 As shown, the intelligent multidimensional data table field processing system called by the integrated MCP tool provided by this invention adopts a layered architecture, including a client layer, an application service layer, a protocol layer, an external service layer, and a data storage layer. The layers communicate and collaborate with each other through standardized interfaces.

[0109] The client layer comprises a user interface, a field configuration module, and a record editing module. Its function is to provide a user interface, receive user configuration inputs and operation commands, and pass user requests to downstream layers. The key technical feature of the client layer is the deep integration of the field configuration and record editing interfaces with the tool's field type calls, ensuring that users configure and use the MCP tool in a way consistent with other field types.

[0110] The application service layer is the core logic layer of the system, comprising a field definition module 10, a trigger management module 20, a parameter construction module 30, a protocol communication module 40, and a response processing module 50. This layer carries the core business logic and coordinates the various modules to complete the entire execution process from triggering to result backfilling. The technical feature of this layer lies in its clear modular division of responsibilities such as field type management, trigger management, parameter construction, protocol communication, and response processing. The modules communicate with each other through data objects, forming a loosely coupled and highly cohesive architecture.

[0111] The protocol layer includes an MCP protocol client 60 and a security verification module 70. The function of this layer is to communicate with the MCP server according to the MCP protocol specification and to perform security verification on requests and responses. The technical features of this layer are that it implements standardized tool calls based on the MCP open protocol, while incorporating multi-layered security protection mechanisms.

[0112] The external service layer includes the MCP server 80 and various tools 81, 82, etc., deployed on it. The function of this layer is to provide specific MCP tool execution capabilities, receive MCP tool call requests, and return execution results. The technical feature of this layer lies in the standardized registration and routing mechanism between tools and the MCP server, enabling the system to interface with any tool that conforms to the MCP specification.

[0113] The data storage layer includes table metadata storage (91), record data storage (92), and audit log storage (93). The function of this layer is to persistently store field configuration information, record data (including MCP tool call results), and audit logs. The key technical feature of this layer is that each of the three storage types performs its specific function, supporting configuration management, data storage, and security auditing requirements respectively.

[0114] In this intelligent multidimensional data table field processing system, the client layer receives user configurations and trigger operations, the application service layer processes business logic, the protocol layer completes protocol interaction and security verification, the external service layer executes MCP tool calls, and the data storage layer persists configurations, data, and logs. This layered architecture ensures that each layer of the system has clear responsibilities and can evolve and be replaced independently.

[0115] Based on an understanding of the overall functions and logical relationships of each layer of the system, the following section provides a more detailed explanation of the specific structural schemes of each component from a technical implementation perspective.

[0116] In the client-side layer, the user interface is implemented using a graphical user interface. In the field configuration interface, the user interface provides a dedicated configuration panel for the field types that the tool calls. This panel contains the following controls: a tool name dropdown selection box, loading a list of available MCP tools from the system whitelist; a parameter template editing area, supporting key-value pair parameter configuration, with value fields supporting auto-completion and syntax hints for field reference placeholder syntax; an execution strategy radio button group, including three options: manual trigger, batch execution with null values ​​priority, and execution of all columns; a source field multi-select dropdown list, listing all fields in the current table for the user to select; and controls such as an automatic update switch.

[0117] In the record editing interface, the cells of the tool call fields display the execution status (including statuses such as "Pending Execution", "Executing", "Completed", "Failed") and the execute button.

[0118] The field configuration module is responsible for collecting the configuration information entered by the user in the configuration interface, performing integrity checks (such as the tool name cannot be empty and the parameter template format must be correct) and validity checks (such as the tool name must be in the whitelist). After the checks are passed, the configuration information is passed to the field definition module 10 for persistent storage.

[0119] The record editing module is responsible for displaying the record data, listening for user clicks on the tool's field execution button, and passing the execution request (including record identifier and field identifier) ​​to the trigger management module 20.

[0120] In the application service layer, the field definition module 10 manages the metadata of table fields and provides interfaces for creating, reading, updating, and deleting fields. For tool call field types, the field definition module 10 is responsible for serializing the field configuration attributes (tool name, parameter template, execution strategy, source field list, automatic update identifier, etc.) and storing them in the table metadata storage 91, and deserializing them into structured objects when reading.

[0121] The trigger management module 20 is the entry manager for the system execution process, and it contains three functional units.

[0122] The manual trigger unit receives a single record manual trigger request from the record editing module, obtains the configuration of the target record and field instance, and starts the execution process.

[0123] The batch execution unit receives batch execution requests from users, calls batch execution logic according to the execution strategy in the field configuration, and starts the execution process one record at a time after determining the list of records to be processed.

[0124] The automatic linkage unit listens for change events from the field change monitoring module, queries dependent fields (the source field list contains the changed field and the auto-update flag is true), and initiates the execution process for each dependent field. This unit also records the set of field flags in the current execution chain. If the field to be triggered already exists in this set, the trigger is skipped and a loop warning log is recorded.

[0125] The parameter construction module 30 is the core computing engine of the system, which contains the following three functional units.

[0126] The template parsing unit receives parameter templates from the field configuration, iterates through the value template string of each parameter, uses regular expression matching to identify field reference placeholders and extracts the list of referenced field identifiers.

[0127] The field reading unit reads the corresponding source field value from the target record based on the field identifier list extracted by the template parsing unit and the field mapping table.

[0128] The interpolation replacement unit replaces the placeholders in the value template string with the corresponding source field values, performs type conversion according to the parameter type configuration, and generates the final parameter object.

[0129] The protocol communication module 40 is responsible for protocol interaction with the MCP server, constructing request bodies, sending HTTP requests, and receiving HTTP responses according to the MCP protocol specifications, and includes request timeout control, retry mechanism, and error handling logic.

[0130] The response processing module 50 is responsible for parsing the response body returned by the MCP server, extracting text-type entries from the content array, performing security filtering on the text content, and then calling the update interface of the record data storage 92 to write the results to the target cell.

[0131] In the protocol layer, the MCP protocol client 60 implements the client logic of the MCP protocol specification, which includes a request builder and a response parser.

[0132] The request builder receives the tool name and parameter object, assembles the request body (including the tool name field and parameter field) according to the MCP protocol specification, and generates a complete HTTP request after adding the authentication token.

[0133] The response parser receives the HTTP response, parses the response body according to the MCP protocol specification, and extracts the text content from the content array.

[0134] The security verification module 70 performs security verification before the request is sent and after the response is received.

[0135] The pre-request validation includes: verifying whether the tool name is in the system configuration whitelist, verifying whether each parameter value matches the configured type, and verifying whether the field reference is in the source field list whitelist.

[0136] The post-response verification performs security filtering on the extracted text content to remove potential malicious scripts.

[0137] In the external service layer, MCP server 80 is deployed on a remote server, exposing an HTTP interface for tool invocation that conforms to the MCP protocol specification. MCP server 80 maintains a list of registered MCP tools and their metadata (including tool name, description, parameter structure definition, etc.), receives MCP tool invocation requests, routes them to the corresponding tool executor, collects the execution results, and returns a response in the MCP protocol format.

[0138] Tools 81, 82, etc., are specific implementations of MCP tools, including but not limited to translation tools, web search tools, code execution tools, and AI inference tools. Each MCP tool implements specific business logic, receives structured parameter input, and returns structured result output.

[0139] In the data storage layer, table metadata storage 91 is implemented using a relational database or document database, storing the table's structure definition, including a list of fields, the type of each field, and configuration attributes. For fields invoked by tools, their complete configuration objects are stored.

[0140] Data storage (92) is implemented using a relational database or document database to store tabular records. Each record is a data object, with the key being the database field name and the value being the field value. The execution result of the MCP tool that calls a field is stored as text in the cell corresponding to that field.

[0141] The audit log storage 93 is implemented using a log database or file system, storing a complete audit record of each MCP tool call. The audit record includes call time, record identifier, field identifier, tool name, request parameters (after anonymization), response result (after truncation), execution time (milliseconds), status (success / failure / timeout), and error information (if any). The audit log storage 93 supports multi-dimensional retrieval by time, record, field, tool, etc., facilitating security audits and anomaly tracing.

[0142] The intelligent multidimensional data table field processing system, which integrates the MCP tool based on the above scheme, operates according to the following logical process: During the configuration phase, the user creates a new field through the user interface and the field configuration module, selects the "tool call" field type, and fills in the various configuration options. The field configuration module then performs integrity and validity checks on the configuration information. After successful verification, the field configuration module calls the field definition module 10 to serialize the configuration information and store it in the table metadata storage 91.

[0143] During the execution phase, after receiving a trigger event, the trigger management module 20 loads the complete configuration of the field from the table metadata storage 91 based on the field identifier. Then, the trigger management module 20 calls the parameter construction module 30, which performs field reference interpolation based on the parameter template in the configuration and the field value of the target record to generate the final parameter object.

[0144] After the parameters are constructed, the parameter construction module 30 passes the parameter object to the protocol communication module 40. After verification by the security verification module 70, the protocol communication module 40's request constructor constructs a request body according to the MCP protocol and adds a security token. The MCP protocol client 60 then sends the request to the MCP server 80.

[0145] After receiving the request and returning a response, the MCP server 80 parses the response body and extracts the text content for security filtering. Finally, the response processing module 50 calls the update interface of the record data storage 92 to write the result to the target cell, and simultaneously writes the complete information of this call to the audit log storage 93.

[0146] If the result backfilling triggers a field change event, the automatic linkage unit will intervene again, querying other tools that depend on that field to call the field and triggering a new round of execution. The circular dependency protection mechanism ensures system security.

[0147] In batch execution scenarios, the batch execution unit first scans the record data storage (92 records) according to the execution strategy (empty value priority batch strategy or full column execution strategy) and filters out a list of records that meet the criteria. Then, the batch execution unit traverses this list, repeating the above execution process for each record. Finally, the batch execution unit summarizes the execution results (success count / failure count) and provides feedback to the user.

[0148] The following example, using cross-language content translation, illustrates in detail the application and implementation process of the present invention.

[0149] A content management platform needs to translate article titles and author information into multiple languages. The platform maintains an article table containing fields such as article ID, original title (in Chinese), author name, English title, and Japanese title. Users want the system to automatically generate English and Japanese titles using the MCP translation tool when adding articles or modifying the original title, eliminating the need for manual translation by the user.

[0150] The system deployment includes the following: Deploying the system described in this invention, comprising a client layer, an application service layer, a protocol layer, and a data storage layer. Deploying an MCP translation server, registering MCP translation tools on this server, and having these tools receive two parameters: the text to be translated and the target language, and return the translated text. Configuring a tool whitelist within the system to ensure that only whitelisted MCP tools are allowed to be called.

[0151] The administrator added two tool call fields to the article table.

[0152] The English title field is configured as follows: the tool name is MCP Translation Tool; in the parameter template, the source text value template string is "Article Title: {Original Title}, Author: {Author}", and the target language is fixed as "English"; the source field list specifies the original title and author; the execution strategy is a null-first batch strategy with 20 rows; and automatic update is set to true.

[0153] The configuration for the Japanese title field is as follows: the tool name is MCP Translation Tool; in the parameter template, the source text value template string is "{Original Title}", and the target language is fixed as "Japanese"; the source field list specifies the original title; the execution strategy is to execute all columns; and automatic update is set to true.

[0154] In the initial batch population scenario, the table already contains 100 article records, with original titles and authors already in the data, but the English and Japanese titles are empty. When the user clicks the "Batch Execute" button in the English title field, the system filters out the first 20 records with non-empty source fields and empty English titles based on the empty value priority batch strategy, and then executes them one by one.

[0155] For each record, the system loads the configuration, replaces the original title and author placeholders in the parameter template with the actual values ​​for the current record, constructs an MCP request and sends it to the MCP translation server. Upon receiving the response, the system extracts the translation result text and writes it into the English title cell of that record, while simultaneously logging the audit log. After all 20 records are completed, the user will see that the English titles of the first 20 records have been automatically filled in.

[0156] In an automatic update scenario, a user edits an existing record, changing the original title from "AI Technology Development Trends" to "Machine Learning Application Practices." The field change monitoring module detects the change in the original title field and finds that the source field lists for both the English and Japanese title fields contain the original title and are automatically updated to true. Therefore, it automatically triggers the re-execution of these two fields.

[0157] The system performs parameter interpolation on the two fields, sends an MCP request, parses the response, and backfills the results, writing the new translation results into the corresponding cells. Both MCP tool calls are logged in the audit log. The entire update process requires no additional user intervention.

[0158] In the scenario of regenerating the entire column, if the MCP translation tool is upgraded and the user wants to regenerate the Japanese titles of all articles, they can click the "Batch Execution" button in the Japanese title field. The system will then filter out all records (100 records) whose original titles are not empty according to the full column execution strategy, call the MCP translation tool again for each record, and write the new results back to complete the full update.

[0159] The above application examples fully demonstrate the technical effectiveness of the present invention. Automation is significantly improved; translation results are automatically updated when the original title or author changes, reducing most manual operations. Data consistency is guaranteed; the linkage mechanism ensures that translation results are always synchronized with the source data. Batch processing is flexible and efficient; the null value priority strategy avoids overwriting existing data, and the full column execution strategy facilitates full updates. Security is comprehensively guaranteed through whitelists, type validation, response filtering, and audit logs. Based on the MCP open protocol, the system can be easily extended to connect with other MCP tools.

[0160] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for processing intelligent multidimensional data table fields by integrating MCP tool calls, characterized in that, The method includes the following steps: The first step is that the method defines a tool call field type in the field system of the intelligent multidimensional data table. The tool call field type has field configuration attributes, which include tool identifier, parameter template and execution strategy. In the second step, the method, in response to a configuration operation on a field instance of a field type invoked by the tool, stores the field configuration attribute of the field instance; The third step is that, in response to a tool call trigger event, the method obtains the field configuration attribute of the field instance in the target record; The fourth step involves the method constructing tool call request parameters based on the parameter template in the field configuration attribute and the field value of the target record. Fifth step: The method sends a tool invocation request containing the tool identifier and the tool invocation request parameters to an external tool server; The sixth step is that the method receives the tool invocation response returned by the external tool server; The seventh step is to extract result data from the tool call response; The eighth step is to write the result data into the cell corresponding to the field instance in the target record.

2. The method according to claim 1, characterized in that, In the fourth step, the method constructs tool call request parameters based on the parameter template in the field configuration attribute and the field value of the target record, specifically including: The first sub-step involves the method obtaining the value template string for each parameter in the parameter template, wherein the value template string contains field reference placeholders identified in a predetermined format; The second sub-step involves the method parsing the value template string, identifying the field reference placeholders, and extracting the source field identifiers referenced therein. The third sub-step involves the method reading the corresponding source field value from the target record based on the source field identifier. The fourth sub-step involves replacing the field reference placeholders in the value template string with the source field values ​​to generate the final parameter values.

3. The method according to claim 1, characterized in that, In the third step, the tool invokes a batch execution trigger event; the method further includes: The range of records to be processed is determined based on the execution strategy in the field configuration attributes. The execution strategy includes at least one of the following: The manual triggering strategy is a method that responds to a user's manual triggering operation for a single record by invoking a tool. The null-value priority batch strategy filters out records where the source field is not null and the corresponding cell of the field instance is empty, and selects the first N records in a predetermined order to execute the tool call; The full-column execution strategy involves calling the tool on all records where the source field is not empty.

4. The method according to claim 1, characterized in that, The field configuration attributes also include a list of source fields and an auto-update flag; the method also includes: The method listens for field value change events in the intelligent multidimensional data table; When a change in the value of any of the fields in the source field list is detected, the method determines whether it needs to be automatically re-executed based on the automatic update flag; If automatic re-execution is required, the method automatically triggers a tool call to the field instance in the record where the change occurs.

5. The method according to claim 1, characterized in that, In the fifth step, the method sends a tool invocation request containing the tool identifier and the tool invocation request parameters to an external tool server, specifically including: The method constructs a request body according to the MCP protocol specification, and the request body includes a tool name field and a parameter field; The method sends the request body to the external tool server.

6. The method according to claim 1, characterized in that, In the seventh step, the method extracts result data from the tool call response, specifically including: The method parses the response body of the tool call response; The method extracts text-type entries from the output content array of the response body; The method uses the extracted text content as the result data.

7. A smart multidimensional data table field processing system integrating MCP tool calls, characterized in that, The system includes: The field definition module is used to define tool call field types in the field system of the intelligent multidimensional data table and manage the field configuration attributes of field instances of that type. The field configuration attributes include tool identifier, parameter template and execution strategy. A configuration storage module is configured to store the field configuration attributes of the field instance; The trigger management module is used to respond to tool call trigger events and obtain the field configuration attributes of the field instance in the target record; The parameter construction module is used to construct tool call request parameters based on the parameter template in the field configuration attribute and the field value of the target record. The protocol communication module is used to send a tool call request containing the tool identifier and the tool call request parameters to an external tool server, and to receive a tool call response returned by the external tool server; The response processing module is used to extract result data from the tool call response and write the result data into the cell corresponding to the field instance in the target record.

8. The system according to claim 8, characterized in that, The parameter construction module includes: The template parsing unit is used to obtain the value template string of each parameter in the parameter template, parse and identify the field reference placeholders and the source field identifiers referenced therein; A field reading unit, configured to read the corresponding source field value from the target record based on the source field identifier; An interpolation replacement unit is used to replace the field reference placeholder in the value template string with the source field value to generate the final parameter value.

9. The system according to claim 8, characterized in that, The trigger management module includes: A manual triggering unit, which is used to respond to a user's call to a tool to perform a manual triggering operation on a single record; A batch execution unit is configured to determine the range of records to be processed according to the execution strategy and to batch execute tool calls. The execution strategy includes at least one of a null value priority batch strategy and a full column execution strategy. An automatic linkage unit is used to listen for field value change events. When a change in the value of any field in the source field list in the field configuration attribute is detected, the tool is automatically invoked.

Images