Agent decision-making responsibility management and control method and system based on declarative primitive state machine

Abstract

This application provides a method and system for controlling the decision-making authority and responsibility of an intelligent agent based on a declarative primitive state machine, comprising: parsing a declarative primitive request for an intelligent agent decision-making task to generate an authorization primitive instruction unit, and generating an authorization state authority and responsibility record based on the authorization primitive instruction unit; parsing the authorization state authority and responsibility record for judgment admission elements to obtain a judgment admission element description, performing admission matching between the judgment primitive instruction unit and the judgment admission element description to generate an evidence-bound judgment record; reconstructing the evidence-bound judgment record for verification elements to obtain a verification comparison element description, generating a verification state authority and responsibility record based on the verification comparison element description and the determined verification primitive instruction unit to be verified; reconstructing the verification state authority and responsibility record for termination elements to generate a termination constraint element description, and generating an intelligent agent decision-making authority and responsibility control evidence storage report based on the termination constraint element description.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence governance, intelligent agent decision control, and automated auditing technologies. Specifically, it relates to a method and system for intelligent agent decision-making responsibility management based on declarative primitive state machines. Background Technology

[0002] With the development of large-scale intelligent agents, multi-agent collaborative systems, and automated task execution platforms, intelligent agents can automatically invoke tools, generate decision results, and execute subsequent operations based on input tasks in complex business scenarios. This operational mode places higher demands on system access control, operation process auditing, decision result verification, and permission revokement after task completion. Especially in scenarios of continuous multi-agent collaboration, the system needs to maintain a traceable technical record of every task trigger, decision output, verification feedback, and task termination process.

[0003] Existing intelligent agent operation management solutions typically employ a combination of ordinary access tokens and operation logs. This solution assigns an access token to the intelligent agent before the task begins and records interface call logs during the agent's execution. When the intelligent agent generates a decision output, the system writes the output content and call time to the log file. After the task ends, the permission management component invalidates the access token based on the session state.

[0004] However, the aforementioned solutions primarily focus on access tokens and time-sequence logs for control, making it difficult to establish a continuous, stateful relationship between task requests, authorization triggers, decision-making access, evidence sources, verification feedback, and task termination. When an agent makes unauthorized calls, provides inconsistent evidence sources, disagrees with the verification feedback and decision output, or retains permissions after the task ends, existing solutions typically only allow for post-event investigations from scattered logs. This lack of a comprehensive accountability chain that spans the entire process of authorization, decision-making, verification, and termination results in unclear boundaries of responsibility in the agent's decision-making process, and the evidence storage report exhibits low completeness in expressing responsibility assumption and permission revocation status. Summary of the Invention

[0005] This application provides a method and system for managing the decision-making authority and responsibility of intelligent agents based on declarative primitive state machines, so as to at least alleviate the above-mentioned technical problems.

[0006] A method for managing the decision-making responsibilities of intelligent agents based on declarative primitive state machines includes: Step 1: Perform declarative primitive parsing on the agent's decision task request to generate authorization primitive instruction units, and generate authorization state responsibility records based on the authorization primitive instruction units; Step 2: Perform judgment access element parsing on the authorized state rights and responsibilities record to obtain the judgment access element description, and perform access matching between the judgment primitive instruction unit and the judgment access element description to generate evidence binding judgment record; Step 3: Reconstruct the verification elements of the evidence binding judgment record to obtain the verification comparison element description. Based on the verification comparison element description and the determined verification primitive instruction unit to be verified, generate the verification status responsibility record. Step 4: Reconstruct the final elements of the verification state rights and responsibilities record to generate a final constraint element description, and generate an agent decision-making rights and responsibilities management and evidence storage report based on the final constraint element description.

[0007] Optionally, in step 1, an authorization state responsibility record is generated based on the authorization primitive instruction unit, which includes the following steps: The authorization primitive instruction unit is subjected to authorization trigger verification to obtain the authorization trigger verification result; Based on the authorization trigger verification result, the authorization elements of the authorization primitive instruction unit are extracted to obtain the authorization element description; The authorization element descriptions are encapsulated into authorization relationships to form a dynamic authorization mapping; Based on the authorization trigger verification result, the flow elements of the authorized primitive instruction unit are extracted to obtain the primitive flow order; Based on the primitive flow order and the dynamic authorization mapping, an authorization state responsibility record is generated.

[0008] Optionally, based on the primitive flow order and the dynamic authorization mapping, an authorization state responsibility record is generated, including: The dynamic authorization mapping is registered with a status entry according to the primitive flow order to form a primitive flow credential; The dynamic authorization mapping, the primitive flow credential, and the authorization trigger verification result are processed by signature digest to generate a signature identity digest; The authorization primitive instruction unit, the authorization trigger verification result, the dynamic authorization mapping, the primitive flow credential, and the signature identity digest are written into the declarative primitive state machine to generate an authorization state responsibility record.

[0009] Optionally, step 2: Parse the authorized state rights and responsibilities record for judgment access elements to obtain a judgment access element description; perform access matching between the judgment primitive instruction unit to be verified and the judgment access element description to generate an evidence binding judgment record, including: The decision primitive instruction unit is received based on the authorized state rights and responsibilities record, and is used as the decision primitive instruction unit to be verified; The primitive instruction unit to be verified is matched with the description of the judgment access element to obtain the judgment access verification result. The description of the judgment admission elements is parsed to obtain the judgment flow constraints. The judgment flow verification is performed on the judgment primitive instruction unit to be verified according to the judgment flow constraints to generate a judgment flow verification result; In response to the judgment access verification result and the judgment transfer verification result, the evidence elements of the judgment primitive instruction unit to be verified are extracted to determine the judgment evidence description; Based on the description of the evidence in the judgment, an evidence-binding judgment record is generated.

[0010] Optionally, based on the description of the judgment evidence, an evidence-binding judgment record is generated, including: The original evidence data is read according to the description of the judgment evidence to obtain the evidence data to be compared; The evidence description in the judgment is compared with the evidence data to be compared, and a source verification is performed to obtain the evidence source verification result. The judgment result is extracted from the primitive instruction unit to be verified to obtain the judgment result; The judgment access verification result, the judgment flow verification result, the judgment evidence description, the evidence source verification result, and the judgment result are subjected to evidence binding processing to generate an evidence binding judgment record.

[0011] Optionally, in step 3, a verification status responsibility record is generated based on the description of the verification comparison elements and the determined verification primitive instruction units to be verified, including: The verification entry verification is performed on the verification primitive instruction unit to be verified according to the verification comparison element description, so as to obtain the verification entry verification result; The verification conclusion is extracted from the verification primitive instruction unit to be verified in order to obtain the verification conclusion; The consistency between the verification conclusion and the description of the verification comparison elements is verified to generate a verification status responsibility record.

[0012] Optionally, the consistency between the verification conclusion and the description of the verification comparison elements is verified to generate a verification status responsibility record, including: The consistency between the verification conclusion and the description of the verification comparison element is verified to generate a verification status determination result; Based on the verification results of the verification entry and the verification status determination results, the audit status of the evidence binding judgment record is rewritten to generate an audit judgment record; Based on the verification status determination result, the audit judgment record is configured to assume responsibility, so as to generate a responsibility assumption identifier; The responsibility assignment identifier is written into the audit decision record to obtain the verification status responsibility record.

[0013] Optionally, step 4: Based on the description of the termination constraint elements, generate an agent decision-making responsibility control and evidence storage report, including: Based on the verification state responsibility record, the terminal primitive instruction unit is received to obtain the terminal primitive instruction unit to be verified; According to the description of the termination constraint elements, the termination entry verification is performed on the termination primitive instruction unit to be verified to obtain the termination entry verification result; Based on the termination entry verification result, the session boundary is located for the termination constraint element description to determine the authorization record to be terminated; The pending authorization records are marked as invalid to form authorization invalidation records; Based on the authorization expiration records, generate a report on the management and evidence preservation of the decision-making authority and responsibility of the intelligent agent.

[0014] Optionally, based on the authorization expiration records, a report on the management and evidence preservation of agent decision-making authority and responsibility is generated, including: Based on the authorization expiration record, the authorization record to be terminated is erased to obtain the authorization erasure record; Write the authorized erasure record into the verification state responsibility record to obtain the final state responsibility record; The terminal state rights and responsibilities records are stored and encapsulated to generate an agent decision-making rights and responsibilities management evidence report.

[0015] A decision-making responsibility management system for intelligent agents based on declarative primitive state machines includes an authorized state responsibility record generation unit, an evidence binding judgment record generation unit, a verification state responsibility record generation unit, and a responsibility management evidence storage report generation unit configured on a processor. The authorized state responsibility record generation unit is used to execute step 1 of the above method; the evidence binding judgment record generation unit is used to execute step 2 of the above method; the verification state responsibility record generation unit is used to execute step 3 of the above method; and the responsibility management evidence storage report generation unit is used to execute step 4 of the above method.

[0016] Technical advantages of the technical solution provided in this application This application presents a method for controlling the decision-making responsibilities of intelligent agents based on declarative primitive state machines. Addressing the technical deficiency in existing intelligent agent operation control schemes where there is a lack of continuous state-based association between task requests, authorization triggers, decision access, evidence sources, verification feedback, and task termination, this method performs declarative primitive parsing on the intelligent agent's decision-making task requests to generate authorization primitive instruction units. Based on these instruction units, an authorization state responsibility record is generated, ensuring that the intelligent agent's decision-making task forms a state-based authorization phase record before proceeding to subsequent decision processing. Compared to traditional methods relying solely on ordinary access tokens, this approach establishes a clear association between task requests and authorization phase records, reducing the risk of subsequent decision primitive instruction units being processed outside of authorization constraints.

[0017] Furthermore, this application performs judgment access element analysis on the authorization state rights and responsibilities record to obtain a judgment access element description, and performs access matching between the judgment primitive instruction unit and the judgment access element description to generate an evidence-bound judgment record. Therefore, the judgment processing does not directly rely on scattered logs or single interface call results, but rather forms a judgment access element description based on the authorization state rights and responsibilities record, and incorporates the judgment primitive instruction unit into the evidence-bound judgment record through access matching. Compared to the traditional approach that only records decision output and call time, this approach enables a clearer technical continuity between the judgment stage and the authorization stage, and provides the judgment processing result with a more complete evidentiary basis.

[0018] Furthermore, this application reconstructs the verification elements of the evidence-bound judgment record to obtain a description of the verification comparison elements, and generates a verification status responsibility record based on the description of the verification comparison elements and the determined verification primitive instruction units to be verified. Through this process, the verification stage is not a post-event record independent of the judgment result, but rather a reconstruction of the description of the verification comparison elements from the evidence-bound judgment record, and then processing of the verification primitive instruction units based on this description of the verification comparison elements. Compared with the traditional approach where the correlation between verification feedback and the original judgment record is weak, this approach enables a clearer closed-loop record between the verification conclusion, the judgment content, and the evidence correlation, thereby alleviating the problem of unclear responsibility assignment.

[0019] Furthermore, this application reconstructs the verification state rights and responsibilities record using termination elements to generate a termination constraint element description, and generates an agent decision-making rights and responsibilities management and evidence storage report based on the termination constraint element description. Through this process, the task termination phase can inherit the verification state rights and responsibilities record, instead of only handling permission invalidation based on the session end state; the termination constraint element description participates in generating the agent decision-making rights and responsibilities management and evidence storage report, forming a continuous management and control link between authorization, decision, verification, and termination. Compared to the traditional approach of separating the permission revoke state from the preceding decision and verification records, this approach improves the completeness of the evidence storage report's expression of rights and responsibilities boundaries, evidence association, and permission termination state. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating an intelligent agent decision-making responsibility management method based on a declarative primitive state machine, according to an embodiment of this application.

[0021] Figure 2 This is a schematic diagram of the structure of an intelligent agent decision-making responsibility management system based on a declarative primitive state machine, according to an embodiment of this application.

[0022] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application.

[0023] Figure 4 This application provides a schematic diagram of an intelligent agent decision-making responsibility management system based on a declarative primitive state machine. Detailed Implementation

[0024] like Figure 1 As shown in the figure, this application provides an embodiment of a method for managing the decision-making responsibilities of an intelligent agent based on a declarative primitive state machine, which includes: Step 1: Perform declarative primitive parsing on the agent's decision task request to generate authorization primitive instruction units, and generate authorization state responsibility records based on the authorization primitive instruction units; Step 2: Perform judgment access element parsing on the authorized state rights and responsibilities record to obtain the judgment access element description, and perform access matching between the judgment primitive instruction unit and the judgment access element description to generate evidence binding judgment record; Step 3: Reconstruct the verification elements of the evidence binding judgment record to obtain the verification comparison element description. Based on the verification comparison element description and the determined verification primitive instruction unit to be verified, generate the verification status responsibility record. Step 4: Reconstruct the final elements of the verification state rights and responsibilities record to generate a final constraint element description, and generate an agent decision-making rights and responsibilities management and evidence storage report based on the final constraint element description.

[0025] Optionally, in step 1, an authorization state responsibility record is generated based on the authorization primitive instruction unit, which includes the following steps: The authorization primitive instruction unit is subjected to authorization trigger verification to obtain the authorization trigger verification result; Based on the authorization trigger verification result, the authorization elements of the authorization primitive instruction unit are extracted to obtain the authorization element description; The authorization element descriptions are encapsulated into authorization relationships to form a dynamic authorization mapping; Based on the authorization trigger verification result, the flow elements of the authorized primitive instruction unit are extracted to obtain the primitive flow order; Based on the primitive flow order and the dynamic authorization mapping, an authorization state responsibility record is generated.

[0026] Preferably, in the specific technical implementation of step 1, when performing declarative primitive parsing processing on the agent decision task request, the agent decision task request is first written into the request receiving buffer, and the request receiving buffer is then segmented according to the message header, primitive declaration field, task payload field, and permission payload field to form a request field offset table. The request field offset table is used to record the storage location of the message header, primitive declaration field, task payload field, and permission payload field in the request receiving buffer. The task payload field includes the task initiation field, task receiving field, and task scope field, and the permission payload field includes the permission invocation field. Subsequently, the primitive declaration field is read according to the request field offset table, and the primitive declaration field is written into the primitive scheduling queue for primitive type identification to separate the authorized primitive instruction unit. After the authorization primitive instruction unit is formed, it carries the field positions corresponding to the primitive declaration field, task payload field and permission payload field in the request field offset table. This enables subsequent authorization trigger verification to directly read the task initiation field, task reception field, task scope field and permission invocation field corresponding to the authorization primitive instruction unit from the request reception buffer based on the request field offset table, without having to rescan the complete agent decision task request.

[0027] Preferably, in step 1, when performing authorization trigger verification on the authorization primitive instruction unit, the primitive header marker is first read from the primitive declaration field in the authorization primitive instruction unit, and the primitive header marker is matched with the pre-set authorization entry marker in the declarative primitive state machine to obtain the primitive entry matching result; then, according to the request field offset table carried by the authorization primitive instruction unit, the task initiation field, task reception field, and task scope field are read from the task payload field in the request reception buffer, and the permission invocation field is read from the permission payload field in the request reception buffer. The field integrity verification is performed on the task initiation field, task reception field, task scope field, and permission invocation field to obtain the field integrity verification result. The primitive entry matching result is used to characterize whether the primitive header tag matches the authorization entry tag, and the field integrity verification result is used to characterize whether the task initiation field, task reception field, task scope field, and permission invocation field can constitute the authorization processing basis of the authorization primitive instruction unit. The primitive entry matching result and the field integrity verification result are combined into the authorization trigger verification result. The authorization trigger verification result continues to participate in the authorization element extraction, so that subsequent authorization element extraction is only performed under the condition that the authorization primitive instruction unit satisfies the primitive entry matching and field integrity verification.

[0028] Preferably, in the process of extracting authorized elements from the authorization trigger verification result, the following steps are taken: First, based on the authorization trigger verification result and the request field offset table, the authorized subject identifier is read from the task initiation field of the authorization primitive instruction unit; then, based on the authorization trigger verification result and the request field offset table, the entrusted subject identifier is read from the task receiving field of the authorization primitive instruction unit; subsequently, based on the authorization trigger verification result and the request field offset table, the task boundary description is read from the task scope field of the authorization primitive instruction unit; and finally, based on the authorization trigger verification result and the request field offset table, the computing power permission description is read from the permission invocation field of the authorization primitive instruction unit. The authorized subject identifier, entrusted subject identifier, task boundary description, and computing power permission description are written into the same authorization element buffer, forming an authorization element description in the authorization element buffer; the authorization element description continues to enter the authorization relationship encapsulation, so that the authorized subject identifier, entrusted subject identifier, task boundary description, and computing power permission description do not participate in subsequent processing as scattered fields, but participate in the formation of dynamic authorization mapping in the form of authorization element descriptions.

[0029] Preferably, when encapsulating the authorization relationship for the authorization element description, firstly, an authorization mapping slot corresponding to the authorization element description is opened in the pre-configured memory mapping table, and the authorization subject identifier in the authorization element description is written to the authorization source position of the authorization mapping slot; then, the entrusted subject identifier in the authorization element description is written to the authorization acceptance position of the authorization mapping slot; subsequently, the task boundary description in the authorization element description is written to the task boundary position of the authorization mapping slot, and the computing power permission description in the authorization element description is converted into a permission bitmap and written to the permission payload position of the authorization mapping slot to form a dynamic authorization mapping. The permission bitmap is used to express the permission enabling range in the computing power permission description in a bit state that can be read by the machine, and the dynamic authorization mapping binds the authorization subject identifier, entrusted subject identifier, task boundary description and permission bitmap to the same authorization mapping slot; the memory mapping table stores the dynamic authorization mapping through the authorization mapping slot, so that the dynamic authorization mapping can be located and read by the declarative primitive state machine in the subsequent state entry registration; after the dynamic authorization mapping is formed, the dynamic authorization mapping continues to participate in the state entry registration, so that the dynamic authorization mapping can be recognized by the declarative primitive state machine as a state object in the authorization stage.

[0030] Preferably, when extracting the flow elements of the authorized primitive instruction unit based on the authorization trigger verification result, the primitive arrangement field is first read from the primitive declaration field in the authorized primitive instruction unit according to the authorization trigger verification result and the request field offset table, and then written into the primitive scheduling queue. The primitive scheduling queue performs state sequence parsing on the primitive arrangement field according to the state order of authorization, decision, verification, and termination to obtain the primitive flow order. The primitive flow order is used to characterize the state transition path after the authorized primitive instruction unit enters the declarative primitive state machine. The primitive flow order then participates in the state entry registration together with the dynamic authorization mapping, so that the dynamic authorization mapping not only includes the authorization relationship, but also the flow position of the authorization relationship in authorization, decision, verification, and termination. After forming the primitive flow order, the primitive scheduling queue continues to retain the correspondence between the primitive arrangement field and the primitive flow order, so that the primitive flow credentials generated subsequently can be traced back to the primitive arrangement field in the authorized primitive instruction unit.

[0031] Preferably, when generating the authorization state responsibility record based on the primitive flow order and dynamic authorization mapping, the authorization entry position is first located in the state table of the declarative primitive state machine, and the dynamic authorization mapping is written into the state table entry corresponding to the authorization entry position. Then, according to the primitive flow order, the flow relationship between the dynamic authorization mapping and the decision entry, verification entry, and termination entry is registered in the state table entry to form a primitive flow credential. The primitive flow credential records the authorization entry position of the dynamic authorization mapping in the declarative primitive state machine, as well as the state positions that the dynamic authorization mapping can subsequently migrate to the decision entry, verification entry, and termination entry. After the primitive flow credential is formed, it continues to participate in signature digest processing, so that the signature digest processing can simultaneously cover the authorization relationship in the dynamic authorization mapping and the state transition relationship in the primitive flow credential.

[0032] Preferably, when performing signature digest processing on the dynamic authorization mapping, primitive flow credentials, and authorization trigger verification results, the dynamic authorization mapping is first written to the authorization digest cache area, and the primitive flow credentials are written to the flow digest cache area; then the authorization trigger verification results are written to the trigger digest cache area; subsequently, digest calculation is performed according to the cache order of the authorization digest cache area, the flow digest cache area, and the trigger digest cache area to generate a signature identity digest. The authorization digest cache is used to cache the authorizing entity identifier, entrusted entity identifier, task boundary description, and permission bitmap in the dynamic authorization mapping. The flow digest cache is used to cache the authorization entry position, decision entry, verification entry, and termination entry in the primitive flow credentials. The trigger digest cache is used to cache the primitive entry matching result and field integrity verification result in the authorization trigger verification result. The signature identity digest is not an identity record generated simply for the authorizing entity identifier or entrusted entity identifier, but a digest result jointly generated by the dynamic authorization mapping, primitive flow credentials, and authorization trigger verification result. The signature identity digest continues to participate in the writing of the authorization state rights and responsibilities record, so that the authorization state rights and responsibilities record can reflect the authorization relationship, flow order, and trigger verification status simultaneously when it is read later.

[0033] Preferably, when writing the authorization primitive instruction unit, authorization trigger verification result, dynamic authorization mapping, primitive flow credential, and signature identity digest into the declarative primitive state machine, the authorization state record position is first generated in the state table of the declarative primitive state machine, and the authorization primitive instruction unit is written into the primitive storage area of ​​the authorization state record position; then the authorization trigger verification result is written into the trigger verification area of ​​the authorization state record position; subsequently, the dynamic authorization mapping is written into the authorization mapping area of ​​the authorization state record position, and the primitive flow credential is written into the flow registration area of ​​the authorization state record position; finally, the signature identity digest is written into the digest verification area of ​​the authorization state record position to generate the authorization state responsibility record. The primitive storage area, trigger verification area, authorization mapping area, circulation registration area, and digest verification area in the authorization state rights and responsibilities record all correspond to the same authorization state record location. The primitive storage area stores the authorization primitive instruction unit, the trigger verification area stores the authorization trigger verification result, the authorization mapping area stores the dynamic authorization mapping, the circulation registration area stores the primitive circulation voucher, and the digest verification area stores the signature identity digest, so that the authorization state rights and responsibilities record can be used as a unified reading object for subsequent judgment access element parsing.

[0034] Preferably, in a scenario where multiple agents collaboratively invoke tools, the agent's decision-making task request may carry the authorized entity identifier of the task scheduling node, the entrusted entity identifier of the execution node, the task boundary description corresponding to the tool invocation scope, and the computing power permission description corresponding to the computing power invocation scope. The task scheduling node enters the agent's decision-making task request through the task initiation field, the execution node enters the agent's decision-making task request through the task receiving field, the tool invocation scope enters the agent's decision-making task request through the task scope field, and the computing power invocation scope enters the agent's decision-making task request through the permission invocation field. After the authorization primitive instruction unit is separated from the agent's decision-making task request, the authorization trigger verification result limits whether the authorization primitive instruction unit can enter the authorization element extraction. The authorization element description organizes the authorization subject identifier, entrusted subject identifier, task boundary description, and computing power permission description into encapsulated data content. The dynamic authorization mapping converts the authorization element description into a registrable memory mapping relationship. The primitive flow sequence incorporates the dynamic authorization mapping into the state transition path of authorization, judgment, verification, and termination. The authorization state responsibility record writes the authorization primitive instruction unit, authorization trigger verification result, dynamic authorization mapping, primitive flow credential, and signature identity digest into the declarative primitive state machine, so that when performing judgment access element parsing on the authorization state responsibility record, the access basis corresponding to the judgment primitive instruction unit can be read from the authorization state responsibility record.

[0035] Optionally, based on the primitive flow order and the dynamic authorization mapping, an authorization state responsibility record is generated, including: The dynamic authorization mapping is registered with a status entry according to the primitive flow order to form a primitive flow credential; The dynamic authorization mapping, the primitive flow credential, and the authorization trigger verification result are processed by signature digest to generate a signature identity digest; The authorization primitive instruction unit, the authorization trigger verification result, the dynamic authorization mapping, the primitive flow credential, and the signature identity digest are written into the declarative primitive state machine to generate an authorization state responsibility record.

[0036] Preferably, in step 1, when generating the authorization status and responsibility record based on the primitive flow order and dynamic authorization mapping, the authorization subject identifier, entrusted subject identifier, task boundary description, and permission bitmap are first read from the authorization mapping slot where the dynamic authorization mapping is located. These identifiers are then written into the status registration buffer to form an authorization registration data block. Next, the primitive flow order is written into the same status registration buffer, ensuring a sequential association between the primitive flow order and the authorization registration data block within the same cache address range. The status registration buffer is used to receive the dynamic authorization mapping and the primitive flow order. The authorization registration data block expresses the authorization relationship content in the dynamic authorization mapping, and the primitive flow order expresses the state transition path of the authorization registration data block into authorization, judgment, verification, and termination. The cache address range limits the continuous storage location of the authorization registration data block and the primitive flow order in the status registration buffer, enabling subsequent status entry registrations to continuously read the authorization registration data block and the primitive flow order according to the cache address range. Therefore, subsequent state entry registration no longer registers the dynamic authorization mapping and primitive flow order separately, but registers based on the authorization registration data block and primitive flow order already associated in the state registration buffer, and maintains the underlying storage association between the dynamic authorization mapping and primitive flow order through the state registration buffer.

[0037] Preferably, in step 1, when registering the state entry of the dynamic authorization mapping according to the primitive flow order, the authorization entry position is first located in the state table of the declarative primitive state machine, and a state table entry is allocated to the authorization registration data block according to the authorization entry position; then, the dynamic authorization mapping is written into the authorization mapping area of ​​the state table entry, and the primitive flow order is written into the flow order area of ​​the state table entry to form a state entry index. The state table is used to store each state table entry in the declarative primitive state machine. The state table entry is used to carry the dynamic authorization mapping and the primitive flow order. The authorization mapping area is used to store the dynamic authorization mapping, the flow order area is used to store the primitive flow order, and the state entry index is used to identify the authorization entry position of the dynamic authorization mapping in the state table entry, and record the state transition path that the dynamic authorization mapping can sequentially migrate to the decision entry, verification entry, and termination entry. After the state entry index is formed, it continues to participate in the generation of primitive flow credentials, so that the primitive flow credentials can simultaneously reflect the storage location of the dynamic authorization mapping and the state transition path of the dynamic authorization mapping.

[0038] Preferably, in step 1, when forming the primitive flow credential, the state entry index is processed by credential writing. Specifically, the authorization entry position, decision entry, verification entry, and termination entry in the state entry index are read, and these positions are written into the flow registration area according to the primitive flow order to form the primitive flow credential. The flow registration area is used to store the readable credential content transformed from the state entry index. The primitive flow credential not only records the primitive flow order but also records the authorization entry position, decision entry, verification entry, and termination entry of the dynamic authorization mapping in the state table entry. Therefore, the primitive flow credential can include the dynamic authorization mapping and state transition path in the digest scope during subsequent signature digest processing. Through the primitive flow credential, the dynamic authorization mapping is no longer just a dynamic mapping relationship in memory but is registered by the declarative primitive state machine as a stateful record that can be read by subsequent decision, verification, and termination. This stateful record continues to participate in the signature digest processing, enabling the dynamic authorization mapping, primitive flow credential, and authorization trigger verification result to be jointly bound to the signature identity digest.

[0039] Preferably, after the primitive flow credential is formed, when performing signature digest processing on the dynamic authorization mapping, primitive flow credential, and authorization trigger verification result, the dynamic authorization mapping is first written to the authorization digest cache, and authorization digest data blocks are formed in the authorization digest cache according to the order of the authorizing entity identifier, the entrusted entity identifier, the task boundary description, and the permission bitmap; then, the primitive flow credential is written to the flow digest cache, and flow digest data blocks are formed in the flow digest cache according to the state order of the authorization entry position, the decision entry, the verification entry, and the termination entry; subsequently, the authorization trigger verification result is written to the trigger digest cache, and trigger digest data blocks are formed in the trigger digest cache. The authorization digest cache is used to cache the dynamic authorization mapping, and the authorization digest data blocks originate from the dynamic authorization mapping in the authorization digest cache; the flow digest cache is used to cache the primitive flow credential, and the flow digest data blocks originate from the primitive flow credential in the flow digest cache; the trigger digest cache is used to cache the authorization trigger verification result, and the trigger digest data blocks originate from the authorization trigger verification result in the trigger digest cache. The authorization digest data block, the flow digest data block, and the trigger digest data block are derived from the dynamic authorization mapping, the primitive flow credential, and the authorization trigger verification result, respectively. Subsequent digest calculations are performed based on the authorization digest data block, the flow digest data block, and the trigger digest data block, so that the signature identity digest can cover the authorization relationship content in the dynamic authorization mapping, the status flow content in the primitive flow credential, and the authorization trigger verification content in the authorization trigger verification result.

[0040] Preferably, in step 1, when generating the signature identity digest, the authorization digest data block, the circulation digest data block, and the trigger digest data block are sequentially written into the digest concatenation buffer, and digest calculation is performed on the data in the digest concatenation buffer to generate a digest value to be signed; then, the signature key handle associated with the authorization primitive instruction unit is called to sign the digest value to be signed to generate the signature identity digest. The digest concatenation buffer is used to form the input content to be signed according to the writing order of the authorization digest data block, the circulation digest data block, and the trigger digest data block. The digest value to be signed is formed by digest calculation of the input content to be signed in the digest concatenation buffer; the signature key handle is triggered by the authorization subject identifier in the authorization primitive instruction unit. The signature key handle does not directly replace the authorization subject identifier, but is used to sign the digest value to be signed. After the signature identity digest is generated, the signature identity digest continues to participate in the writing of the authorization state rights and responsibilities record, so that the authorization state rights and responsibilities record can be verified by the signature identity digest to check whether the dynamic authorization mapping, primitive circulation credentials, and authorization trigger verification results maintain the same source association when subsequently read.

[0041] Preferably, when writing the authorization primitive instruction unit, authorization trigger verification result, dynamic authorization mapping, primitive transfer credential, and signature identity digest into the declarative primitive state machine, firstly, an authorization state record position is allocated for the authorization state responsibility record in the state table, and the authorization state record position is address-bound to the aforementioned state table entry; then, the authorization primitive instruction unit is written into the primitive storage area of ​​the authorization state record position, the authorization trigger verification result is written into the trigger verification area of ​​the authorization state record position, the dynamic authorization mapping is written into the authorization mapping area of ​​the authorization state record position, the primitive transfer credential is written into the transfer registration area of ​​the authorization state record position, and the signature identity digest is written into the digest verification area of ​​the authorization state record position. The authorization state record location is used to carry the authorization state rights and responsibilities record; the primitive storage area is used to store the authorization primitive instruction unit; the trigger verification area is used to store the authorization trigger verification result; the authorization mapping area is used to store the dynamic authorization mapping; the circulation registration area is used to store the primitive circulation credential; and the digest verification area is used to store the signature identity digest. The authorization state record location, primitive storage area, trigger verification area, authorization mapping area, circulation registration area, and digest verification area together constitute the underlying storage structure of the authorization state rights and responsibilities record, enabling the authorization state rights and responsibilities record to enter the subsequent judgment access element parsing as a unified read object.

[0042] Preferably, to avoid incomplete status table entries during the writing process of the authorized state rights and responsibilities record, segmented writing and write status marking are performed on the authorized state record position when generating the authorized state rights and responsibilities record. Segmented writing first writes the authorized primitive instruction unit and the authorized trigger verification result, then writes the dynamic authorized mapping and primitive flow credential, and finally writes the signature identity digest. After each segment is written, the write status mark corresponding to that segment is updated in the authorized state record position, and after the signature identity digest is written, the write status marks are combined into an authorized state readable mark. The write status mark is used to record the writing completion status of the authorized primitive instruction unit, the authorized trigger verification result, the dynamic authorized mapping, the primitive flow credential, and the signature identity digest in the authorized state record position. The authorized state readable mark is formed by combining the write status marks and is used to control whether the authorized state rights and responsibilities record can be parsed and read by subsequent decision access elements. After the authorized state readable marker is written to the authorized state record location, the authorized state rights and responsibilities record is allowed to be read as the reading object for subsequent decision access element parsing. As a result, the subsequently read authorized state rights and responsibilities record contains the authorized primitive instruction unit, the authorized trigger verification result, the dynamic authorization mapping, the primitive flow credential, and the signature identity digest.

[0043] Preferably, in a scenario where multiple agents collaboratively invoke tools, the dynamic authorization mapping corresponds to the mapping relationship where the task scheduling node temporarily grants tool invocation permissions to the execution node. The primitive flow order corresponds to the state transition path of the dynamic authorization mapping, which sequentially passes through authorization, judgment, verification, and termination. The task scheduling node corresponds to the authorizing entity identifier in the dynamic authorization mapping, the execution node corresponds to the entrusted entity identifier in the dynamic authorization mapping, the tool invocation scope corresponds to the task boundary description in the dynamic authorization mapping, and the tool invocation permission corresponds to the permission bitmap in the dynamic authorization mapping. After registering the state entry of the dynamic authorization mapping according to the primitive flow order, the resulting primitive flow credential can write the task scheduling node, execution node, tool invocation scope, tool invocation permission, and state transition path into the authorization state responsibility record. Then, the dynamic authorization mapping, primitive flow credential, and authorization trigger verification result are bound to the same authorization state responsibility record through the signature identity digest. The authorized state rights and responsibilities record generated in this way is not a simple access token record, but a stateful record that includes dynamic authorization mapping, state transition path, authorization trigger verification result and signature identity digest. When performing judgment access element analysis on the authorized state rights and responsibilities record in the future, the access basis corresponding to the judgment primitive instruction unit can be read from the authorized state rights and responsibilities record.

[0044] Optionally, step 2: Parse the authorized state rights and responsibilities record for judgment access elements to obtain a judgment access element description; perform access matching between the judgment primitive instruction unit to be verified and the judgment access element description to generate an evidence binding judgment record, including: The decision primitive instruction unit is received based on the authorized state rights and responsibilities record, and is used as the decision primitive instruction unit to be verified; The primitive instruction unit to be verified is matched with the description of the judgment access element to obtain the judgment access verification result. The description of the judgment admission elements is parsed to obtain the judgment flow constraints. The judgment flow verification is performed on the judgment primitive instruction unit to be verified according to the judgment flow constraints to generate a judgment flow verification result; In response to the judgment access verification result and the judgment transfer verification result, the evidence elements of the judgment primitive instruction unit to be verified are extracted to determine the judgment evidence description; Based on the description of the evidence in the judgment, an evidence-binding judgment record is generated.

[0045] Preferably, the specific implementation process of step 2 is as follows: When receiving the judgment primitive instruction unit based on the authorized state responsibility record, the authorized state readable marker is first read from the authorized state record position corresponding to the authorized state responsibility record. The authorized state readable marker comes from the write status markers for the authorized primitive instruction unit, authorization trigger verification result, dynamic authorization mapping, primitive flow credential, and signature identity digest in the authorized state record position. After the authorized state readable marker indicates that the authorized state responsibility record has been written, the judgment primitive instruction unit is written to the judgment primitive receiving buffer. The judgment primitive receiving buffer caches the judgment primitive instruction unit to form a judgment primitive instruction unit to be verified. The judgment primitive instruction unit to be verified carries the associated read address corresponding to the authorized state record position. The associated read address comes from the storage address of the authorized state record position in the state table of the declarative primitive state machine, and the associated read address continues to participate in the subsequent judgment admission element parsing, so that the judgment primitive instruction unit to be verified can read the authorized state responsibility record through the associated read address, rather than relying solely on the time order to associate with the authorized state responsibility record.

[0046] Preferably, in the specific technical implementation of step 2, when parsing the decision access elements of the authorization state rights and responsibilities record, the location of the authorization state record is first located based on the associated read address carried by the decision primitive instruction unit to be verified. Then, the dynamic authorization mapping is read from the authorization mapping area in the authorization state record location, and the primitive flow credential is read from the flow registration area in the authorization state record location. The dynamic authorization mapping is used to provide the authorization subject identifier, entrusted subject identifier, task boundary description, and permission bitmap required for decision access. The primitive flow credential is used to provide the state transition path required for the decision primitive instruction unit to enter the decision stage. Subsequently, the dynamic authorization mapping and primitive flow credential are written into the decision access parsing cache area to form the decision access element description. The decision admission parsing cache is used to temporarily store the dynamic authorization mapping and primitive flow credentials parsed from the authorization state rights and responsibilities record. The decision admission element description is formed by the dynamic authorization mapping and primitive flow credentials in the decision admission parsing cache and continues to participate in the admission matching of the decision primitive instruction unit to be verified, making the decision admission element description an intermediate technical object connecting the authorization state rights and responsibilities record and the decision primitive instruction unit to be verified.

[0047] Preferably, when performing access matching between the judgment primitive instruction unit to be verified and the judgment access element description, the judgment initiation identifier, judgment task boundary, and judgment permission request are first read from the judgment primitive instruction unit to be verified, and then the entrusted entity identifier, task boundary description, and permission bitmap are read from the judgment access element description. The judgment initiation identifier is used to perform subject matching with the entrusted entity identifier to form a subject matching result; the judgment task boundary is used to perform boundary matching with the task boundary description to form a boundary matching result; and the judgment permission request is used to perform permission matching with the permission bitmap to form a permission matching result. The subject matching result, boundary matching result, and permission matching result are written to the same judgment access verification cache to obtain the judgment access verification result. The judgment access verification cache is used to store the subject matching result, boundary matching result, and permission matching result. The judgment access verification result continues to participate in the triggering determination of subsequent evidence element extraction, so that the judgment primitive instruction unit to be verified only enters the subsequent evidence element extraction processing when the judgment initiation identifier corresponds to the entrusted entity identifier, the judgment task boundary corresponds to the task boundary description, and the judgment permission request corresponds to the permission bitmap.

[0048] Preferably, when parsing the flow constraint of the decision admission element description, the primitive flow credentials in the decision admission element description are read, and the state transition path between the authorization entry position, the decision entry, the verification entry, and the termination entry is extracted from the primitive flow credentials. The state transition path is written into the decision flow constraint cache to form a decision flow constraint; the decision flow constraint cache is used to store the state transition path parsed from the primitive flow credentials, and the decision flow constraint is used to characterize whether the decision primitive instruction unit to be verified can enter the decision entry from the authorization entry position, and to further restrict the write position of the decision primitive instruction unit to be verified in the decision stage of the declarative primitive state machine. The decision flow constraint originates from the primitive flow credentials in the decision admission element description and continues to participate in the decision flow verification, enabling the decision flow constraint to pass the state transition path generated in the authorization stage to the entry verification processing in the decision stage.

[0049] Preferably, when performing judgment flow verification on the judgment primitive instruction unit to be verified according to the judgment flow constraints, the judgment entry point is first read from the judgment flow constraints, and then the judgment entry point declaration is read from the judgment primitive instruction unit to be verified. The judgment entry point declaration and the judgment entry point are then checked for entry point consistency to generate a judgment entry point verification result. Subsequently, the order relationship between the authorized entry point position and the judgment entry point is read from the judgment flow constraints, and the primitive arrival order is read from the judgment primitive instruction unit to be verified. The order relationship and the primitive arrival order are checked for order consistency to generate a judgment order verification result. The primitive arrival order comes from the cache registration order formed when the judgment primitive instruction unit to be verified enters the judgment primitive receiving buffer. After the primitive arrival order participates in the order consistency verification, it, together with the order relationship, affects the judgment order verification result. The judgment entry point verification result and the judgment order verification result are written into the judgment flow verification buffer to generate the judgment flow verification result. The results of the judgment transfer verification continue to participate in the triggering determination of subsequent evidence element extraction together with the results of the judgment access verification, so that the primitive instruction unit of the judgment to be verified satisfies both the authorization access relationship and the judgment entry relationship registered by the declarative primitive state machine.

[0050] Preferably, in response to the judgment access verification result and the judgment flow verification result, when extracting evidence elements from the judgment primitive instruction unit to be verified, the evidence hash field is first read from the judgment primitive instruction unit to be verified, and the length format of the evidence hash field is parsed to obtain the length format verification result. Then, the summary format of the evidence hash field is parsed to obtain the summary format verification result. Based on the length format verification result and the summary format verification result, the evidence hash is extracted from the evidence hash field. Next, the evidence index field is read from the judgment primitive instruction unit to be verified, and the storage address of the evidence index field is parsed to obtain the original evidence index. Subsequently, the judgment payload field is read from the judgment primitive instruction unit to be verified, and the judgment content of the judgment payload field is parsed to obtain the judgment result. The evidence hash field, as the source of the evidence hash, continues to participate in the evidence hash extraction; the evidence index field, as the source of the original evidence index, continues to participate in the original evidence index extraction; and the judgment payload field, as the source of the judgment result, continues to participate in the judgment result extraction. The evidence hash, the original evidence index, and the judgment result are written into the judgment evidence cache area, and a judgment evidence description is formed in the judgment evidence cache area. The description of evidence in the judgment continues to participate in the generation of the evidence binding judgment record, so that the evidence hash, the original evidence index, and the judgment result can be processed as evidence-related content of the same judgment primitive instruction unit.

[0051] Preferably, when generating an evidence-bound judgment record based on the judgment evidence description, the evidence storage location of the original evidence data is first located according to the original evidence index in the judgment evidence description, and the original evidence data is read from the evidence storage location. The evidence storage location originates from the storage address resolution result corresponding to the original evidence index, and the evidence storage location continues to participate in the reading of the original evidence data, ensuring that the original evidence data corresponds to the original evidence index in the judgment evidence description. Subsequently, a digest calculation is performed on the original evidence data to form an evidence recalculation digest, and the evidence recalculation digest is compared with the evidence hash in the judgment evidence description for a source verification result. The evidence recalculation digest originates from the original evidence data, and after participating in the source verification, it, along with the evidence hash, affects the source verification result. The source verification result is written to the judgment evidence cache area and forms the evidence-bound input content with the judgment result in the judgment evidence description. The evidence-bound input content continues to enter the evidence-bound process, enabling the evidence-bound judgment record to simultaneously store the judgment result, evidence hash, original evidence index, and evidence source verification result.

[0052] Preferably, during the evidence binding process, the judgment access verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result are sequentially written into the judgment record splicing cache. A judgment binding data block is then generated according to the field order among these three parameters. The judgment record splicing cache stores the evidence binding input content and the associated judgment access verification result and judgment flow verification result. The judgment binding data block originates from the judgment access verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result in the judgment record splicing cache. The judgment binding data block continues to be written into the judgment record position corresponding to the judgment entry in the declarative primitive state machine. The judgment record position is used to store the evidence binding judgment record generated by the judgment primitive instruction unit to be verified. After the evidence binding judgment record is formed, the evidence binding judgment record retains the judgment access verification result and the judgment flow verification result, so that when the verification elements of the evidence binding judgment record are reconstructed in the future, the verification comparison element description required by the verification primitive instruction unit to be verified can be read.

[0053] Preferably, in a scenario where multiple agents collaboratively invoke tools, the authorization state responsibility record can record the dynamic authorization mapping of the task scheduling node granting tool invocation permissions to the execution node, and the judgment primitive instruction unit to be verified can represent the judgment request made by the execution node for the tool invocation result. After generating a judgment access element description based on the authorization state responsibility record, the entrusted subject identifier in the judgment access element description is matched with the judgment initiator identifier in the judgment primitive instruction unit to be verified, the task boundary description in the judgment access element description is matched with the judgment task boundary in the judgment primitive instruction unit to be verified, and the permission bitmap in the judgment access element description is matched with the judgment permission request in the judgment primitive instruction unit to be verified; the processing results of subject matching, boundary matching, and permission matching form the judgment access verification result. Then, based on the judgment flow constraints, the judgment flow verification is performed on the judgment primitive instruction unit to be verified to generate a judgment flow verification result, and under the condition that the judgment access verification result and the judgment flow verification result are jointly participated in, the judgment evidence description is extracted, and then the evidence binding judgment record is generated. In this way, the evidence-bound judgment record is not simply a log that stores the judgment result, but a judgment stage status record formed by the authorization status and responsibility record, the judgment access element description, the judgment access verification result, the judgment flow verification result, and the judgment evidence description. The judgment stage status record, as the state expression of the evidence-bound judgment record at the judgment stage, continues to participate in the subsequent verification element reconstruction.

[0054] Optionally, based on the description of the judgment evidence, an evidence-binding judgment record is generated, including: The original evidence data is read according to the description of the judgment evidence to obtain the evidence data to be compared; The evidence description in the judgment is compared with the evidence data to be compared, and a source verification is performed to obtain the evidence source verification result. The judgment result is extracted from the primitive instruction unit to be verified to obtain the judgment result; The judgment access verification result, the judgment flow verification result, the judgment evidence description, the evidence source verification result, and the judgment result are subjected to evidence binding processing to generate an evidence binding judgment record.

[0055] Preferably, the specific implementation process of step 2 is as follows: When reading the original evidence data according to the judgment evidence description, the original evidence index is first read from the judgment evidence description, and the index type of the original evidence index is parsed to form an evidence reading request. The evidence reading request carries the storage address range corresponding to the original evidence index and writes the storage address range into the evidence reading cache. The evidence reading cache locates the evidence storage location according to the storage address range and reads the original evidence data corresponding to the original evidence index from the evidence storage location. After the original evidence data is read, the original evidence data is written into the evidence reading cache to form the evidence data to be compared. The evidence data to be compared continues to participate in the subsequent same-source verification, so that the original evidence index in the judgment evidence description can form a continuous evidence reading link through the evidence reading request, the evidence storage location, and the evidence data to be compared. The evidence reading link is used to express the reading association between the original evidence index, the evidence reading request, the evidence storage location, and the evidence data to be compared, and continues to participate in the formation of the evidence source verification result in the subsequent same-source verification through the evidence storage location and the evidence data to be compared.

[0056] Preferably, in the specific technical implementation of step 2, before reading the original evidence data, the evidence reading cache performs page boundary verification on the storage address range corresponding to the original evidence index to obtain the page boundary verification result; the page boundary verification result is used to determine whether the storage address range falls within the readable evidence storage area. After the page boundary verification result meets the reading conditions, the evidence reading cache applies a read lock mark to the evidence storage location, and reads the original evidence data from the evidence storage location when the read lock mark is in a valid state to form the evidence data to be compared. The read lock mark is written into the evidence reading cache along with the evidence data to be compared, so that subsequent same-source verification can read the read lock mark corresponding to the evidence data to be compared, thereby distinguishing that the evidence data to be compared comes from a continuous reading process from the same evidence storage location, rather than from scattered record content. The page boundary verification result and the read lock mark jointly participate in the formation of the evidence reading link, so that the evidence reading link includes not only the address relationship between the original evidence index and the evidence storage location, but also the boundary verification state and lock state of the evidence storage location during reading.

[0057] Preferably, when performing source verification between the judgment evidence description and the evidence data to be compared, the evidence hash is first read from the judgment evidence description and written into the evidence summary cache. Then, the evidence data to be compared is written into the same evidence summary cache, and a summary calculation is performed on the evidence data to form a recalculated evidence summary. The recalculated evidence summary originates from the evidence data to be compared, and the evidence hash originates from the judgment evidence description. The evidence summary cache places the evidence hash and the recalculated evidence summary within the same verification address range, enabling subsequent source verification to read the evidence hash and the recalculated evidence summary within the same verification address range. After source verification, the evidence hash and the recalculated evidence summary form the evidence source verification result. The evidence source verification result continues to participate in the evidence binding process, enabling the evidence source relationship between the evidence data to be compared and the judgment evidence description to be written into the evidence binding judgment record. The evidence source relationship is jointly expressed by the evidence hash, the recalculated evidence summary, the evidence storage location, and the read lock flag, and enters the judgment binding cache along with the evidence source verification result during the evidence binding process.

[0058] Preferably, in the specific execution of the same-source verification, the evidence storage location corresponding to the evidence data to be compared is first read according to the original evidence index in the judgment evidence description. Then, an evidence recalculation digest is formed based on the evidence data to be compared, and the original evidence index, evidence storage location, evidence hash, and evidence recalculation digest are written into the same-source verification data block. The same-source verification data block organizes the data in the order of original evidence index, evidence storage location, evidence hash, and evidence recalculation digest, and performs a consistency comparison on the evidence hash and evidence recalculation digest to form a same-source verification mark. The same-source verification mark, together with the original evidence index and evidence storage location, is written into the evidence source verification result, so that the evidence source verification result not only indicates the comparison status between the evidence hash and the evidence recalculation digest, but also retains the address information of the evidence storage location from which the evidence data to be compared originates. After the evidence source verification result is formed, it is sent to the judgment binding cache area and participates in the evidence binding process together with the subsequently extracted judgment result. The same source verification data block continues to serve as the source data of the evidence source verification result after the evidence source verification result is formed, so that the evidence binding judgment record can trace the original evidence index, evidence storage location, evidence hash and evidence recalculation summary on which the evidence source verification result is based.

[0059] Preferably, when extracting the judgment result from the judgment primitive instruction unit to be verified, the judgment payload field is first read from the judgment primitive instruction unit to be verified, and the judgment payload reading position is formed according to the field offset position of the judgment payload field in the judgment primitive receiving buffer. The judgment payload reading position is used to locate the judgment content area in the judgment payload field; then the judgment content is read from the judgment content area, and field boundary resolution is performed on the judgment content to generate the judgment result. The judgment result is written to the judgment result buffer and a judgment result address association is formed with the judgment payload reading position; the judgment result address association is used to express the correspondence between the judgment result and the source position of the judgment payload field in the judgment primitive receiving buffer. The judgment result buffer continues to participate in the evidence binding process, so that the judgment result can be written into the evidence binding judgment record together with the judgment payload reading position and the judgment result address association, avoiding the judgment result from existing independently of the judgment primitive instruction unit to be verified.

[0060] Preferably, after the judgment result is extracted, the judgment access verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result are written into the judgment binding cache. The judgment binding cache first receives the judgment access verification result and the judgment flow verification result to form judgment entry constraint data; the judgment entry constraint data is used to express the access verification status and flow verification status when the judgment primitive instruction unit to be verified enters the judgment stage. The judgment binding cache then receives the judgment evidence description and the evidence source verification result to form evidence source constraint data; the evidence source constraint data is used to express the evidence source relationship between the evidence hash, the original evidence index, and the evidence data to be compared in the judgment evidence description. The judgment binding cache then receives the judgment result to form judgment content data; the judgment content data is used to express the judgment result extracted from the judgment primitive instruction unit to be verified. The judgment entry constraint data, evidence source constraint data, and judgment content data are arranged in the judgment binding cache in the order of judgment entry constraint data, evidence source constraint data, and judgment content data to generate a judgment binding data block. The judgment binding data block is then written to the judgment record position, so that the evidence binding judgment record can simultaneously express the technical relationship between the judgment access verification result, the judgment flow verification result, the judgment evidence description, the evidence source verification result, and the judgment result.

[0061] Preferably, when generating an evidence-bound judgment record, the judgment record position is first located in the state table of the declarative primitive state machine based on the judgment entry point in the judgment flow verification result, and an address inheritance relationship is established between the judgment record position and the primitive flow credential in the authorized state rights and responsibilities record. Then, the judgment-bound data block is written to the judgment record position to form the evidence-bound judgment record. The judgment record position is used to store the judgment-bound data block, which stores the judgment access verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result. The address inheritance relationship is used to express the state transition inheritance relationship between the judgment record position and the primitive flow credential. After the evidence-bound judgment record is formed, it retains the address inheritance relationship between the judgment record position and the primitive flow credential, enabling subsequent reconstruction of the verification elements of the evidence-bound judgment record to read the judgment access verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result along the judgment record position, and to trace back to the primitive flow credential in the authorized state rights and responsibilities record along the address inheritance relationship.

[0062] Preferably, when writing the evidence-binding judgment record to the judgment record location, the judgment-binding data block is written in segments. The segmented writing first writes the judgment access verification result and the judgment circulation verification result to the access circulation area of ​​the judgment record location; then, it writes the judgment evidence description and the evidence source verification result to the evidence verification area of ​​the judgment record location; finally, it writes the judgment result to the judgment content area of ​​the judgment record location. The access circulation area, the evidence verification area, and the judgment content area all belong to the judgment record location; the access circulation area is used to store the judgment access verification result and the judgment circulation verification result; the evidence verification area is used to store the judgment evidence description and the evidence source verification result; and the judgment content area is used to store the judgment result. The evidence-bound judgment record is stored in partitions within the judgment record location, including the access flow area, evidence verification area, and judgment content area. This allows subsequent verification element reconstruction to read the data content from the access flow area, evidence verification area, and judgment content area respectively. The access flow area, evidence verification area, and judgment content area correspond to the judgment entry constraint data, evidence source constraint data, and judgment content data, respectively, ensuring that the write structure of the judgment-bound data block in the judgment record location is consistent with the arrangement structure in the judgment-bound cache area.

[0063] Preferably, in a scenario where multiple agents collaboratively invoke tools, the judgment request made by the execution node corresponding to the judgment primitive instruction unit to be verified in response to the tool invocation result is recorded in the judgment evidence description, which includes the evidence hash and original evidence index carried by the judgment request. When reading the original evidence data according to the judgment evidence description, the original evidence index points to the evidence storage location corresponding to the tool invocation result, and the original evidence data in the evidence storage location is read as the evidence data to be compared. After performing a source verification between the judgment evidence description and the evidence data to be compared, the obtained evidence source verification result is used to characterize the evidence source relationship between the evidence hash carried by the judgment request and the evidence data to be compared corresponding to the tool invocation result. Subsequently, the judgment result is extracted from the judgment primitive instruction unit to be verified, and the judgment access verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result are jointly written into the evidence-bound judgment record. Thus, the evidence-bound judgment record does not separately store the execution record of the judgment result, but rather is a stateful record that binds the judgment access verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result to the judgment record location; this stateful record continues to participate in the subsequent verification element reconstruction as the evidence-bound judgment record.

[0064] Optionally, in step 3, a verification status responsibility record is generated based on the description of the verification comparison elements and the determined verification primitive instruction units to be verified, including: The verification entry verification is performed on the verification primitive instruction unit to be verified according to the verification comparison element description, so as to obtain the verification entry verification result; The verification conclusion is extracted from the verification primitive instruction unit to be verified in order to obtain the verification conclusion; The consistency between the verification conclusion and the description of the verification comparison elements is verified to generate a verification status responsibility record.

[0065] Preferably, the specific implementation process of step 3 is as follows: Before generating the verification state responsibility record based on the verification comparison element description and the determined verification primitive instruction unit to be verified, the admission flow area, the evidence verification area, and the judgment content area are read along the judgment record position corresponding to the evidence binding judgment record. The judgment record position is the storage position corresponding to the evidence binding judgment record in the state table of the declarative primitive state machine. The judgment admission verification result and the judgment flow verification result are read from the admission flow area, the judgment evidence description and the evidence source verification result are read from the evidence verification area, and the judgment result is read from the judgment content area. Then, the judgment admission verification result, the judgment flow verification result, the judgment evidence description, the evidence source verification result, and the judgment result are written into the verification reconstruction cache area to form the verification comparison element description. The verification and reconstruction cache is used to receive the judgment access verification results, judgment flow verification results, judgment evidence descriptions, evidence source verification results, and judgment results read from the judgment record position of the evidence-bound judgment record. The verification comparison element descriptions come from the judgment access verification results, judgment flow verification results, judgment evidence descriptions, evidence source verification results, and judgment results in the verification and reconstruction cache, and continue to participate in the verification entry verification of the verification primitive instruction unit to be verified, so that the judgment access verification results, judgment flow verification results, judgment evidence descriptions, evidence source verification results, and judgment results in the evidence-bound judgment record can be continuously read by the verification processing stage.

[0066] Preferably, in the specific technical implementation of step 3, when determining the verification primitive instruction unit to be verified, the verification primitive instruction unit is first received based on the evidence-binding judgment record, and the verification primitive instruction unit is written into the verification primitive receiving buffer. Then, field boundary resolution is performed on the verification primitive instruction unit to form a verification field offset table. The verification field offset table is used to record the storage locations of the verification entry declaration field, verification object index field, and verification conclusion payload field in the verification primitive instruction unit within the verification primitive receiving buffer. Subsequently, the verification object index field is located according to the verification field offset table, the verification object index is read from the verification object index field, and the judgment record position corresponding to the evidence-binding judgment record is read according to the verification object index. The judgment record position is then written into the verification primitive receiving buffer to form the verification primitive instruction unit to be verified. The verification primitive instruction unit to be verified carries the verification field offset table, the verification object index, and the judgment record position, enabling subsequent verification entry verification to read the verification entry declaration field based on the verification field offset table and to read the verification comparison element description in the evidence-binding judgment record based on the judgment record position.

[0067] Preferably, when performing verification entry verification on the verification primitive instruction unit to be verified according to the verification comparison element description, the judgment flow verification result is first read from the verification comparison element description, and the verification entry is parsed from the judgment flow verification result; then, based on the verification field offset table carried by the verification primitive instruction unit to be verified, the verification entry declaration is read from the verification entry declaration field, and the verification entry declaration and the verification entry are verified to form a verification entry correspondence result. The verification entry correspondence result is written to the verification entry verification buffer, which continues to receive the subsequently formed verification object correspondence result, and generates a verification entry verification result based on the verification entry correspondence result and the verification object correspondence result; the verification entry verification result is used to express whether the verification entry declaration in the verification primitive instruction unit to be verified corresponds to the verification entry registered in the evidence binding judgment record.

[0068] Preferably, after forming the verification entry point correspondence result, the verification object index field is located according to the verification field offset table carried by the verification primitive instruction unit to be verified, and the verification object index is read from the verification object index field; then, the verification object index is matched with the judgment record position to form the verification object correspondence result. The verification object correspondence result is used to characterize whether the verification object pointed to by the verification primitive instruction unit to be verified corresponds to the evidence-bound judgment record; the verification object correspondence result and the verification entry point correspondence result are written together into the verification entry point verification cache area to obtain the verification entry point verification result. The verification entry point verification result continues to participate in the triggering determination of the verification conclusion extraction, so that the verification conclusion extraction is executed when the verification primitive instruction unit to be verified enters the verification entry point and points to the evidence-bound judgment record.

[0069] Preferably, when extracting the verification conclusion from the verification primitive instruction unit to be verified, the verification conclusion payload field is first located according to the verification field offset table, and the verification payload data is read from the verification conclusion payload field; then, field boundary resolution is performed on the verification payload data to extract the verification conclusion. The verification conclusion is written to the verification conclusion cache area, and a verification conclusion source association is formed with the storage location of the verification conclusion payload field in the verification primitive receiving cache area. The verification conclusion source association is used to express the source relationship between the verification conclusion and the verification conclusion payload field in the verification primitive instruction unit to be verified; the verification conclusion source association continues to participate in the consistency verification, so that the verification conclusion can be traced back to the verification conclusion payload field in the verification primitive instruction unit to be verified when participating in subsequent processing, rather than participating in verification independently of the verification primitive instruction unit to be verified.

[0070] Preferably, when verifying the consistency between the verification conclusion and the description of the verification comparison elements, the judgment result and the evidence source verification result are first read from the description of the verification comparison elements, and the judgment result, the evidence source verification result, the verification conclusion, and the verification conclusion source association are written into the verification consistency cache. The verification consistency cache organizes data in the order of judgment result, evidence source verification result, verification conclusion, and verification conclusion source association to form a verification consistency data block. The verification consistency data block is used to express the content correspondence between the verification conclusion and the judgment result, the source correspondence between the verification conclusion and the evidence source verification result, and the source relationship between the verification conclusion and the verification conclusion payload field. The verification consistency data block continues to participate in the consistency verification, enabling the consistency verification to simultaneously handle the verification relationships of the three aspects: judgment content, evidence source, and verification conclusion source.

[0071] Preferably, during the consistency verification process, the verification conclusions and judgment results in the consistency data block are first verified for content correspondence to form a content correspondence verification result; then, the verification conclusions and evidence source verification results in the consistency data block are verified for source correspondence to form a source correspondence verification result; subsequently, the verification conclusions and verification conclusion source associations in the consistency data block are verified for source location to form a source location verification result. The content correspondence verification result, source correspondence verification result, and source location verification result are written to the consistency cache and associated with the verification entry verification result to form a verification state record data block. The verification state record data block originates from the verification entry verification result, verification conclusion, judgment result, evidence source verification result, and verification conclusion source association; the verification state record data block continues to participate in the writing of the verification state rights and responsibilities record, enabling the verification state rights and responsibilities record to simultaneously express the verification entry status, verification conclusion content, judgment content correspondence, evidence source correspondence, and verification conclusion source relationship.

[0072] Preferably, when generating the verification state responsibility record, the verification record position corresponding to the evidence-binding judgment record is first located in the state table of the declarative primitive state machine, and an address inheritance relationship is established between the verification record position and the judgment record position. Then, the verification state record data block is written to the verification record position to form the verification state responsibility record. The verification record position is used to store the verification state record data block, which stores the verification entry verification result, verification conclusion, content-corresponding verification result, source-corresponding verification result, and source location verification result. The address inheritance relationship is used to express the state transition inheritance relationship between the verification record position and the judgment record position. After the verification state responsibility record is formed, it retains the address inheritance relationship between the verification record position and the judgment record position, enabling subsequent final element reconstruction to read the verification entry verification result and verification conclusion along the verification record position, and to read the judgment result and evidence source verification result in the evidence-binding judgment record along the address inheritance relationship.

[0073] Preferably, when writing the verification state responsibility record to the verification record location, the verification state record data block is partitioned. Partition writing first writes the verification entry verification result to the entry verification area of ​​the verification record location, then writes the verification conclusion and the verification conclusion source association to the verification conclusion area of ​​the verification record location, and subsequently writes the content-corresponding verification result, the source-corresponding verification result, and the source location verification result to the consistency verification area of ​​the verification record location. The entry verification area, the verification conclusion area, and the consistency verification area all belong to the verification record location; the entry verification area is used to store the verification entry verification result, the verification conclusion area is used to store the verification conclusion and the verification conclusion source association, and the consistency verification area is used to store the content-corresponding verification result, the source-corresponding verification result, and the source location verification result. The verification state responsibility record is partitioned and stored through the entry verification area, the verification conclusion area, and the consistency verification area in the verification record location, enabling subsequent final element reconstruction to read the data content in the entry verification area, the verification conclusion area, and the consistency verification area respectively.

[0074] Preferably, in a scenario where multiple agents collaboratively invoke tools, the evidence-binding judgment record corresponds to the evidence-binding judgment record made by the execution node in response to the tool invocation result, and the verification primitive instruction unit to be verified corresponds to the verification request initiated by the verification node on the evidence-binding judgment record. After reconstructing the verification elements of the evidence-binding judgment record, the resulting verification comparison element description includes the judgment access verification result, the judgment flow verification result, the judgment evidence description, the evidence source verification result, and the judgment result. After verifying the verification entry of the verification primitive instruction unit to be verified according to the verification comparison element description, the obtained verification entry verification result is used to characterize whether the verification primitive instruction unit to be verified points to the verification entry registered in the evidence-binding judgment record. After extracting the verification conclusion from the verification primitive instruction unit to be verified, the verification conclusion is verified against the verification comparison element description to generate a verification state responsibility record. Therefore, the verification status record is not a record that stores the verification conclusion separately. Instead, it writes the verification entry verification result, verification conclusion, verification conclusion source association, content corresponding verification result, source corresponding verification result, source location verification result, and address connection relationship between the verification record location and the decision record location into the status record of the verification record location.

[0075] Optionally, the consistency between the verification conclusion and the description of the verification comparison elements is verified to generate a verification status responsibility record, including: The consistency between the verification conclusion and the description of the verification comparison element is verified to generate a verification status determination result; Based on the verification results of the verification entry and the verification status determination results, the audit status of the evidence binding judgment record is rewritten to generate an audit judgment record; Based on the verification status determination result, the audit judgment record is configured to assume responsibility, so as to generate a responsibility assumption identifier; The responsibility assignment identifier is written into the audit decision record to obtain the verification status responsibility record.

[0076] Preferably, in step 3, when verifying the consistency between the verification conclusion and the verification comparison element description, the verification conclusion and the verification conclusion source association are first read from the verification conclusion cache, and the verification comparison element description is read from the verification reconstruction cache. Then, the judgment admission verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result are separated from the verification comparison element description. The verification conclusion, verification conclusion source association, judgment admission verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result are written into the verification state judgment cache to form a verification state judgment data block. The verification state judgment data block originates from the verification conclusion cache and the verification reconstruction cache. The verification state judgment data block continues to participate in the consistency verification, ensuring that the verification conclusion does not deviate from the verification primitive instruction unit to be verified, and that the judgment result and evidence source verification result do not deviate from the evidence-bound judgment record. This allows the verification conclusion, verification conclusion source association, judgment admission verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result to be placed in the same verification state judgment cache for subsequent processing.

[0077] Preferably, when a verification-state judgment data block enters the consistency verification process, the verification conclusions and judgment results in the verification-state judgment data block are first verified for content correspondence to form a content correspondence verification result; then, the verification conclusions and evidence source verification results in the verification-state judgment data block are verified for source correspondence to form a source correspondence verification result; subsequently, the source association of the verification conclusions in the verification-state judgment data block is verified for source location to form a source location verification result. The content correspondence verification result, source correspondence verification result, and source location verification result are all written to the verification-state judgment cache and arranged in association with the verification entry verification result to generate the verification-state judgment result. The verification-state judgment result originates from the verification entry verification result, content correspondence verification result, source correspondence verification result, and source location verification result. The verification-state judgment result continues to participate in the audit status rewriting of the evidence-binding judgment record, ensuring that the audit status rewriting of the evidence-binding judgment record is simultaneously constrained by the verification entry verification result, content correspondence verification result, source correspondence verification result, and source location verification result.

[0078] Preferably, when rewriting the audit status of the evidence-bound judgment record based on the verification entry verification result and the verification status determination result, the access flow area, evidence verification area, and judgment content area are first read according to the judgment record position corresponding to the evidence-bound judgment record, and the data content in the access flow area, evidence verification area, and judgment content area is written into the audit status rewriting cache area; then, the verification entry verification result and the verification status determination result are written into the same audit status rewriting cache area to form an audit rewriting data block. The audit rewriting data block includes the judgment access verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result in the evidence-bound judgment record, as well as the verification entry verification result and the verification status determination result; the audit rewriting data block continues to participate in the rewriting of the audit status fields, so that the judgment access verification result, judgment flow verification result, judgment evidence description, evidence source verification result, and judgment result in the evidence-bound judgment record form a writable data association with the verification entry verification result and the verification status determination result in the same audit status rewriting cache area.

[0079] Preferably, in the specific execution of the audit status field rewriting, the audit status field is first located in the judgment record position, and the original audit status of the audit status field is read; then, based on the verification entry verification result and verification status judgment result in the audit rewriting data block, the original audit status is rewritten to form the rewritten audit status. The rewritten audit status is written back to the audit status field in the judgment record position, and a regional association is established with the access flow area, evidence verification area, and judgment content area in the judgment record position; the regional association is used to express the storage correspondence between the audit status field and the access flow area, evidence verification area, and judgment content area. The regional association continues to participate in the generation of the audit judgment record to generate the audit judgment record. The audit judgment record originates from the judgment record position of the evidence-bound judgment record. The audit judgment record adds the rewritten audit status and regional association on the basis of the evidence-bound judgment record. The audit judgment record continues to participate in the responsibility assignment configuration, so that the responsibility assignment identifier can be configured based on the rewritten audit status, regional association, judgment result, and evidence source verification result in the audit judgment record.

[0080] Preferably, when configuring responsibility assignment for audit judgment records based on verification status determination results, the rewritten audit status, regional association, judgment result, evidence source verification result, and judgment flow verification result are first read from the audit judgment record. Then, the content-corresponding verification result, source-corresponding verification result, and source location verification result are read from the verification status determination result. Subsequently, the rewritten audit status, regional association, judgment result, evidence source verification result, judgment flow verification result, content-corresponding verification result, source-corresponding verification result, and source location verification result are written into the responsibility assignment cache area to form a responsibility assignment configuration data block. The responsibility assignment configuration data block originates from the audit judgment record and verification status determination results. This data block continues to participate in the generation of the responsibility assignment identifier, ensuring that the responsibility assignment identifier does not originate solely from the verification conclusion but is jointly formed by the rewritten audit status, regional association, judgment result, evidence source verification result, and judgment flow verification result from the audit judgment record, as well as the content-corresponding verification result, source-corresponding verification result, and source location verification result from the verification status determination result.

[0081] Preferably, when generating the responsibility acceptance identifier, the rewritten audit status is first read from the responsibility acceptance configuration data block, and the status segment of the responsibility acceptance identifier is determined based on the rewritten audit status. Then, the judgment result and evidence source verification result are read from the responsibility acceptance configuration data block, and the judgment evidence segment of the responsibility acceptance identifier is determined based on the judgment result and evidence source verification result. Subsequently, the content-correspondence verification result, source-correspondence verification result, and source location verification result are read from the responsibility acceptance configuration data block, and the verification judgment segment of the responsibility acceptance identifier is determined based on the content-correspondence verification result, source-correspondence verification result, and source location verification result. The status segment, judgment evidence segment, and verification judgment segment are written to the responsibility acceptance cache to generate the responsibility acceptance identifier. The responsibility acceptance identifier continues to participate in the writing of audit judgment records, enabling the audit judgment records to continue recording the responsibility acceptance identifier beyond the rewritten audit status.

[0082] Preferably, when writing the responsibility assignment identifier into the audit judgment record, the responsibility assignment field is first located in the corresponding judgment record position of the audit judgment record, and the responsibility assignment identifier is written into the responsibility assignment field; then, the responsibility assignment field is associated with the audit status field, the evidence verification area, and the judgment content area to form a verification-state responsibility record. The responsibility assignment field is used to store the responsibility assignment identifier, the audit status field is used to store the rewritten audit status, the evidence verification area is used to store the judgment evidence description and the evidence source verification result, and the judgment content area is used to store the judgment result; the verification-state responsibility record originates from the audit judgment record after the responsibility assignment identifier is written, and continues to serve as the reading object for subsequent final element reconstruction, enabling subsequent final element reconstruction to read the responsibility assignment identifier, the rewritten audit status, the evidence source verification result, and the judgment result along the verification-state responsibility record.

[0083] Preferably, after the verification state responsibility record is formed, a verification state record position is assigned to the state table of the declarative primitive state machine, and an address inheritance relationship is established between the verification state record position and the corresponding judgment record position of the audit judgment record. Subsequently, the verification entry verification result, verification state judgment result, audit judgment record, and responsibility inheritance identifier are written into the verification state record position to form the state table entry of the verification state responsibility record. The verification state record position is used to store the verification entry verification result, verification state judgment result, audit judgment record, and responsibility inheritance identifier, and the address inheritance relationship is used to express the state transition inheritance relationship between the verification state record position and the judgment record position. The state table entry of the verification state responsibility record continues to participate in the subsequent reconstruction of the final elements, so that the final stage can read the verification state responsibility record from the verification state record position and can trace back to the audit judgment record and the evidence binding judgment record along the address inheritance relationship.

[0084] Preferably, when writing the verification status responsibility record to the verification status record location, the verification status responsibility record is written in partitions. Partition writing first writes the verification entry verification result to the entry verification area of ​​the verification status record location, then writes the verification status judgment result to the judgment result area of ​​the verification status record location, subsequently writes the audit judgment record to the audit record area of ​​the verification status record location, and finally writes the responsibility assignment identifier to the responsibility assignment area of ​​the verification status record location. The entry verification area, judgment result area, audit record area, and responsibility assignment area all belong to the verification status record location; the entry verification area is used to store the verification entry verification result, the judgment result area is used to store the verification status judgment result, the audit record area is used to store the audit judgment record, and the responsibility assignment area is used to store the responsibility assignment identifier. The verification status responsibility record is stored in partitions through the entry verification area, judgment result area, audit record area, and responsibility assignment area in the verification status record location, enabling subsequent final element reconstruction to read the data content of the entry verification area, judgment result area, audit record area, and responsibility assignment area respectively.

[0085] Preferably, in a scenario where multiple agents collaboratively invoke tools, the evidence-binding judgment record corresponds to the evidence-binding judgment record formed by the execution node's tool invocation result, and the verification request initiated by the verification node by the verification primitive instruction unit to the evidence-binding judgment record. After verifying the consistency between the verification conclusion and the verification comparison element description, the resulting verification state judgment result is used to express the corresponding state between the verification conclusion and the judgment result, the evidence source verification result, and the verification conclusion source association. After rewriting the audit state of the evidence-binding judgment record according to the verification entry verification result and the verification state judgment result, an audit judgment record is generated. Then, the audit judgment record is configured with responsibility assignment according to the verification state judgment result to generate a responsibility assignment identifier. Finally, the responsibility assignment identifier is written into the audit judgment record to obtain the verification state responsibility record. Thus, the verification state responsibility record does not store the verification conclusion separately, but rather writes the verification state judgment result, the audit judgment record, and the responsibility assignment identifier together into the verification state record position in the declarative primitive state machine, and continues to participate in the subsequent final element reconstruction through the verification state record position.

[0086] Optionally, step 4: Based on the description of the termination constraint elements, generate an agent decision-making responsibility control and evidence storage report, including: Based on the verification state responsibility record, the terminal primitive instruction unit is received to obtain the terminal primitive instruction unit to be verified; According to the description of the termination constraint elements, the termination entry verification is performed on the termination primitive instruction unit to be verified to obtain the termination entry verification result; Based on the termination entry verification result, the session boundary is located for the termination constraint element description to determine the authorization record to be terminated; The pending authorization records are marked as invalid to form authorization invalidation records; Based on the authorization expiration records, generate a report on the management and evidence preservation of the decision-making authority and responsibility of the intelligent agent.

[0087] Preferably, the specific implementation process of step 4 is as follows: Before generating the agent decision-making responsibility control evidence storage report based on the description of the termination constraint elements, first read the entry verification area, judgment result area, audit record area, and responsibility acceptance area along the verification state record position corresponding to the verification state responsibility record, and read the verification entry verification result from the entry verification area, the verification state judgment result from the judgment result area, the audit judgment record from the audit record area, and the responsibility acceptance identifier from the responsibility acceptance area; then, read the judgment corresponding to the evidence binding judgment record along the address inheritance relationship between the verification state record position and the judgment record position. The system records the location and reads the judgment flow verification result and evidence source verification result from the judgment record location. Then, following the address inheritance relationship between the judgment record location and the authorization state record location, it reads the corresponding authorization state record location for the authorization state responsibility record and reads the dynamic authorization mapping, primitive flow credential, and signature identity digest from the authorization state record location. Finally, it writes the verification entry verification result, verification state judgment result, audit judgment record, responsibility inheritance identifier, judgment flow verification result, evidence source verification result, dynamic authorization mapping, primitive flow credential, and signature identity digest into the termination constraint reconstruction cache to form the termination constraint element description. The termination constraint element description originates from the address inheritance reading results of the verification state responsibility record, evidence binding judgment record, and authorization state responsibility record. The termination constraint element description continues to participate in the termination entry verification of the termination primitive instruction unit to be verified, enabling the termination stage to read the status records during the authorization, judgment, and verification processes along the verification state responsibility record, evidence binding judgment record, and authorization state responsibility record.

[0088] Preferably, when receiving a termination primitive instruction unit based on the verification state responsibility record, the responsibility acceptance area in the verification state record position is first read. After a responsibility acceptance identifier exists in the responsibility acceptance area, the termination primitive instruction unit is written into the termination primitive receiving buffer. Subsequently, field boundary resolution is performed on the termination primitive instruction unit to form a termination field offset table. The termination field offset table is used to record the storage locations of the termination entry declaration field, termination object index field, and termination authorization location field in the termination primitive instruction unit in the termination primitive receiving buffer. Then, the termination object index field is read according to the termination field offset table, and the verification state record position corresponding to the termination object index field is written into the termination primitive receiving buffer to form a termination primitive instruction unit to be verified. The termination primitive instruction unit to be verified carries the termination field offset table and the verification state record position, enabling subsequent termination entry verification to simultaneously read the termination entry declaration field in the termination primitive instruction unit and the primitive flow credentials in the termination constraint element description.

[0089] Preferably, when performing terminal entry verification on the terminal primitive instruction unit to be verified according to the terminal constraint element description, the terminal entry is first read from the primitive flow credential in the terminal constraint element description, and the terminal entry declaration is read from the terminal entry declaration field according to the terminal field offset table carried by the terminal primitive instruction unit to be verified; then, the terminal entry declaration and the terminal entry are verified for entry correspondence to form a terminal entry correspondence result. The terminal entry correspondence result is written to the terminal entry verification buffer, which continues to receive the subsequently formed terminal object correspondence result and terminal acceptance correspondence result, and generates a terminal entry verification result based on the terminal entry correspondence result, the terminal object correspondence result, and the terminal acceptance correspondence result; the terminal entry verification result is used to express whether the terminal primitive instruction unit to be verified has entered the terminal entry registered by the primitive flow credential.

[0090] Preferably, after forming the termination entry correspondence result, the termination object index field is located according to the termination field offset table, and the termination object index is read from the termination object index field; then, the termination object index is verified against the verification state record position to form the termination object correspondence result. The termination object correspondence result is used to express whether the termination object pointed to by the termination primitive instruction unit to be verified is the verification state responsibility record corresponding to the verification state record position; subsequently, the responsibility acceptance identifier is read from the termination constraint element description, and the termination acceptance declaration is read from the termination authorization positioning field in the termination primitive instruction unit; then, the termination acceptance declaration and the responsibility acceptance identifier are verified against the acceptance correspondence to form the termination acceptance correspondence result. Both the termination object correspondence result and the termination acceptance correspondence result are written to the termination entry verification cache area, and together with the termination entry correspondence result, generate the termination entry verification result; the termination entry verification result continues to participate in session boundary positioning, so that session boundary positioning is performed when the termination primitive instruction unit to be verified corresponds to the verification state responsibility record and the responsibility acceptance identifier.

[0091] Preferably, when locating the session boundary of the termination constraint element description based on the termination entry verification result, the termination entry verification result is first written into the session boundary location cache, and the corresponding results of the termination entry, the termination object, and the termination acceptance are read from the session boundary location cache. Then, the dynamic authorization map, primitive flow credential, and signature identity digest are read from the termination constraint element description, and written into the same session boundary location cache to form a session boundary location data block. The session boundary location data block originates from the termination entry verification result and the termination constraint element description. The session boundary location data block continues to participate in the determination of the authorization record to be terminated, ensuring that the authorization record to be terminated is located not only based on the termination primitive instruction unit, but simultaneously based on the termination entry verification result, the dynamic authorization map, the primitive flow credential, and the signature identity digest.

[0092] Preferably, when determining the authorization record to be terminated, the authorization mapping slot is first read from the dynamic authorization mapping in the session boundary positioning data block, and the authorization subject identifier, entrusted subject identifier, task boundary description, and permission bitmap are read from the authorization mapping slot. Then, the authorization entry position, decision entry, verification entry, and termination entry are read from the primitive flow credentials in the session boundary positioning data block, and the authorization entry position, decision entry, verification entry, and termination entry are verified against the termination entry verification result to form a session boundary correspondence result. Subsequently, the authorization subject identifier, entrusted subject identifier, task boundary description, permission bitmap, primitive flow credentials, signature identity digest, and session boundary correspondence result are written into the authorization record cache area to determine the authorization record to be terminated. The authorization record to be terminated originates from the dynamic authorization mapping, primitive flow credentials, signature identity digest, and session boundary correspondence result. The authorization record to be terminated continues to participate in the invalidation marking process, enabling the invalidation marking process to locate the authorization relationship content that needs to be terminated in the authorization state responsibility record.

[0093] Preferably, when performing invalidation marking on the authorization record to be terminated, the authorization record to be terminated is first read from the authorization record cache area, and the authorization state record position is located along the authorization entry position in the authorization record to be terminated; then, the authorization mapping area is located in the authorization state record position, and the dynamic authorization mapping in the authorization mapping area is read; then, an invalidation mark field is added to the authorization mapping area, and the termination entry verification result, responsibility assignment identifier, session boundary correspondence result, primitive flow credential, and signature identity digest are written into the invalidation mark field to form an authorization invalidation record. The authorization invalidation record originates from the authorization record to be terminated and the termination entry verification result. The authorization invalidation record retains the authorization subject identifier, entrusted subject identifier, task boundary description, and permission bitmap corresponding to the dynamic authorization mapping, and retains the primitive flow credential and signature identity digest in the authorization record to be terminated, while adding an invalidation mark field; the authorization invalidation record continues to participate in the generation of the agent decision-making responsibility control evidence storage report, enabling the agent decision-making responsibility control evidence storage report to read the authorization relationship content, state flow content, and termination state content in the authorization invalidation record.

[0094] Preferably, after an authorization failure record is formed, it is written to an authorization failure record cache, and a write status flag is added to the authorization failure record in the cache. The write status flag records whether the authorization failure record has been written to the authorization mapping area in the authorization state record location. After the write status flag and the authorization failure record are written to the authorization failure record cache, the authorization failure record cache continues to send the authorization failure record, the write status flag, the termination entry verification result, and the session boundary correspondence result to the evidence storage report assembly cache. The evidence storage report assembly cache reads the verification state responsibility record, audit decision record, and evidence binding decision record based on the authorization failure record, enabling the intelligent agent decision responsibility control evidence storage report to inherit the verification state responsibility record, audit decision record, and evidence binding decision record formed before the authorization failure record.

[0095] Preferably, when generating an agent decision-making responsibility management evidence report based on authorization failure records, the authorization failure records are first written into the evidence report assembly cache, and the authorization entry point, judgment entry point, and verification entry point are read according to the primitive flow credentials in the authorization failure records; then, the authorization state responsibility record is read according to the authorization entry point, the evidence binding judgment record is read according to the judgment entry point, and the verification state responsibility record is read according to the verification entry point; subsequently, the authorization state responsibility record, the evidence binding judgment record, the verification state responsibility record, the termination entry verification result, the pending termination authorization record, and the authorization failure record are written into the same evidence report assembly cache to form an evidence report data block. The evidence report data block originates from the authorization state responsibility record, the evidence binding judgment record, the verification state responsibility record, the pending termination authorization record, and the authorization failure record. The evidence report data block continues to participate in the report field encapsulation to generate the agent decision-making responsibility management evidence report.

[0096] Preferably, in the specific execution of the report field encapsulation, the authorized state rights and responsibilities record is first read from the evidence storage report data block, and an authorization field is formed based on the authorized state rights and responsibilities record; then, the evidence binding judgment record is read from the evidence storage report data block, and a judgment field is formed based on the evidence binding judgment record; subsequently, the verification state rights and responsibilities record is read from the evidence storage report data block, and a verification field is formed based on the verification state rights and responsibilities record; finally, the authorization invalidation record is read from the evidence storage report data block, and a termination field is formed based on the authorization invalidation record. The authorization field, judgment field, verification field, and termination field are written into the report field cache to form the agent decision-making rights and responsibilities control evidence storage report; the agent decision-making rights and responsibilities control evidence storage report expresses the state succession relationship between the authorized state rights and responsibilities record, the evidence binding judgment record, the verification state rights and responsibilities record, and the authorization invalidation record through the authorization field, judgment field, verification field, and termination field, respectively.

[0097] Preferably, in a scenario where multiple agents collaboratively invoke tools, the verification state responsibility record corresponds to the record formed after the verification node verifies the tool invocation result, and the termination primitive instruction unit corresponds to the termination request from the task scheduling node to terminate the tool invocation permission. After receiving the termination primitive instruction unit based on the verification state responsibility record, the resulting termination primitive instruction unit to be verified carries the termination field offset table and the position of the verification state record; the termination constraint element description is reconstructed from the verification state responsibility record, the evidence binding judgment record, and the authorization state responsibility record; after verifying the termination entry point of the termination primitive instruction unit to be verified according to the termination constraint element description, the termination entry point verification result is obtained; then, based on the termination entry point verification result, the termination constraint element description is used to locate the session boundary to determine the authorization record to be terminated; subsequently, the authorization record to be terminated is marked as invalid to form an authorization invalidation record; finally, an agent decision responsibility control evidence storage report is generated based on the authorization invalidation record. Therefore, the intelligent agent decision-making responsibility control and evidence storage report does not save the termination request record separately, but is formed by the authorization state responsibility record, evidence binding judgment record, verification state responsibility record and authorization failure record. The authorization failure record expresses the state processing result of the dynamic authorization mapping in the termination stage.

[0098] Optionally, based on the authorization expiration records, a report on the management and evidence preservation of agent decision-making authority and responsibility is generated, including: Based on the authorization expiration record, the authorization record to be terminated is erased to obtain the authorization erasure record; Write the authorized erasure record into the verification state responsibility record to obtain the final state responsibility record; The terminal state rights and responsibilities records are stored and encapsulated to generate an agent decision-making rights and responsibilities management evidence report.

[0099] Preferably, in step 4, when performing permission erasure processing on the authorization record to be terminated based on the authorization expiration record, the authorization expiration record is first read from the authorization expiration record cache, and the authorization entry position, termination entry verification result, responsibility assignment identifier, session boundary correspondence result, primitive flow credential, and signature identity digest are parsed from the authorization expiration record; then, the authorization record to be terminated in the authorization record to be terminated cache is read according to the authorization entry position, and the authorization subject identifier, entrusted subject identifier, task boundary description, permission bitmap, primitive flow credential, and signature identity digest are read from the authorization record to be terminated. The primitive flow credential in the authorization expiration record is verified to correspond to the primitive flow credential in the authorization record to be terminated to form a flow same-origin verification result; the signature identity digest in the authorization expiration record is verified to correspond to the signature identity digest in the authorization record to be terminated to form a digest same-origin verification result. The same-origin verification results and digest same-origin verification results are written to the permission erasure cache and continue to participate in the permission erasure process. This makes the permission erasure process not only based on the termination request, but also based on the state inheritance relationship between the authorization expiration record and the authorization record to be terminated.

[0100] Preferably, after receiving the authorization record to be terminated, the permission erasure cache first locates the authorization state record position based on the authorization entry position in the authorization record to be terminated, and reads the authorization mapping area from the authorization state record position; then, it reads the dynamic authorization mapping from the authorization mapping area, and locates the permission payload position corresponding to the permission bitmap from the dynamic authorization mapping. The permission payload position is used to represent the storage address range of the permission bitmap in the authorization mapping area, and the permission bitmap is used to represent the range of computing power permissions that can be invoked in the authorization record to be terminated. The permission erasure cache writes the permission payload position, permission bitmap, termination entry verification result, responsibility assignment identifier, flow same-source verification result, and digest same-source verification result into the permission erasure data block. The permission erasure data block continues to participate in the permission erasure process, so that the erasure object, erasure basis, and erasure status of the permission bitmap are associated in the same permission erasure data block.

[0101] Preferably, when performing permission erasure processing on the permission bitmap, firstly, an erasure lock mark is applied to the authorization mapping area according to the permission payload position in the permission erasure data block. After the erasure lock mark is written to the authorization mapping area, the permission bitmap in the permission payload position is read to form the permission bitmap before erasure. Subsequently, the permission bit status of the permission bitmap before erasure is cleared to form the permission bitmap after erasure, and the permission bitmap after erasure is written back to the permission payload position. The permission bitmap before erasure is used to express the permission enabled range of the authorization record to be terminated before termination, and the permission bitmap after erasure is used to express the permission closed state of the authorization record to be terminated after permission erasure processing. The permission bitmap before erasure and the permission bitmap after erasure are written to the permission erasure buffer at the same time, so that the permission erasure processing can retain the input content before the permission bit status clearing processing and the output content after the permission bit status clearing processing.

[0102] Preferably, after the erased permission bitmap is written back to the permission payload location, a readback process is performed on the permission payload location to read the erased permission bitmap after the write-back. The erased permission bitmap obtained from the readback is then compared with the erased permission bitmap in the permission erase cache to verify readback consistency, thus forming a permission erase readback result. The permission erase readback result is written to the permission erase data block and, together with the termination entry verification result, responsibility assignment identifier, session boundary correspondence result, flow same-origin verification result, and digest same-origin verification result, forms an erase state description. The erase state description continues to participate in the generation of the authorization erase record, enabling the authorization erase record to express that the permission bitmap has undergone permission bit status clearing, write-back, and readback processing, rather than simply recording a termination status identifier.

[0103] Preferably, when generating an authorization erasure record, the following are first read from the authorization erasure data block: the authorization subject identifier, the entrusted subject identifier, the task boundary description, the permission bitmap before erasure, the permission bitmap after erasure, the permission erasure readback result, the termination entry verification result, the responsibility assignment identifier, the session boundary correspondence result, the primitive flow credential, and the signature identity digest. Then, the above content is written into the authorization erasure record cache in the order of authorization relationship content, permission erasure content, termination verification content, and digest verification content to form the authorization erasure record. The authorization relationship content includes the authorization subject identifier, the entrusted subject identifier, and the task boundary description; the permission erasure content includes the permission bitmap before erasure, the permission bitmap after erasure, and the permission erasure readback result; the termination verification content includes the termination entry verification result, the responsibility assignment identifier, and the session boundary correspondence result; and the digest verification content includes the primitive flow credential and the signature identity digest. The authorization erasure record expresses the authorization relationship, permission erasure status, termination verification status, and digest verification status of the authorization record to be terminated during the termination stage through the authorization relationship content, permission erasure content, termination verification content, and digest verification content. After the authorization erase record is formed, it continues to participate in the writing of the verification state rights and responsibilities record, so that the final stage can inherit the authorization erase processing result into the verification state rights and responsibilities record.

[0104] Preferably, when writing the authorized erasure record into the verification state responsibility record, the verification state record location is first located along the primitive flow voucher in the authorized erasure record, and the entry verification area, judgment result area, audit record area, and responsibility assignment area in the verification state record location are read; then, a verification state authorized erasure area is created in the verification state record location, and the authorized erasure record is written into the verification state authorized erasure area. The verification state authorized erasure area belongs to the verification state record location. The verification state authorized erasure area is used to store the authorized erasure record, the entry verification area is used to store the verification entry verification result, the judgment result area is used to store the verification state judgment result, the audit record area is used to store the audit judgment record, and the responsibility assignment area is used to store the responsibility assignment identifier. After the authorized erasure record is written into the verification state authorized erasure area, the verification state authorized erasure area and the responsibility assignment area establish an address association, so that the authorized erasure record and the responsibility assignment identifier can form a readable final assignment relationship in the verification state responsibility record.

[0105] Preferably, after the authorized erasure record is written to the verification-state authorized erasure area, the verification entry verification result, verification state judgment result, audit decision record, responsibility assignment identifier, and authorized erasure record in the verification state record location are read, and these are written to the final state record cache to form a final state record data block. The final state record cache is used to cache the final stage assignment content read from the verification-state rights and responsibilities record and the authorized erasure record. The final state record data block originates from the verification entry verification result, verification state judgment result, audit decision record, responsibility assignment identifier, and authorized erasure record in the final state record cache. The final state record data block continues to participate in the generation of the final state rights and responsibilities record, enabling the final state rights and responsibilities record to simultaneously assume the rights and responsibilities status of the verification stage and the permission erasure status of the final stage.

[0106] Preferably, when generating the final state responsibility record, the final state record position corresponding to the verification state record position is first located in the state table of the declarative primitive state machine, and an address inheritance relationship is established between the final state record position and the verification state record position. Then, the final state record data block is written to the final state record position to obtain the final state responsibility record. The final state record position is used to store the final state record data block, which is used to store the verification entry verification result, verification state judgment result, audit decision record, responsibility inheritance identifier, and authorized erasure record. The address inheritance relationship is used to express the state transition relationship between the final state record position and the verification state record position. After the final state responsibility record is formed, it continues to participate in evidence preservation encapsulation, enabling the evidence preservation encapsulation to read the final state responsibility record formed in the final stage along the final state record position.

[0107] Preferably, when writing the final state responsibility record to the final state record location, the final state responsibility record is written in partitions. Partition writing first writes the verification entry verification result and the verification state judgment result to the verification acceptance area of ​​the final state record location, then writes the audit judgment record to the audit acceptance area of ​​the final state record location, subsequently writes the responsibility acceptance identifier to the final state responsibility acceptance area of ​​the final state record location, and finally writes the authorized erasure record to the final state authorized erasure area of ​​the final state record location. The verification acceptance area, audit acceptance area, final state responsibility acceptance area, and final state authorized erasure area all belong to the final state record location; the verification acceptance area is used to store the verification entry verification result and the verification state judgment result, the audit acceptance area is used to store the audit judgment record, the final state responsibility acceptance area is used to store the responsibility acceptance identifier, and the final state authorized erasure area is used to store the authorized erasure record. The final state responsibility record participates in evidence encapsulation through the partitioned storage method in the final state record location, enabling the evidence encapsulation to read the data content of the verification acceptance area, audit acceptance area, final state responsibility acceptance area, and final state authorized erasure area respectively.

[0108] Preferably, when encapsulating the final state rights and responsibilities record for evidence storage, the final state rights and responsibilities record is first read from its location, and the verification entry verification result, verification state judgment result, audit judgment record, responsibility assignment identifier, and authorized erasure record are extracted from it. Then, based on the primitive flow credential in the authorized erasure record, the authorized state rights and responsibilities record and the evidence binding judgment record are read, and these records are written into the evidence storage encapsulation cache to form an evidence storage encapsulation data block. The evidence storage encapsulation cache is used to cache the authorized state rights and responsibilities record, the evidence binding judgment record, the verification state rights and responsibilities record, the final state rights and responsibilities record, and the authorized erasure record. The evidence storage encapsulation data block originates from the authorized state rights and responsibilities record, the evidence binding judgment record, the verification state rights and responsibilities record, the final state rights and responsibilities record, and the authorized erasure record in the evidence storage encapsulation cache. The evidence storage encapsulation data block continues to participate in the encapsulation of report fields, enabling the intelligent agent's decision-making rights and responsibilities management evidence storage report to express the state succession relationship between the authorization, judgment, verification, and termination stages.

[0109] Preferably, when encapsulating the report fields, the authorized state responsibility record is first read from the evidence encapsulation data block, and an authorization field is formed based on the authorized state responsibility record; then, the evidence binding judgment record is read from the evidence encapsulation data block, and a judgment field is formed based on the evidence binding judgment record; subsequently, the verification state responsibility record is read from the evidence encapsulation data block, and a verification field is formed based on the verification state responsibility record; then, the final state responsibility record is read from the evidence encapsulation data block, and a final field is formed based on the final state responsibility record; finally, the authorization erasure record is read from the evidence encapsulation data block, and an authorization erasure field is formed based on the authorization erasure record. The authorization field, judgment field, verification field, final field, and authorization erasure field are written into the report field cache area to generate an agent decision responsibility management evidence report; the report field cache area is used to store the authorization field, judgment field, verification field, final field, and authorization erasure field. The intelligent agent decision-making authority and responsibility management and evidence storage report expresses authorized state authority and responsibility records through the authorization field, evidence-bound judgment records through the judgment field, verification state authority and responsibility records through the verification field, termination state authority and responsibility records through the termination field, and authorized erasure records through the permission erasure field.

[0110] Preferably, in scenarios involving multi-agent collaborative tool invocation, the pending authorization record can represent the authorization relationship that needs to be terminated after the task scheduling node grants tool invocation permissions to the execution node, and the authorization failure record can represent that the authorization relationship has entered the termination entry point and completed the failure marking. When performing permission erasure processing on the pending authorization record based on the authorization failure record, the permission bitmap in the dynamic authorization mapping is first located along the pending authorization record, and then the permission bit state of the permission bitmap is cleared to form an authorization erasure record; subsequently, the authorization erasure record is written into the verification state responsibility record to obtain the termination state responsibility record; finally, the termination state responsibility record is encapsulated for evidence storage to generate an agent decision responsibility management evidence storage report. Thus, the agent decision responsibility management evidence storage report does not separately store the permission termination result record, but rather encapsulates the authorization state responsibility record, evidence binding judgment record, verification state responsibility record, termination state responsibility record, and authorization erasure record together into a readable stateful report.

[0111] like Figure 2 As shown in the figure, an intelligent agent decision-making responsibility management system based on a declarative primitive state machine is provided in this application embodiment. It includes an authorized state responsibility record generation unit, an evidence binding judgment record generation unit, a verification state responsibility record generation unit, and a responsibility management evidence storage report generation unit configured on a processor. The authorized state responsibility record generation unit is used to execute step 1 of the method described in any one of the above claims of this application; the evidence binding judgment record generation unit is used to execute step 2 of the method described in any one of the above claims of this application; the verification state responsibility record generation unit is used to execute step 3 of the method described in any one of the above claims of this application; and the responsibility management evidence storage report generation unit is used to execute step 4 of the method described in any one of the above claims of this application.

[0112] Figure 3 This is a schematic diagram of an electronic device according to an embodiment of this application. Figure 3 As shown, it includes a processor and memory; The memory is used to store computer programs; the processor is used to execute the programs stored in the memory to implement the functions of each module of the device described in the embodiments of this application, or to implement the steps of the method described.

[0113] like Figure 4 As shown, this application provides an intelligent agent decision-making responsibility management system based on a declarative primitive state machine. The system includes a task access unit, a processor, a cache unit, a state machine storage unit, an evidence storage unit, and a bus interconnection unit. The task access unit, processor, cache unit, state machine storage unit, and evidence storage unit interact with each other through the bus interconnection unit, enabling the intelligent agent's decision-making task request to form a continuous hardware acceptance path between the authorization, judgment, verification, and termination stages.

[0114] The task access unit is used to receive agent decision task requests and write them into the cache unit; the processor is used to read agent decision task requests from the cache unit and perform declarative primitive parsing processing on the agent decision task requests to generate authorization primitive instruction units; the state machine storage unit is used to receive authorization state responsibility records generated by the processor according to the authorization primitive instruction units and allocate corresponding state storage locations for the authorization state responsibility records.

[0115] The processor is also used to read the authorized state rights and responsibilities record from the state machine storage unit and perform judgment admission element parsing on the authorized state rights and responsibilities record to obtain the judgment admission element description; the cache unit is used to cache the judgment primitive instruction unit and the judgment admission element description; the processor is also used to perform admission matching between the judgment primitive instruction unit and the judgment admission element description to generate the evidence binding judgment record; the state machine storage unit is also used to store the evidence binding judgment record, so that the evidence binding judgment record can inherit the authorized state information in the authorized state rights and responsibilities record.

[0116] The processor is also used to read the evidence-binding judgment record from the state machine storage unit and reconstruct the verification elements of the evidence-binding judgment record to obtain the verification comparison element description; the cache unit is also used to cache the verification comparison element description and the verification primitive instruction unit to be verified; the processor is also used to generate the verification state responsibility record based on the verification comparison element description and the verification primitive instruction unit to be verified; the state machine storage unit is also used to store the verification state responsibility record, so that the verification state responsibility record can inherit the judgment state information and evidence association information in the evidence-binding judgment record.

[0117] The processor is also used to read the verification state rights and responsibilities record from the state machine storage unit and reconstruct the terminal elements of the verification state rights and responsibilities record to generate a terminal constraint element description; the cache unit is also used to cache the terminal constraint element description; the processor is also used to generate an agent decision rights and responsibilities management evidence storage report based on the terminal constraint element description; the evidence storage unit is used to store the agent decision rights and responsibilities management evidence storage report, so that a continuous state-based storage relationship is formed between the authorized state rights and responsibilities record, the evidence binding judgment record, the verification state rights and responsibilities record, and the agent decision rights and responsibilities management evidence storage report.

[0118] In this hardware architecture, the task access unit is responsible for external task input, the cache unit is responsible for temporary storage of data in stages, the processor is responsible for data parsing, matching, reconstruction and generation processing in each stage, the state machine storage unit is responsible for saving state records in the authorization, judgment and verification process, and the evidence storage unit is responsible for saving the final intelligent agent decision-making responsibility management evidence report.

[0119] Figures 2-4 For an exemplary description, please refer to the above. Figure 1 This will not be elaborated upon here.

Claims

1. A method for controlling the decision-making authority and responsibility of intelligent agents based on declarative primitive state machines, characterized in that, include: Step 1: Perform declarative primitive parsing on the agent's decision task request to generate authorization primitive instruction units, and generate authorization state responsibility records based on the authorization primitive instruction units; Step 2: Perform judgment access element parsing on the authorized state rights and responsibilities record to obtain the judgment access element description, and perform access matching between the judgment primitive instruction unit and the judgment access element description to generate evidence binding judgment record; Step 3: Reconstruct the verification elements of the evidence binding judgment record to obtain the verification comparison element description. Based on the verification comparison element description and the determined verification primitive instruction unit to be verified, generate the verification status responsibility record. Step 4: Reconstruct the final elements of the verification state rights and responsibilities record to generate a final constraint element description, and generate an agent decision-making rights and responsibilities management and evidence storage report based on the final constraint element description.

2. The method for controlling the decision-making authority and responsibility of an intelligent agent based on a declarative primitive state machine according to claim 1, characterized in that, Step 1 generates an authorization state responsibility record based on the authorization primitive instruction unit, which includes the following steps: The authorization primitive instruction unit is subjected to authorization trigger verification to obtain the authorization trigger verification result; Based on the authorization trigger verification result, the authorization elements of the authorization primitive instruction unit are extracted to obtain the authorization element description; The authorization element descriptions are encapsulated into authorization relationships to form a dynamic authorization mapping; Based on the authorization trigger verification result, the flow elements of the authorized primitive instruction unit are extracted to obtain the primitive flow order; Based on the primitive flow order and the dynamic authorization mapping, an authorization state responsibility record is generated.

3. The method for controlling the decision-making authority and responsibility of an intelligent agent based on a declarative primitive state machine according to claim 2, characterized in that, Based on the primitive flow order and the dynamic authorization mapping, an authorization state responsibility record is generated, including: The dynamic authorization mapping is registered with a status entry according to the primitive flow order to form a primitive flow credential; The dynamic authorization mapping, the primitive flow credential, and the authorization trigger verification result are processed by signature digest to generate a signature identity digest; The authorization primitive instruction unit, the authorization trigger verification result, the dynamic authorization mapping, the primitive flow credential, and the signature identity digest are written into the declarative primitive state machine to generate an authorization state responsibility record.

4. The method for controlling the decision-making authority and responsibility of an intelligent agent based on a declarative primitive state machine according to claim 1, characterized in that, Step 2: Parse the authorized state rights and responsibilities record for judgment access elements to obtain judgment access element descriptions. Perform access matching between the judgment primitive instruction unit to be verified and the judgment access element descriptions to generate evidence-binding judgment records, including: The decision primitive instruction unit is received based on the authorized state rights and responsibilities record, and is used as the decision primitive instruction unit to be verified; The primitive instruction unit to be verified is matched with the description of the judgment access element to obtain the judgment access verification result. The description of the judgment admission elements is parsed to obtain the judgment flow constraints. The judgment flow verification is performed on the judgment primitive instruction unit to be verified according to the judgment flow constraints to generate a judgment flow verification result; In response to the judgment access verification result and the judgment transfer verification result, the evidence elements of the judgment primitive instruction unit to be verified are extracted to determine the judgment evidence description; Based on the description of the evidence in the judgment, an evidence-binding judgment record is generated.

5. The method for controlling the decision-making authority and responsibility of an intelligent agent based on a declarative primitive state machine according to claim 4, characterized in that, Based on the aforementioned description of the judgment evidence, an evidence-bound judgment record is generated, including: The original evidence data is read according to the description of the judgment evidence to obtain the evidence data to be compared; The evidence description in the judgment is compared with the evidence data to be compared, and a source verification is performed to obtain the evidence source verification result. The judgment result is extracted from the primitive instruction unit to be verified to obtain the judgment result; The judgment access verification result, the judgment flow verification result, the judgment evidence description, the evidence source verification result, and the judgment result are subjected to evidence binding processing to generate an evidence binding judgment record.

6. The method for controlling the decision-making authority and responsibility of an intelligent agent based on a declarative primitive state machine according to claim 1, characterized in that, In step 3, based on the description of the verification comparison elements and the determined verification primitive instruction units to be verified, a verification state responsibility record is generated, including: The verification entry verification is performed on the verification primitive instruction unit to be verified according to the verification comparison element description, so as to obtain the verification entry verification result; The verification conclusion is extracted from the verification primitive instruction unit to be verified in order to obtain the verification conclusion; The consistency between the verification conclusion and the description of the verification comparison elements is verified to generate a verification status responsibility record.

7. The method for controlling the decision-making authority and responsibility of an intelligent agent based on a declarative primitive state machine according to claim 6, characterized in that, The consistency between the verification conclusion and the description of the verification comparison elements is verified to generate a verification status responsibility record, including: The consistency between the verification conclusion and the description of the verification comparison element is verified to generate a verification status determination result; Based on the verification results of the verification entry and the verification status determination results, the audit status of the evidence binding judgment record is rewritten to generate an audit judgment record; Based on the verification status determination result, the audit judgment record is configured to assume responsibility, so as to generate a responsibility assumption identifier; The responsibility assignment identifier is written into the audit decision record to obtain the verification status responsibility record.

8. The method for controlling the decision-making authority and responsibility of an intelligent agent based on a declarative primitive state machine according to claim 1, characterized in that, Step 4: Based on the description of the termination constraint elements, generate an agent decision-making responsibility control and evidence storage report, including: Based on the verification state responsibility record, the terminal primitive instruction unit is received to obtain the terminal primitive instruction unit to be verified; According to the description of the termination constraint elements, the termination entry verification is performed on the termination primitive instruction unit to be verified to obtain the termination entry verification result; Based on the termination entry verification result, the session boundary is located for the termination constraint element description to determine the authorization record to be terminated; The pending authorization records are marked as invalid to form authorization invalidation records; Based on the authorization expiration records, generate a report on the management and evidence preservation of the decision-making authority and responsibility of the intelligent agent.

9. The method for controlling the decision-making authority and responsibility of an intelligent agent based on a declarative primitive state machine according to claim 8, characterized in that, Based on the authorization expiration records, generate a report on the management and evidence preservation of the agent's decision-making authority and responsibility, including: Based on the authorization expiration record, the authorization record to be terminated is erased to obtain the authorization erasure record; Write the authorized erasure record into the verification state responsibility record to obtain the final state responsibility record; The terminal state rights and responsibilities records are stored and encapsulated to generate an agent decision-making rights and responsibilities management evidence report.

10. A decision-making and responsibility management system for intelligent agents based on declarative primitive state machines, characterized in that, The method includes an authorization state rights and responsibilities record generation unit, an evidence binding judgment record generation unit, a verification state rights and responsibilities record generation unit, and a rights and responsibilities management and evidence storage report generation unit configured on a processor; the authorization state rights and responsibilities record generation unit is used to perform step 1 of the method according to any one of claims 1-8; the evidence binding judgment record generation unit is used to perform step 2 of the method according to any one of claims 1-8; the verification state rights and responsibilities record generation unit is used to perform step 3 of the method according to any one of claims 1-8; and the rights and responsibilities management and evidence storage report generation unit is used to perform step 4 of the method according to any one of claims 1-8.

Images