[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81759-en":3,"doc-seo-81759-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},81759,4398048949847,"Eliana","https://ap-avatar.wpscdn.com/avatar/400002536579ef2da7f?_k=1778318612642679267",8,"Research & Report","Mnemosyne: Agentic Transaction Processing for Validating and Repairing AI-generated Workflows","LLMs generate workflow actions, repairs, and plans, yet individual actions can be syntactically valid while stale, infeasible, conflicting, or destructive to the evidence that triggered repair. Agentic Transaction Processing (ATP) treats generated actions as untrusted proposals until deterministic admission under executable constraints. Mnemosyne operationalizes ATP via an append-only transition log, effective-state projection, dependency-safe compensation, and active commitment records. Safety properties cover authority separation, serial-equivalent admission, evidence-preserving repair, obligation containment, plus bounded reactive repair with localized edits.","Mnemosyne: Agentic Transaction Processing for Validating and Repairing AI-generated Workflows  \nEdward Y. Chang  \nStanford University  \nLongling Geng  \nStanford University  \nEmily J. Chang  \nQuadriumAI  \narXiv :2607 .00269v2 [ cs .AI ] 5 Jul 2026  \nAbstract  \nLLMs increasingly generate workflow actions, repairs, and plans, but a generated action may be syntactically valid yet stale, infeasible, conflicting, or destructive of the evidence that triggered a repair. We introduce Agentic Transaction Processing (ATP), a transaction model that treats generated actions as untrusted proposals until they pass deterministic admission under a declared, executable constraint set  \nC. The governing principle is two-sided: a proposal is not truth, and no proposal foresees every disruption. Anything may propose, but only the runtime admits and commits; when an unforeseen disruption strikes, it repairs reactively within bounds rather than trusting a fresh proposal. Relative to C , committedstate correctness becomes independent of the competence, honesty, or learning of the proposing layer. We realize ATP in Mnemosyne, a runtime with an appendonly transition log, effective-state projection, dependency-safe compensation, and active commitment records, and prove four safety properties relative to C (authority separation, serial-equivalent generative admission, evidence-preserving repair, and obligation containment) together with a bounded-reactive-repair guarantee for its localized repair protocol (LCRP) . A reproducible artifact rejects the targeted violations across nine falsification tests while still admitting valid work, at under 6% projection-and-validation overhead, and bounded local repair edits an order of magnitude fewer operations than global recompute. In bounded live-proposer pilots, 80 static plan-entry and mid-execution repair proposals from four heterogeneous LLMs pass through the same admission boundary and are scored by an external cross-episode harness with zero invalid commits; the gate admits 24 of  \n40 live repair proposals and rejects 16, four of them explicit safety rejections of unsafe or over-broad rollback. Mnemosyne is open source: [https://github](https://github) .  \ncom/eyuchang/Mnemosyne/tree/mnemosyne-atp-postgres-rerun.  \nKeywords: Agentic transaction processing, LLM agents, Transaction processing, Workflow validation and repair, Data integrity  \n1 Introduction  \nDatabase systems have long separated logical correctness from physical execution through transactions, recovery, isolation, logging, and integrity constraints Gray and Reuter [1993], Bernstein and Newcomer [2009], Weikum and Vossen [2001], Mohan et al. [1992] . Workflow systems extend these ideas to long-running activities with retries, timeouts, idempotency, and compensation GarciaMolina and Salem [1987], Ludscher et al. [2006], Russell et al. [2005] . Both families, however, usually assume that the submitted unit of work is meaningful enough to be treated as a transaction request: it may conflict, fail, or abort, but it is not itself a hallucinated, stale, or semantically invalid proposal produced by an untrusted reasoning process.  \nAgentic workflows violate this assumption. The candidate transaction is generated by an LLM, often embedded in a multi-agent planning system that already adds validation and transaction-style  \nPreprint.  \nguarantees Chang and Geng [2025], Yao et al. [2023b,a], Shinn et al. [2023], Wang et al. [2023a], Park et al. [2023], Wang et al. [2023b], Chang [2025] . Yet a well-formed proposal can still be wrong in ways a syntactic check accepts.  \nConsider a concrete case from multi-agent planning Chang and Geng [2025] . In a family’s Thanksgiving plan, the father lands in Boston and picks up Grandma on the way home so dinner can start on time. His flight is delayed, so an LLM agent reactively replans the afternoon’s pickups and drives. The new schedule parses as valid, yet two faults slip through. It reuses the off-peak drive time ","cbCaibnol3z7GjkF","https://ap.wps.com/l/cbCaibnol3z7GjkF","pdf",462517,2,1,41,"English","en",105,"# Introduction\n## Problem: Syntactic validity vs semantic disruption\n## Agentic Transaction Processing (ATP) and deterministic admission\n## Bounded local repair and evidence-preserving commitment","[{\"question\":\"What is the core idea of Agentic Transaction Processing (ATP)?\",\"answer\":\"ATP treats LLM-generated workflow actions as untrusted proposals until they pass deterministic admission against a declared, executable constraint set on the current effective state.\"},{\"question\":\"How does Mnemosyne ensure safety when a proposal is wrong but syntactically valid?\",\"answer\":\"Mnemosyne requires runtime admission to reject infeasible actions and prevents committed obligations from being silently dropped, recording rejection reasons when constraints are violated.\"},{\"question\":\"What happens when an unforeseen disruption occurs during execution?\",\"answer\":\"Mnemosyne triggers a bounded local repair using the localized cascading repair protocol (LCRP), then routes the repair back through the same admission gate so incorrect repairs are caught while the blast radius remains contained.\"}]",1784175867,103,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"mnemosyne-agentic-transaction-processing-for-validating-and-repairing-ai-generated-workflows","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/mnemosyne-agentic-transaction-processing-for-validating-and-repairing-ai-generated-workflows/81759/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-26","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What is the core idea of Agentic Transaction Processing (ATP)?","Question",{"text":75,"@type":76},"ATP treats LLM-generated workflow actions as untrusted proposals until they pass deterministic admission against a declared, executable constraint set on the current effective state.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does Mnemosyne ensure safety when a proposal is wrong but syntactically valid?",{"text":80,"@type":76},"Mnemosyne requires runtime admission to reject infeasible actions and prevents committed obligations from being silently dropped, recording rejection reasons when constraints are violated.",{"name":82,"@type":73,"acceptedAnswer":83},"What happens when an unforeseen disruption occurs during execution?",{"text":84,"@type":76},"Mnemosyne triggers a bounded local repair using the localized cascading repair protocol (LCRP), then routes the repair back through the same admission gate so incorrect repairs are caught while the blast radius remains contained.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]