Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ESAA: Event Sourcing for Autonomous Agents in LLM-Based Software Engineering

Elzo Brito dos Santos Filho

TL;DR

The ESAA (Event Sourcing for Autonomous Agents) architecture is presented, which separates the agent's cognitive intention from the project's state mutation, inspired by the Event Sourcing pattern, ensuring the immutability of completed tasks and forensic traceability.

Abstract

Autonomous agents based on Large Language Models (LLMs) have evolved from reactive assistants to systems capable of planning, executing actions via tools, and iterating over environment observations. However, they remain vulnerable to structural limitations: lack of native state, context degradation over long horizons, and the gap between probabilistic generation and deterministic execution requirements. This paper presents the ESAA (Event Sourcing for Autonomous Agents) architecture, which separates the agent's cognitive intention from the project's state mutation, inspired by the Event Sourcing pattern. In ESAA, agents emit only structured intentions in validated JSON (agent.result or issue.report); a deterministic orchestrator validates, persists events in an append-only log (activity.jsonl), applies file-writing effects, and projects a verifiable materialized view (roadmap.json). The proposal incorporates boundary contracts (AGENT_CONTRACT.yaml), metaprompting profiles (PARCER), and replay verification with hashing (esaa verify), ensuring the immutability of completed tasks and forensic traceability. Two case studies validate the architecture: (i) a landing page project (9 tasks, 49 events, single-agent composition) and (ii) a clinical dashboard system (50 tasks, 86 events, 4 concurrent agents across 8 phases), both concluding with run.status=success and verify_status=ok. The multi-agent case study demonstrates real concurrent orchestration with heterogeneous LLMs (Claude Sonnet 4.6, Codex GPT-5, Antigravity/Gemini 3 Pro, and Claude Opus 4.6), providing empirical evidence of the architecture's scalability beyond single-agent scenarios.

ESAA: Event Sourcing for Autonomous Agents in LLM-Based Software Engineering

TL;DR

The ESAA (Event Sourcing for Autonomous Agents) architecture is presented, which separates the agent's cognitive intention from the project's state mutation, inspired by the Event Sourcing pattern, ensuring the immutability of completed tasks and forensic traceability.

Abstract

Autonomous agents based on Large Language Models (LLMs) have evolved from reactive assistants to systems capable of planning, executing actions via tools, and iterating over environment observations. However, they remain vulnerable to structural limitations: lack of native state, context degradation over long horizons, and the gap between probabilistic generation and deterministic execution requirements. This paper presents the ESAA (Event Sourcing for Autonomous Agents) architecture, which separates the agent's cognitive intention from the project's state mutation, inspired by the Event Sourcing pattern. In ESAA, agents emit only structured intentions in validated JSON (agent.result or issue.report); a deterministic orchestrator validates, persists events in an append-only log (activity.jsonl), applies file-writing effects, and projects a verifiable materialized view (roadmap.json). The proposal incorporates boundary contracts (AGENT_CONTRACT.yaml), metaprompting profiles (PARCER), and replay verification with hashing (esaa verify), ensuring the immutability of completed tasks and forensic traceability. Two case studies validate the architecture: (i) a landing page project (9 tasks, 49 events, single-agent composition) and (ii) a clinical dashboard system (50 tasks, 86 events, 4 concurrent agents across 8 phases), both concluding with run.status=success and verify_status=ok. The multi-agent case study demonstrates real concurrent orchestration with heterogeneous LLMs (Claude Sonnet 4.6, Codex GPT-5, Antigravity/Gemini 3 Pro, and Claude Opus 4.6), providing empirical evidence of the architecture's scalability beyond single-agent scenarios.
Paper Structure (38 sections, 1 figure, 4 tables)

This paper contains 38 sections, 1 figure, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Tool-Using Agents and Interleaved Reasoning
  4. Multi-Agent Frameworks
  5. Context Limits and Long-Horizon Degradation
  6. Event Sourcing and CQRS
  7. Constrained Output and Syntactic Structure
  8. ESAA Architecture
  9. Canonical Artifacts
  10. Trace-First Model and Immutability
  11. Determinism through Canonicalization and Hashing
  12. Multi-Agent Dispatch and Concurrency
  13. Methodology
  14. Architectural Intervention Design
  15. Orchestrator Implementation and Validation Pipeline
  16. ...and 23 more sections

Figures (1)

  • Figure 1: ESAA Orchestration Cycle: the agent emits structured intentions validated by boundary contracts and JSON Schema; the orchestrator persists events in the append-only store and projects the read-model. The agent receives only a purified view, never the raw state.