Releases: UndiFineD/PyAgent
v3.7.0
v3.6.0
PyAgent v3.6.0 (Phase 317 - "Synaptic Acceleration") focuses on the transition from a monolithic core to a high-performance modular architecture.
Release Message: PyAgent v3.6.0
Title: 🚀 v3.6.0: Synaptic Acceleration - Modular Core & Complexity Guard
Summary:
This release marks a major architectural milestone in the PyAgent fleet. By transitioning to a decentralized Mixin-based architecture, we have achieved a 90% reduction in core complexity while restoring critical observability layers. Version 3.6.0 establishes a "debt-free" baseline for autonomous self-improvement.
Key Changes:
Synaptic Modularization: Refactored the monolithic BaseAgent (Complexity: 135) into five domain-specific Mixins: Identity, Persistence, Knowledge, Orchestration, and Governance.
Observability Restoration: Resolved circular dependencies in the stats module, restoring real-time nanosecond-precision telemetry.
Infrastructure Hardening: Patched the SelfImprovementOrchestrator to stabilize fleet-wide scans following the GitHub Copilot CLI deprecation.
Git Synchronization: Unified 8 legacy restoration branches into a clean, synchronized main lineage.
Phase 317 Complexity Guard: Successfully scanned 1,046 files to identify and queue 24 high-complexity debt clusters for autonomous refactoring.
Technical Metadata:
Phase: 317 (Synaptic Acceleration)
Stability Score: 1.000
Primary Focus: Logic Complexity & Type Safety
Core Seal: GOLDEN_MASTER_SEAL (v3.6.0)
v3.5.1
PyAgent v3.5.1 | Release Notes
"Structural Integrity & Type Safety Phase"
Summary
This release focuses on eliminating structural debt in the observability and logic layers while establishing a rigorous type-checking baseline across the project.
Key Improvements
Massive Type Optimization: Integrated 352 Mypy fixes using advanced automation (fix_mypy_issues_v4.py).
Structural Repair: Fixed broken relative imports in 17 core modules (specifically targeted incorrectly auto-extracted classes in observability and logic).
Infrastructure Stability: Created the missing init.py to restore corrupted package paths that were blocking the pytest runner.
Rust Interoperability: Resolved rust_core redefinition conflicts by implementing type-safe fallback mechanisms for optimized modules.
Consistency: Standardized type-hint imports (typing.Any, pathlib.Path) across the entire SDK.
Technical Stats
Evolution Phase: 316
Stability Score: 1.000 (Golden Master Seal Active)
Test Result: 1,456 Tests Passed (100% success rate on current suite).
Compatibility: Fully backward compatible with version 3.5.0.
Commit Hash
fc77b55: Refactor: Integrated 350+ Mypy fixes, resolved broken relative imports, and bumped version to 3.5.1
PyAgent v3.4.2 - Concurrency & Discovery
PyAgent v3.4.2 - Concurrency & Discovery
Core Evolution: Phase 241 & 242
Dynamic Capability Discovery: Implemented get_capabilities() across the agent hierarchy. Agents now auto-register their specialized skills (e.g., bash, security-audit, python) with the SignalRegistry upon initialization, enabling intelligent tool routing.
Signal Registry Hub: Upgraded the SignalRegistry to act as a real-time discovery service and event bus, maintaining a global index of active agent capabilities.
Security Hardening: Mitigated RUSTSEC-2025-0020 by upgrading the Rust core to PyO3 v0.24.1 and refactoring the bridge to use the modern, thread-safe Bound API.
Legal Compliance Lifecycle: Integrated the LegalAuditAgent into the AsyncFleetManager pre-commit hooks, ensuring all parallel agent outputs are scanned for license violations (GPL/AGPL) and liability risks.
Privacy Impact Assessment (PIA): Added automated PIA scoring to ComplianceAgent.py to evaluate data retention, encryption, and PII exposure risks.
Infrastructure Initialization: Established the Consciousness Registry to index agent thought streams and laid the groundwork for the Distributed Lock Manager to ensure resource safety in parallel workflows.
Status: Stable. Optimized for high-concurrency swarm operations and automated discovery.
PyAgent v3.2.0 "Proxima"
PyAgent v3.2.0 "Proxima" Release Summary
Version 3.2.0 marks a major architectural milestone in the PROXIMA Evolution, focusing on Logic Extraction, Infrastructure Unification, and Rust-Readiness. This release transitions the fleet from a monolith-manager structure into a high-performance "Core/Shell" architecture.
🏗️ 1. Core Architectural Shift (Rust Readiness)
The primary objective of v3.2.0 was the isolation of pure logic into standalone "Cores." These modules are 100% typed and side-effect-free, audited for immediate conversion to Rust (via PyO3) to enable low-latency execution.
New Cores: GPUMonitorCore, LoggingCore, PluginManager, ShardedKnowledgeCore, and StabilityCore.
Dependency De-coupling: Implemented lazy-loaded manager resolution to break long-standing circular import chains between BaseAgent and the fleet managers.
📊 2. High-Velocity Data & Serialization
Binary Communication: Integrated CBOR (Concise Binary Object Representation) with Zlib/Zstd compression for intra-agent communication, significantly reducing network overhead compared to JSON.
Knowledge Sharding: Upgraded the knowledge store to 1024-bucket sharding using Adler-32 stable hashing.
Data Ingest: Added Apache Parquet export support to ShardedKnowledgeCore for efficient LLM training and large-scale data offloading.
🔋 3. Infrastructure & Resilience
GPUMonitorCore: Introduced pure logic for real-time VRAM pressure calculation, thermal throttling detection, and optimal GPU workload routing.
Unified Fleet Harness: A new CLI entry point (FleetHarness.py) provides a consistent interface for heal, restore, and improve operations across the entire swarm.
Circuit Breakers: Enhanced resilience with "Wait-for-Success" logic in the HALF_OPEN state, preventing rapid flap transitions during system instability.
🔌 4. Plugin Lifecycle Stabilization
Lifecycle Hooks: AgentPluginBase now mandates health_check() and shutdown() methods, ensuring third-party extensions do not create zombie processes or resource leaks.
Strict Manifests: Enforced a new strictly-typed manifest.json format for the plugins directory, including enforced SDK version compatibility mapping.
🔍 5. Observability & Security
Sensitive Data Masking: The StructuredLogger now features an automated redaction engine for API keys (OpenAI, GitHub) and Bearer tokens.
High-Throughput Logs: Extracted LoggingCore to handle RFC3339 timestamping and masking logic at the performance-critical level.
📈 Evolution Status
Current Phase: 228 (Swarm Self-Documentation).
Stability Score: 0.997.
Git Commit: 6663d4f (main).
v3.0.0
PyAgent 3.0.0 "The Core Realization" — Release Announcement
We are proud to announce the release of PyAgent 3.0.0, representing a monumental shift from a class-based framework to a fully tiered, modular ecosystem for autonomous agent fleets.
🚀 Key Highlights
🏛️ Modular Tiered Architecture
The codebase has been completely reorganized into five distinct layers to improve scalability and maintainability:
Core: Stable versioning and SDK foundations.
Logic: The "brains" — divided into Cognitive, Development, Intelligence, Security, Swarm, and System agents.
Infrastructure: Fleet management, shard mapping, secret orchestration, and multi-cloud scaling.
Interface: Unified GUI (PyAgent-GUI), Web (Topology Viewer), and Mobile scaffolding.
Observability: Real-time stats, error-impact analysis, and automated improvement loops.
🧠 Advanced Cognitive Expansion
Latent Reasoning Protocol: Implementation of world-model-based reasoning for complex task planning.
Knowledge Trinity: A new synergy between memory consolidation, holographic context, and relational graph knowledge.
Theory of Mind (ToM): Specialized agents can now model the state and intent of other agents in the swarm.
🛡️ Security & Resilience
Byzantine Consensus: Enhanced stability for decentralized decision-making.
Privacy Guard: Real-time data privacy enforcement and compliance auditing.
Self-Healing Fleet: Autonomous fleet deployment and infrastructure repair agents are now standard.
📈 Infrastructure & Scaling
Quantum Sharding: New ShardManager for massive parallelization of knowledge and tasks.
GPU Scaling: Dynamic resource allocation for heavy model inference.
MCP Native: Deep integration with the Model Context Protocol for tool discovery.
🛠️ Internal Rebirth
Test Suite 2.0: Reorganized into Adversarial, Specialist, Performance, and Phase-specific validation.
Zero-Bloat History: Cleaned repository history to exclude massive index files while maintaining full auditing of the 3.0.0 transition.
SDK Stability: Standardized on SDK version 3.0.0 with a verified Stability Score of 1.0.
v2.1.4
"v2.1.4-stable: Global Security Hardening & Logic Isolation" --notes "Phase 115: Global mitigation for decompression bombs and redirect vulnerabilities. Upgraded urllib3 to 2.6.3. Isolated logic in MemoryConsolidator, MetacognitiveMonitor, TheoryOfMind, and ContextCompressor to support future Rust porting. Evaluated and fixed Pylance diagnostics in AST formula evaluation."
2.0.0
version 1.0.0
version 1.0.0
after making it more modular, splitting out classes and adding a gui,
we finally have our first major release.
the gui needs a lot of work and the agents need more testing,
but the whole should already become very capable.
the work of a local own designed rag llm is also in progress.