v3.0.0

🌿 SME v3.0.0 "Odyssey" Release Notes
Date: February 24, 2026
Status: Major Release

🚀 Major Highlights
The Odyssey Migration: A comprehensive overhaul of the data and extension layers to improve performance and stability across the entire forensic lab.

Pydantic v2 Integration: Fully migrated to Pydantic v2 for faster data validation and better compatibility with the latest Python toolchains (up to 3.14).

Local OCR Microservice: Introduced sme-ocr-service, a dedicated local-first OCR layer for processing documents without external API reliance.

Unified Extension Manager: Completed the Base Plugin migration for all 18 core extensions, ensuring a standardized lifecycle for every forensic tool.

🛠️ Key Fixes & Improvements
Dependency Shielding: Updated waybackpy to ==3.0.6 to resolve versioning conflicts that were breaking certain ingestion pipelines.

Optimization for 1660 Ti: Refined memory management in the Gephi Bridge and Data Guard Auditor to ensure stable performance within the 6GB VRAM limit of the NVIDIA GeForce GTX 1660 Ti.

Control Room Enhancements: Updated the Glassmorphism dashboard with improved telemetry for CPU, RAM, and VRAM monitoring.

📦 Vendoring Note (Reminder)
As per your project requirements, the faststylometry library has been successfully vendored into the source tree. This bypasses broken PyPI dependency requirements and locks the version to your specific Python 3.10 setup.

📥 Getting Started
To update to the Odyssey release while bypassing local dependency issues, use the updated Docker blueprint:

Bash
curl -O https://raw.githubusercontent.com/SpectreDeath/Semantic-Memory-Engine/main/docker-compose.yaml
docker-compose up -d --build

v2.3.3

Full Changelog: v2.3.2...v2.3.3

SME v2.1.0: The Rhetorical Fingerprint (Lawnmower Man)

🌿 SME v2.1.0: The Rhetorical Fingerprint (Lawnmower Man)
This release shifts the Semantic-Memory-Engine from simple data ingestion to Forensic Stylometry. By integrating Microsoft's MarkItDown and formalizing archival diffs, SME can now track not just what was said, but how it was said and how it changed over time.

🧬 New Forensic Capabilities
Rhetorical Fingerprinting: Introduced active_persona.json mapping. The engine now identifies stylistic markers in ingested text using the vendored faststylometry suite, bypassing PyPI dependency issues.

Archival Diff Specification: Formalized the logic for tracking "Semantic Drift." This allows the engine to compare two versions of a document and highlight changes in intent or sentiment, not just text.

MarkItDown Integration: Switched to a more robust markdown conversion for web ingestion, ensuring high-fidelity "atomic facts" for the Knowledge Graph.

⚡ Hardware Optimizations (GTX 1660 Ti / 6GB VRAM)
Memory Footprint Audit: Verified that the auditor.py (Isolation Forest) maintains a <5MB RAM footprint.

VRAM Safeguards: The gephi_bridge.py now strictly enforces the 2000-node limit to prevent OOM crashes during large codebase visualizations.

Sidecar Isolation: The Dual-Python architecture is now standard, allowing the Logic Layer (Operator) to run on Python 3.10 while the Inference Layer (Sidecar) handles heavier ML tasks.

🧪 Validation Results
Master Forensic Test Suite: 100% Pass rate on all forensic utilities.

Container Health: Unified docker-compose now successfully orchestrates the Operator, Sidecar, and the Vite-based Control Room.

v2.0.0: The Control Room

Full Changelog: v1.3.0...v2.0.0

1.3.0

integrating Cosine Delta and KL Divergence