docs: add comprehensive Technologies Used section to README.md

Release Notes
This release improves project documentation by adding a dedicated Technologies Used section in README.md.

What’s included

• Added a categorized stack overview covering:

1. Language/runtime
2. Backend framework
3. Frontend technologies
4. Database
5. ML and NLP libraries
6. Model/runtime utilities
7. Testing and code quality tools
8. DevOps and deployment tooling

Why this helps

• Makes the project’s technical stack easier to understand at a glance
• Helps contributors onboard faster
• Improves clarity for reviewers, recruiters, and collaborators evaluating the repository

LedgerGuard v0.4.1 — Ops Feedback Loop, Rules Overlay, and Deployment Readiness

Release notes

This release marks the first complete, end-to-end version of LedgerGuard as a production-minded fraud detection and transaction categorisation system for a neo-bank use case. The focus of v0.4.1 is operational correctness, auditability, and deployment readiness rather than model novelty.

Key features

• End-to-end ML pipeline that auto-labels merchant transactions and assigns fraud risk scores
• Ops Dashboard for CSV ingestion, scoring, KPI review, filtering, and inspection of results
• Stable row identifiers enabling durable analyst feedback across filtering, ordering, and truncation
• Inline category editing with merged session persistence and exportable feedback data
• Rules overlay for deterministic business overrides with explicit audit tagging and clear precedence
• Read-only Customer Dashboard that exposes safe transaction views without Ops controls or model internals
• Performance page driven by a model registry, surfacing model metadata and metrics
• Lightweight /health/ endpoint and production toggles for deployment probes

Operational guarantees

• Explicit precedence model: analyst edit over rule override over model output
• Category provenance tracked per row for audit and downstream retraining
• Deterministic previews for large uploads with preserved totals and flagged prioritisation
• CI-backed test suite with coverage enforcement

Deployment readiness

• Containerised application with Gunicorn entrypoint
• Environment-driven configuration for debug, hosts, and static assets
• Healthcheck suitable for platform liveness checks
• Documented deployment steps and reproducible local and CI workflows

Known limitations

• Categorisation confidence calibration is not implemented
• Feedback exports are not yet wired into automated retraining
• Large result sets rely on preview limits rather than full pagination

What’s next
Future work will focus on persistence for edits and flags, automated retraining pipelines, monitoring for ingestion and scoring drift, and background job orchestration.

Maintainer
Adrian Adewunmi

LedgerGuard v0.4.0

Week 4 milestone.

• Durable feedback loop with stable identifiers
• Rules overlay with explicit precedence and audit tagging
• Read-only customer site with privacy safeguards
• Deployment-ready packaging with healthcheck
• CI stability and coverage enforcement