suriyasandy/accounting

Purpose and Scope

This repository is a multi-prototype workspace for an accounting-control and break-analysis system. It appears to cover ingestion of accounting data, cleansing and feature preparation, break/factor/LNA analysis, model training and prediction, and analyst-facing applications. Several ZIP archives and parallel application variants indicate active experimentation rather than a single clearly canonical deployment.

Primary Entry Points

Likely execution surfaces are the orchestration and UI directories:

• Breakanalyisi/Main

• Breakanalyisi/Cli

• Breakanalyisi/Streamlit app

• Accounting_fastapi.zip

• Accounting_streamlit.zip

Main and Cli likely support batch or command-driven workflows; the Streamlit and FastAPI artifacts suggest separate interactive and API-serving paths. No README content identifies a definitive production entry point.

Data Ingestion and Storage Boundary

The system’s first stage is accounting-data acquisition and normalization:

• Breakanalyisi/Clean data and read excel

• Breakanalyisi/Excel to parquet

• Local loader

• Faster data download

Excel is a material input format, with Parquet apparently serving as an optimized analytical representation. Loader/download components likely isolate local files or remote data retrieval from downstream analytical logic.

Core Analytical Pipeline

Break identification and accounting analysis appear to be the central domain workflow:

• Breakanalyisi/Data engine

• Breakanalyisi/Factor analysis

• Breakanalyisi/Lna analysis

• Breakanalyisi/Pipeljne

• Mi analyis

The data engine and pipeline likely coordinate prepared records through break, factor, and LNA analysis. “Break chaining” is additionally documented by cashfoare break chaining.pdf, suggesting relationships between breaks may be modeled as a separate analytical concern.

Modeling, Training, and Inference

Predictive capabilities are organized into distinct training and serving stages:

• Breakanalyisi/Modeling

• Breakanalyisi/Train

• Breakanalyisi/Prediction

• Train_model

• Live inference

• Updated predictor

The expected flow is normalized accounting data → feature selection/modeling → trained model artifacts → batch or live predictions. Feature selection, Query pred, and Record from prediction indicate prediction results may be filtered, queried, and persisted for analyst review.

Analyst Applications and Workflow Integrations

Multiple UI variants appear to expose the same or evolving analytical workflow:

• Breakanalyisi/P1app original

• Breakanalyisi/P1appnewversion

• Breakanalyisi/P1appwithjira

• Ui/Single prediction

• Ui/Query main

• Ui/Query output

These applications likely separate single-record scoring from query-driven investigation. The Jira-specific variant implies an external issue-management boundary for escalation or case tracking, though the integration implementation is not visible in the tree.

Configuration, Evaluation, and Governance

Operational controls and quality assessment are represented by:

• Breakanalyisi/Config

• Breakanalyisi/P1config

• Breakanalyisi/Evaluation

• Breakanalyisi/AI-Governance-Usecase.md

• Breakanalyisi/Technical-Workflow.md

Configuration is separated from analytical stages, while evaluation and governance materials suggest model validation, traceability, and control requirements are important. Breakanalyisi/Accounting_Control_AI_Gap_Analysis.docx indicates the project is positioned as an AI enhancement to accounting-control processes.

Repository Structure Risk

The repository contains many overlapping ZIPs, app generations, and named experiments, including Accounting_Control.zip, End-To-End.zip, and phase1 final.zip. Before extending it, identify the intended canonical pipeline and UI/API variant; otherwise changes may land in a deprecated prototype rather than the deployed architecture.