# karitselmuthu/enterprise_ticket_transfomer architecture

> Architecture diagram of the GitHub repository karitselmuthu/enterprise_ticket_transfomer, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-07). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/karitselmuthu/enterprise_ticket_transfomer
- Repository: https://github.com/karitselmuthu/enterprise_ticket_transfomer

## Overview

This is an educational ticket-classification project spanning V1–V7 training and evaluation, plus a V8 FastAPI inference service. The main workflow reads labeled tickets, fixes a reproducible split, trains a selected model generation, evaluates it, and loads an explicitly selected artifact for prediction. The graph preserves the distinct baseline, embedding, recurrent, Transformer, and pretrained-model stages and the production promotion check. Relationships for unsampled modules are included only where the README or SOURCE INDEX supports them; the source excerpts do not establish every internal training detail.

Main entry points: [`src/training/split.py`](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/split.py) (Locked split), [`src/inference/runtime.py`](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/inference/runtime.py) (Artifact runtime), [`src/training/train_neural.py`](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/train_neural.py) (V3–V5 trainer), [`api/main.py`](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/api/main.py) (Prediction API)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_data["Data preparation"]
  node_tickets["Ticket CSV<br/>[tickets.csv]"]
  node_split["Locked split<br/>[split.py]"]
  node_tokenizer["Tokenizer<br/>[tokenizer.py]"]
  node_vocabulary["Neural vocabulary<br/>[vocabulary.py]"]
end

subgraph group_models["Model training"]
  node_v1train["V1 trainer<br/>[train.py]"]
  node_v1model["Naive Bayes<br/>[traditional.py]"]
  node_v2train["V2 trainer<br/>[train_v2.py]"]
  node_v2model["Word2Vec classifier<br/>[word2vec.py]"]
  node_neuraltrain["V3–V5 trainer<br/>[train_neural.py]"]
  node_pretrainedprep["Checkpoint preparation"]
  node_pretrainedtrain["V6–V7 trainer"]
end

subgraph group_assess["Evaluation and promotion"]
  node_evaluation["Model evaluation<br/>[evaluate.py]"]
  node_audit["Data audit<br/>[data_audit.py]"]
  node_compare["Generation comparison<br/>[compare_all.py]"]
  node_promotion["Promotion validation<br/>[promotion.py]"]
end

subgraph group_serve["Inference service"]
  node_pretrainedinfer["Pretrained inference<br/>[pretrained.py]"]
  node_runtime["Artifact runtime<br/>[runtime.py]"]
  node_api["Prediction API<br/>[main.py]"]
end

subgraph group_internals["Model internals"]
  node_embeddings["Skip-gram vectors<br/>[word2vec.py]"]
  node_lstm["LSTM models<br/>[lstm.py]"]
  node_transformer["Transformer classifier<br/>[model.py]"]
  node_attention["Self-attention<br/>[attention.py]"]
  node_position["Positional encoding"]
end

node_operator(("Engineer"))

node_operator -->|"provides data"| node_tickets
node_operator -->|"creates split"| node_split
node_tickets -->|"partitions rows"| node_split
node_split -->|"supplies manifest"| node_v1train
node_tickets -->|"loads rows"| node_v1train
node_v1train -->|"fits classifier"| node_v1model
node_v1model -->|"tokenizes text"| node_tokenizer
node_operator -->|"selects generation"| node_v2train
node_split -->|"supplies manifest"| node_v2train
node_v2train -->|"trains classifier"| node_v2model
node_v2model -->|"trains vectors"| node_embeddings
node_embeddings -->|"tokenizes text"| node_tokenizer
node_operator -->|"selects generation"| node_neuraltrain
node_split -->|"loads split"| node_neuraltrain
node_neuraltrain -->|"builds vocabulary"| node_vocabulary
node_vocabulary -->|"tokenizes text"| node_tokenizer
node_neuraltrain -->|"constructs V3–V4"| node_lstm
node_neuraltrain -->|"constructs V5"| node_transformer
node_transformer -->|"uses attention"| node_attention
node_transformer -->|"adds positions"| node_position
node_operator -->|"prepares checkpoint"| node_pretrainedprep
node_pretrainedprep -->|"provides local weights"| node_pretrainedtrain
node_operator -->|"selects V6–V7"| node_pretrainedtrain
node_split -->|"supplies manifest"| node_pretrainedtrain
node_operator -->|"requests evaluation"| node_evaluation
node_evaluation -->|"checks fingerprint"| node_split
node_operator -->|"audits data"| node_audit
node_audit -->|"uses split logic"| node_split
node_operator -->|"compares models"| node_compare
node_api -->|"loads predictor"| node_runtime
node_api -->|"validates production"| node_promotion
node_runtime -->|"dispatches V1"| node_v1model
node_runtime -->|"dispatches V2"| node_v2model
node_runtime -->|"constructs neural model"| node_neuraltrain
node_runtime -->|"encodes input"| node_vocabulary
node_runtime -->|"dispatches V6–V7"| node_pretrainedinfer
node_operator -->|"submits ticket"| node_api
node_api -->|"predicts text"| node_runtime

click node_tickets "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/data/samples/tickets.csv"
click node_split "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/split.py"
click node_tokenizer "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/preprocessing/tokenizer.py"
click node_vocabulary "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/preprocessing/vocabulary.py"
click node_v1train "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/train.py"
click node_v1model "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/baselines/traditional.py"
click node_v2train "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/train_v2.py"
click node_v2model "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/baselines/word2vec.py"
click node_embeddings "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/embeddings/word2vec.py"
click node_neuraltrain "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/train_neural.py"
click node_lstm "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/baselines/lstm.py"
click node_transformer "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/transformer/model.py"
click node_attention "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/transformer/attention.py"
click node_position "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/transformer/positional_encoding.py"
click node_pretrainedprep "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/prepare_pretrained.py"
click node_pretrainedtrain "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/train_pretrained.py"
click node_pretrainedinfer "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/inference/pretrained.py"
click node_evaluation "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/evaluation/evaluate.py"
click node_audit "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/evaluation/data_audit.py"
click node_compare "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/evaluation/compare_all.py"
click node_runtime "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/inference/runtime.py"
click node_promotion "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/inference/promotion.py"
click node_api "https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/api/main.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_tickets,node_split,node_tokenizer,node_vocabulary toneBlue
class node_v1train,node_v1model,node_v2train,node_v2model,node_neuraltrain,node_pretrainedprep,node_pretrainedtrain toneAmber
class node_evaluation,node_audit,node_compare,node_promotion toneMint
class node_pretrainedinfer,node_runtime,node_api toneRose
class node_embeddings,node_lstm,node_transformer,node_attention,node_position,node_operator toneIndigo
```

## Components

### Data preparation

- [Ticket CSV](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/data/samples/tickets.csv): `data/samples/tickets.csv`
- [Locked split](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/split.py): `src/training/split.py`
- [Tokenizer](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/preprocessing/tokenizer.py): `src/preprocessing/tokenizer.py`
- [Neural vocabulary](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/preprocessing/vocabulary.py): `src/preprocessing/vocabulary.py`

### Model training

- [V1 trainer](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/train.py): `src/training/train.py`
- [Naive Bayes](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/baselines/traditional.py): `src/baselines/traditional.py`
- [V2 trainer](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/train_v2.py): `src/training/train_v2.py`
- [Word2Vec classifier](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/baselines/word2vec.py): `src/baselines/word2vec.py`
- [V3–V5 trainer](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/train_neural.py): `src/training/train_neural.py`
- [Checkpoint preparation](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/prepare_pretrained.py): `src/training/prepare_pretrained.py`
- [V6–V7 trainer](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/training/train_pretrained.py): `src/training/train_pretrained.py`

### Evaluation and promotion

- [Model evaluation](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/evaluation/evaluate.py): `src/evaluation/evaluate.py`
- [Data audit](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/evaluation/data_audit.py): `src/evaluation/data_audit.py`
- [Generation comparison](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/evaluation/compare_all.py): `src/evaluation/compare_all.py`
- [Promotion validation](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/inference/promotion.py): `src/inference/promotion.py`

### Inference service

- [Pretrained inference](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/inference/pretrained.py): `src/inference/pretrained.py`
- [Artifact runtime](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/inference/runtime.py): `src/inference/runtime.py`
- [Prediction API](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/api/main.py): `api/main.py`

### Model internals

- [Skip-gram vectors](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/embeddings/word2vec.py): `src/embeddings/word2vec.py`
- [LSTM models](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/baselines/lstm.py): `src/baselines/lstm.py`
- [Transformer classifier](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/transformer/model.py): `src/transformer/model.py`
- [Self-attention](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/transformer/attention.py): `src/transformer/attention.py`
- [Positional encoding](https://github.com/karitselmuthu/enterprise_ticket_transfomer/blob/main/src/transformer/positional_encoding.py): `src/transformer/positional_encoding.py`

### Other components

- **Engineer**

## Connections

- Engineer → Ticket CSV: provides data (evidence: `README.md`)
- Engineer → Locked split: creates split (evidence: `README.md`)
- Ticket CSV → Locked split: partitions rows (evidence: `src/training/split.py`)
- Locked split → V1 trainer: supplies manifest (evidence: `src/training/train.py`)
- Ticket CSV → V1 trainer: loads rows (evidence: `src/training/train.py`)
- V1 trainer → Naive Bayes: fits classifier (evidence: `src/training/train.py`)
- Naive Bayes → Tokenizer: tokenizes text (evidence: `src/baselines/traditional.py`)
- Engineer → V2 trainer: selects generation (evidence: `README.md`)
- Locked split → V2 trainer: supplies manifest (evidence: `src/training/train_v2.py`)
- V2 trainer → Word2Vec classifier: trains classifier (evidence: `src/training/train_v2.py`)
- Word2Vec classifier → Skip-gram vectors: trains vectors (evidence: `src/baselines/word2vec.py`)
- Skip-gram vectors → Tokenizer: tokenizes text (evidence: `src/embeddings/word2vec.py`)
- Engineer → V3–V5 trainer: selects generation (evidence: `README.md`)
- Locked split → V3–V5 trainer: loads split (evidence: `src/training/train_neural.py`)
- V3–V5 trainer → Neural vocabulary: builds vocabulary (evidence: `src/training/train_neural.py`)
- Neural vocabulary → Tokenizer: tokenizes text (evidence: `src/preprocessing/vocabulary.py`)
- V3–V5 trainer → LSTM models: constructs V3–V4 (evidence: `src/training/train_neural.py`)
- V3–V5 trainer → Transformer classifier: constructs V5 (evidence: `src/training/train_neural.py`)
- Transformer classifier → Self-attention: uses attention (evidence: `src/transformer/model.py`)
- Transformer classifier → Positional encoding: adds positions (evidence: `src/transformer/model.py`)
- Engineer → Checkpoint preparation: prepares checkpoint (evidence: `README.md`)
- Checkpoint preparation → V6–V7 trainer: provides local weights (evidence: `README.md`)
- Engineer → V6–V7 trainer: selects V6–V7 (evidence: `README.md`)
- Locked split → V6–V7 trainer: supplies manifest (evidence: `src/training/train_pretrained.py`)
- Engineer → Model evaluation: requests evaluation (evidence: `README.md`)
- Model evaluation → Locked split: checks fingerprint (evidence: `src/evaluation/evaluate.py`)
- Engineer → Data audit: audits data (evidence: `README.md`)
- Data audit → Locked split: uses split logic (evidence: `src/evaluation/data_audit.py`)
- Engineer → Generation comparison: compares models (evidence: `README.md`)
- Prediction API → Artifact runtime: loads predictor (evidence: `api/main.py`)
- Prediction API → Promotion validation: validates production (evidence: `api/main.py`)
- Artifact runtime → Naive Bayes: dispatches V1 (evidence: `src/inference/runtime.py`)
- Artifact runtime → Word2Vec classifier: dispatches V2 (evidence: `src/inference/runtime.py`)
- Artifact runtime → V3–V5 trainer: constructs neural model (evidence: `src/inference/runtime.py`)
- Artifact runtime → Neural vocabulary: encodes input (evidence: `src/inference/runtime.py`)
- Artifact runtime → Pretrained inference: dispatches V6–V7 (evidence: `src/inference/runtime.py`)
- Engineer → Prediction API: submits ticket (evidence: `README.md`)
- Prediction API → Artifact runtime: predicts text (evidence: `api/main.py`)
