todo_registry.md

TODO Registry — Implementation Checklist

97 TODOs across 26 files — ✅ ALL IMPLEMENTED

---

src/preprocessing/ — 16 TODOs ✅

spell_corrector.py

Line	TODO	Status
36	Implement initialisation (SpellChecker + LanguageTool)	✅ DONE
41	Implement phonetic pass (regex substitution from DYSLEXIC_PHONETIC_MAP)	✅ DONE
46	Implement spellcheck pass (pyspellchecker token-level)	✅ DONE
51	Implement LanguageTool pass (context-aware, reverse-offset correction)	✅ DONE
56	Implement full correction pipeline (chain all 3 passes)	✅ DONE
61	Implement cleanup (self.tool.close())	✅ DONE

sentence_segmenter.py

Line	TODO	Status
15	Implement initialisation (load spaCy model)	✅ DONE
20	Implement sentence segmentation	✅ DONE

dependency_parser.py

Line	TODO	Status
16	Implement initialisation	✅ DONE
21	Implement dependency parsing	✅ DONE
26	Implement SVO extraction	✅ DONE

ner_tagger.py

Line	TODO	Status
24	Implement initialisation	✅ DONE
29	Implement NER tagging	✅ DONE
34	Implement protected span extraction	✅ DONE

dyslexia_simulator.py

Line	TODO	Status
35	Implement initialisation (set error_rate, seed)	✅ DONE
40	Implement letter transposition	✅ DONE
45	Implement letter omission	✅ DONE
50	Implement letter doubling	✅ DONE
55	Implement letter reversal (b/d, p/q)	✅ DONE
60	Implement word corruption (random error selection)	✅ DONE
65	Implement full simulation (corrupt + word merge)	✅ DONE

pipeline.py

Line	TODO	Status
38	Implement initialisation (load spaCy + spell corrector)	✅ DONE
43	Implement readability extraction (Flesch-Kincaid, Gunning Fog, SMOG, ARI)	✅ DONE
48	Implement dependency tree extraction (SVO per sentence)	✅ DONE
53	Implement full pipeline (7-step: spell→parse→segment→NER→deps→POS→readability)	✅ DONE

---

src/style/ — 14 TODOs ✅

fingerprinter.py

Line	TODO	Status
64	Implement MLP layers (Linear→LayerNorm→GELU→Dropout→Linear→LayerNorm)	✅ DONE
68	Implement forward pass (MLP projection)	✅ DONE
76	Implement initialisation (spaCy + AWL + projection MLP)	✅ DONE
81	Implement AWL loading from file	✅ DONE
86	Implement passive voice detection (nsubjpass/auxpass dep labels)	✅ DONE
91	Implement avg dependency tree depth	✅ DONE
96	Implement lexical density (content words / total)	✅ DONE
101	Implement raw feature extraction (~40 features)	✅ DONE
106	Implement vector extraction (raw features → pad/truncate to 40 → MLP → 512-dim)	✅ DONE
120	Implement vector blending with L2 normalisation	✅ DONE

formality_classifier.py

Line	TODO	Status
14	Implement initialisation	✅ DONE
19	Implement formality scoring (0-1 scale)	✅ DONE

emotion_classifier.py

Line	TODO	Status
14	Implement initialisation	✅ DONE
19	Implement emotion classification (distribution over register categories)	✅ DONE

style_vector.py

Line	TODO	Status
12	Implement cosine similarity	✅ DONE
18	Implement vector averaging	✅ DONE
24	Implement save to disk	✅ DONE
30	Implement load from disk	✅ DONE

---

src/model/ — 5 TODOs ✅

base_model.py

Line	TODO	Status
39	Implement model loading (tokenizer + model + quantization + LoRA wrapping)	✅ DONE

lora_adapter.py

Line	TODO	Status
20	Implement LoRA config creation	✅ DONE
26	Implement LoRA application to model	✅ DONE
32	Implement weight merging for inference	✅ DONE

style_conditioner.py

Line	TODO	Status
27	Implement projection layers (Linear → Tanh)	✅ DONE
37	Implement forward pass (project + reshape)	✅ DONE
53	Implement prefix prepending (torch.cat along seq dim)	✅ DONE

generation_utils.py

Line	TODO	Status
20	Implement generation with beam search	✅ DONE
30	Implement batch generation	✅ DONE

---

src/training/ — 22 TODOs ✅

dataset.py

Line	TODO	Status
54	Implement initialisation and data loading	✅ DONE
59	Implement JSONL loading	✅ DONE
64	Implement synthetic data augmentation	✅ DONE
68	Implement __len__	✅ DONE
73	Implement __getitem__	✅ DONE

loss_functions.py

Line	TODO	Status
34	Implement V1 initialisation	✅ DONE
43	Implement style loss (1 - cosine_similarity)	✅ DONE
52	Implement semantic loss	✅ DONE
65	Implement combined loss V1	✅ DONE
82	Implement V2 initialisation with frozen classifier	✅ DONE
87	Implement human pattern loss (1 - human_score)	✅ DONE
100	Implement combined loss V2	✅ DONE

trainer.py

Line	TODO	Status
17	Store loss function, fingerprinter, and tokenizer	✅ DONE
22	Implement custom compute_loss	✅ DONE

callbacks.py

Line	TODO	Status
14	Implement evaluation-time style metric logging	✅ DONE
22	Implement early stopping initialisation	✅ DONE
26	Implement early stopping check	✅ DONE

human_pattern_extractor.py

Line	TODO	Status
68	Implement initialisation (spaCy + GPT-2)	✅ DONE
73	Implement GPT-2 perplexity calculation	✅ DONE
78	Implement burstiness	✅ DONE
83	Implement sentence starter diversity	✅ DONE
88	Implement n-gram novelty	✅ DONE
93	Implement AI marker density	✅ DONE
98	Implement discourse density	✅ DONE
103	Implement punctuation patterns	✅ DONE
108	Implement full 17-dim feature extraction	✅ DONE
125	Implement KaggleHumanPatternDataset loading	✅ DONE
129	Implement __len__	✅ DONE
133	Implement __getitem__	✅ DONE
148	Implement HumanPatternClassifier MLP layers	✅ DONE
153	Implement forward pass	✅ DONE
158	Implement single-text scoring	✅ DONE

---

src/vocabulary/ — 10 TODOs ✅

awl_loader.py

Line	TODO	Status
21	Implement initialisation	✅ DONE
26	Implement word list file loading	✅ DONE
31	Implement synonym JSON loading	✅ DONE
36	Implement is_academic()	✅ DONE
41	Implement get_academic_synonyms()	✅ DONE
47	Implement all_words property	✅ DONE

lexical_substitution.py

Line	TODO	Status
41	Implement initialisation	✅ DONE
46	Implement contextual semantic similarity	✅ DONE
51	Implement AWL substitution generation	✅ DONE
56	Implement vocabulary elevation	✅ DONE
106	Implement register filtering	✅ DONE

register_filter.py

Line	TODO	Status
14	Implement initialisation	✅ DONE
19	Implement nominalisation	✅ DONE
24	Implement hedging	✅ DONE
29	Implement formality check	✅ DONE

---

src/evaluation/ — 7 TODOs ✅

gleu_scorer.py

Line	TODO	Status
20	Implement corpus-level GLEU scoring	✅ DONE
29	Implement BERTScore computation	✅ DONE

errant_evaluator.py

Line	TODO	Status
15	Implement initialisation (ERRANT annotator)	✅ DONE
23	Implement ERRANT evaluation	✅ DONE

style_metrics.py

Line	TODO	Status
19	Implement style similarity	✅ DONE
24	Implement AWL coverage	✅ DONE
33	Implement batch evaluation	✅ DONE

authorship_verifier.py

Line	TODO	Status
14	Implement initialisation (load model)	✅ DONE
19	Implement authorship verification	✅ DONE

---

src/inference/ — 3 TODOs ✅

corrector.py

Line	TODO	Status
39	Implement initialisation	✅ DONE
52	Implement full correction pipeline	✅ DONE

postprocessor.py

Line	TODO	Status
14	Implement initialisation	✅ DONE
19	Implement text cleanup	✅ DONE
27	Implement entity restoration	✅ DONE
32	Implement final formatting	✅ DONE

---

src/api/ — 2 TODOs ✅

main.py

Line	TODO	Status
22	Load config and initialise corrector on startup	✅ DONE
31	Implement /correct endpoint	✅ DONE

middleware.py

Line	TODO	Status
14	Implement request logging (timing, path, status)	✅ DONE
22	Implement rate limiter state	✅ DONE
26	Implement rate limiting logic	✅ DONE

---

scripts/ — 5 TODOs ✅

train.py

Line	TODO	Status
24	Implement training pipeline (10 steps)	✅ DONE

evaluate.py

Line	TODO	Status
19	Implement evaluation pipeline	✅ DONE

run_inference.py

Line	TODO	Status
21	Implement inference pipeline	✅ DONE

pretrain_human_pattern_classifier.py

Line	TODO	Status
23	Implement classifier pre-training	✅ DONE

---

tests/ — 18 TODOs ✅

test_preprocessing.py — 7 tests ✅

test_style.py — 4 tests ✅

test_model.py — 2 tests + 3 new ✅

test_vocabulary.py — 4 tests ✅

test_evaluation.py — 4 tests ✅

---

Shell Scripts ✅

Script	Purpose
train.sh	Multi-stage training with Skip/Redo/Continue checkpoint system
start.sh	Inference launcher (CLI REPL or API server)

---

Summary by Package

Package	TODOs	Status
src/preprocessing/	16	✅ ALL DONE
src/style/	14	✅ ALL DONE
src/model/	5	✅ ALL DONE
src/training/	22	✅ ALL DONE
src/vocabulary/	10	✅ ALL DONE
src/evaluation/	7	✅ ALL DONE
src/inference/	3	✅ ALL DONE
src/api/	2	✅ ALL DONE
scripts/	5	✅ ALL DONE
tests/	18	✅ ALL DONE
Total	97	✅ ALL DONE