palimpsest-bot

A research-grade bot that rewrites source texts in the style of OLD works and/or renders them into historical or “forgotten” scripts/languages (e.g., Theban alphabet attributed to Agrippa, Celestial alphabet, Classical Latin, Attic/Koine Greek, Coptic). Includes:

• Style adapter for archaic diction and syntactic inversions
• Transliteration engines for historical/occult alphabets
• Optional classical-language generation via open-source NLP
• Provenance + disclaimers baked into output metadata

⚠️ Accuracy note This project does not claim philological authenticity. For real scholarship, keep the provenance blocks and cite original editions. Many “occult” alphabets are cipher scripts mapping modern letters, not actual languages.

Features

• Style transfer: Early-Modern English, Late Latin, Koine Greek-like diction (heuristic)
• Script transliteration: Theban, Celestial; easy plugin interface for new scripts
• Classical language support (where feasible): Latin (CLTK/latindata), Greek (polytonic)
• Prompt/LLM driver or “LLM-free” heuristic mode
• YAML configs for pipelines; CLI + Python API
• Unit tests; reproducible seeds; metadata stamped into every artifact

Quickstart

# 1) create & activate env
python -m venv .venv && source .venv/bin/activate  # Windows: .venv\Scripts\activate

# 2) install
pip install -r requirements.txt

# 3) run a simple rewrite (archaic Early-Modern English)
python src/palimpsest/main.py \
  --input "Knowledge is the seed; discipline the rain." \
  --style early_modern_en \
  --out out/seed_earlymodern.txt

# 4) transliterate same line into Theban alphabet (cipher)
python src/palimpsest/main.py \
  --input "Knowledge is the seed; discipline the rain." \
  --script theban \
  --out out/seed_theban.txt

# 5) pipeline: archaic style → Latin attempt → Celestial script
python src/palimpsest/main.py \
  --input "Light seeks the path through shadow." \
  --style early_modern_en \
  --target_lang la \
  --script celestial \
  --out out/light_pipeline.txt \
  --config configs/pipeline.default.ymll

Examples
	•	style: early_modern_en → “Knowledge is the seed; and discipline the rain, whereby the mind bringeth forth increase.”
	•	script: theban → same sentence rendered in Theban glyphs (see docs/samples/).

Architecture
	•	normalizer → cleans input (unicode, punctuation, casing)
	•	style_transfer → rule-based archaizer + optional LLM helper
	•	classical_gen → Latin/Greek helpers via CLTK or templates
	•	transliterate → pluggable script mappers (Theban, Celestial, user-defined)
	•	provenance → inserts a JSON block (source hash, time, config, model)

Add a new script

Create src/palimpsest/scripts/my_script.py exporting:
NAME = "my_script"
MAPPING = {"a": "…", "b": "…", ...}

Then register it in registry.py.

Ethics & Licensing
	•	Public-domain texts only for full reproduction; otherwise quote sparingly and cite.
	•	Outputs include a Provenance footer by default.
	•	License: MIT (see LICENSE).

Roadmap
	•	Better Latin morphology with lemmas & inflection templates
	•	Coptic transliteration + rendering
	•	Fine-tuned archaic-English n-gram model (LLM-free mode)

# requirements.txt
```txt
regex>=2024.5.10
unidecode>=1.3.8
PyYAML>=6.0.2
click>=8.1.7
rapidfuzz>=3.9.7
cltk>=1.2.0    # classical languages toolkit (Latin/Greek utilities)
# requirements.txt
```txt
regex>=2024.5.10
unidecode>=1.3.8
PyYAML>=6.0.2
click>=8.1.7
rapidfuzz>=3.9.7
cltk>=1.2.0    # classical languages toolkit (Latin/Greek utilities)