| --- |
| license: mit |
| task_categories: |
| - text-generation |
| language: |
| - en |
| - sa |
| tags: |
| - fantasy |
| - biology |
| - worldbuilding |
| pretty_name: "नानायोनि-कामभेद-संग्रहः (Nānāyoni-Kāmabheda-Saṅgrahaḥ)" |
| size_categories: |
| - n < 1K |
| --- |
| |
| # Fantecchi: Nānāyoni-Kāmabheda-Saṅgrahaḥ |
|
|
| A specialized framework and language model designed for **Lore-Strict Biological Roleplay**. This project fine-tunes models to handle complex, non-human anatomical and reproductive systems without defaulting to human biological assumptions. |
|
|
| ## 🛠 Project Structure |
|
|
| The project is divided into two distinct pillars: the **Training Pipeline** (The Engine) and the **Application Layer** (The Showroom). |
|
|
| ### 🏗 Core (Scaffolding & Scripts) |
| - `core/training_pipeline/`: |
| - `lore_refinement/`: Scripts that evolve raw world-building into structured JSON lore. |
| - `data_synth/`: Preprocessing scripts that convert lore into ChatML training data. |
| - `trainers/`: The LoRA/QLoRA fine-tuning logic and model output. |
| - `core/application_layer/`: |
| - `exporters/`: Scripts that generate platform-specific chatbot profiles (JanitorAI, etc.). |
| - `local_ui/`: The Gradio-based local chat interface. |
| - `documentation/`: Automated dictionary and markdown generators. |
|
|
| ### 📊 Data & Assets |
| - `data/setN__.../`: Versioned source data and final processed lore. |
| - `data/training/`: Generated datasets for the fine-tuning process. |
| - `chatbots/`: Exported bot profiles for external platforms. |
| - `docs/dictionary/`: Consolidated markdown dictionaries for human reference. |
|
|
| ## 🧬 How to Use |
|
|
| ### 1. Training (Pillar A) |
| To train the model on current or new lore: |
| ```bash |
| python main.py |
| ``` |
| This will run the full pipeline: Preprocessing -> Training. |
|
|
| ### 2. Application (Pillar B) |
| To chat with the model locally: |
| ```bash |
| python app.py |
| ``` |
|
|
| ## 🚀 Adding New Racial Biology |
| 1. Add a JSON file to `data/set5__culture/`. |
| 2. Run `python main.py` to bake the new biology into the model's weights. |
| 3. Alternatively, paste the biology block directly into a chat prompt for instant "In-Context" recognition. |
|
|
| --- |
| **Authored by नग्नाक्षी (Nagnākṣī)** |
| *(Original character and lore imagined by the Fantecchi team)* |
| *"A Compendium of Erotic Variations Across Many Forms"* |
