datdang-dev/r18-researcher

R18 Research & Downloader Engine

Overview

The R18 Research & Downloader Engine enables the LND Studio workspace to acquire, structure, and analyze adult creative materials. It handles automatic multi-source resource acquisition and analyzes Japanese R18 literature to build structural blueprints for premium prose generation.

Core Philosophy: Authentic reference material breeds authentic creative execution. Kuro operates the downloading pipeline and extracts deep prose insights without compromises.

Capabilities

1. Unified R18 Downloader

Kuro uses the unified python tool studio/tools/r18_downloader.py to fetch:

• Syosetu / Nocturne R18 Novels: Downloads all chapters, prefaces, and afterwords to clean, organized Markdown files.
• Hitomi.la: Interfaces with node-hitomi to download high-fidelity WebP manga assets.
• nhentai: Integrates with the Turkish/English nhentai-downloander package.
• Telegraph / Teletype: Direct thread-pool scraping of high-speed image galleries.

2. Erotic Fetish & Tag Research (RH)

Uses direct authenticated crawler search to extract:

• Popular fetish tag clusters.
• Trend analyses in the R18 Japanese target market.
• Recommendations for specific titles/tags matching the user_fetish_profile.md.

3. Novel Prose Learning (RL)

Analyzes scraped Syosetu Nocturne novels to extract:

• Pacing Topology: Ratio of dialogue, description, internal monologues, and explicit actions.
• Sensory Vocabulary: Key Japanese and localized English verbs/nouns used to depict fluids, heat, physical feedback, and psychological transformations.
• Degradation Patterns: Authentic corner-bracket Japanese formatting and emotional escalation loops.

---

Workflows & Commands

[RD] Run Downloader

Triggers the universal scraper/downloader script: python3 studio/tools/r18_downloader.py <URL_or_ID>

Output is structured and stored:

• Novels: sources/novels/<Title>/chapters/chapter_001.md, etc. + metadata.json
• Manga / Images: sources/manga/<Title>/page_001.webp, etc. + info.json

[RH] Research Hentai Fetishes

To run a research sweep for specific terms or tags:

1. Load {{project_root}}/studio/rules/user_fetish_profile.md to establish core target kinks.
2. Standard web search engine (search_web) is age-gated/censored for Nocturne. DO NOT use search_web to scan Nocturne directly.
3. Instead, run programmatic searches using the tool CLI: python3 studio/tools/r18_downloader.py --search "<Japanese_Query>"
4. Generate a structured research_results.md mapping:
   • Popular tag clusters.
   • Narrative tropes and setups.
   • Recommended novel IDs or doujin tags.

[RL] Novel Learning Mode

To extract prose style patterns from a downloaded Syosetu novel:

1. Run the automated prose analyzer: python3 studio/tools/r18_downloader.py --analyze <Novel_ID_or_Path>
2. Read the generated analysis_prose.md file inside the novel directory to capture:
   • Pacing topology (Dialogue vs Monologue vs Action ratios).
   • Frequency of key R18 verbs, nouns, and SFX.
3. Create a clean prose blueprint sources/novels/<Novel_Title>/analysis_prose.md to help the prose generator.

---

Quality Gates

• [ ] Does the download run without interactive prompts?
• [ ] Are crawled Syosetu novels split into structured chapter files (not a single giant dump)?
• [ ] Are prefaces and afterwords extracted and labeled clearly?
• [ ] Does prose analysis use --analyze and generate a structured markdown report?
• [ ] Are all manga assets saved in high-fidelity WebP/PNG/JPG (no missing pages)?