Multimodal comic archive search

Problem Context

Comic collectors and self-hosters with large offline archives struggle to find specific comics or art styles quickly. Current tools like LANraragi or Komga only support Boolean metadata searches, forcing users to manually tag everything or memorize complex tag systems. When they need to find something like 'a wholesome story with a similar art style to this page,' traditional search fails completely.

Pain Points

Users waste hours manually tagging comics, reinventing the wheel with unstable DIY solutions, or giving up on finding what they need. Boolean search can't handle visual similarity or semantic meaning, and no existing self-hosted tool combines metadata with AI-powered visual search for comics. The frustration of not being able to rely on their archive drives them to abandon projects or waste time on incomplete workarounds.

Impact

The time wasted on manual tagging and failed searches adds up to dozens of hours per year. For collectors with thousands of comics, this means lost productivity, frustration, and even abandoned projects. Professionals who rely on these archives for research or curation face missed opportunities due to inefficient search. The lack of a reliable solution forces users to either accept poor search results or spend weekends building unstable DIY tools.

Urgency

This problem can't be ignored because it directly blocks access to their entire archive. Without a working search system, users can't trust their collection is usable, which is a dealbreaker for serious collectors. The frustration of reinventing the wheel—especially for technically skilled users—makes this a high-priority pain point that demands an immediate solution.

Target Audience

Comic collectors with large offline archives, self-hosting enthusiasts, doujinshi archivists, and researchers who need to search visual media by style or content. This includes hobbyists who spend hundreds on comics monthly, as well as professionals who curate or study comic art. Any user who relies on LANraragi, Komga, or similar tools but needs better search capabilities fits this niche.

Solution Approach

A self-hosted, lightweight multimodal search engine for comic archives that combines visual embeddings (SigLIP) with metadata for fast, intuitive search. Users upload their comics, and the tool indexes them using both human tags and AI-generated visual vectors. This allows searches like 'find comics with a similar dark art style to this page' or 'wholesome stories with tags like X and Y.' The solution is designed for self-hosters, with Docker support and minimal setup requirements.

Key Features

1. Self-Hosted & Lightweight: Runs on a user's own server with Docker, avoiding cloud dependencies or privacy concerns.
2. Automated Tagging Assistance: Uses AI to suggest tags based on visual content, reducing manual tagging workload.
3. FastAPI Backend + Vue Frontend: Simple, performant interface for searching and browsing results with visual previews.

User Experience

Users upload their comics once, and the tool indexes them automatically. From then on, they can search by keywords, tags, or visual similarity with a single query. The interface shows thumbnails of matching comics, ranked by relevance, so they can quickly find what they need. For power users, advanced features like automated tagging or bulk processing are available as paid upgrades, but the core search works out of the box.

Differentiation

Unlike existing tools, this is the *only- self-hosted solution that combines metadata with AI-powered visual search for comics. Most alternatives (LANraragi, Komga) only support Boolean metadata searches, while generic RAG tools lack comic-specific optimizations. The use of SigLIP embeddings ensures accurate visual similarity matching, and the self-hosted model avoids privacy concerns or cloud costs.

Scalability

The product starts with a simple search engine but can expand to include automated tagging, OCR for text extraction, cloud sync for backups, or even recommendation engines. Users with larger archives can scale by adding more server resources, while power users can unlock advanced features via a subscription. The modular design ensures it grows with the user's needs without requiring a full rewrite.

Expected Impact

Users save dozens of hours per year on manual tagging and failed searches. They can finally rely on their archive for research, curation, or personal enjoyment without frustration. For professionals, this restores a critical workflow that was previously broken. The tool pays for itself quickly by eliminating the need for DIY solutions or incomplete workarounds, making it a no-brainer for serious collectors.