Drawing alone means no reviewer nearby, and uploading practice sketches to a cloud service feels wrong. How Dessan-sensei removed both walls with a local vision model on the LAN, using text-only history as context and level-aware prompting.
[Read More]
A Drawing Critique App in Go, SQLite (WAL), React, and an Ollama VLM
Dessan-sensei scores a drawing photo on six axes out of ten using a local vision model. A tour of the four layers — Go, SQLite in WAL mode, React, and Ollama — through the prompt builder, the JSON parser, and the transactional save.
[Read More]
Automating the Split of Stream Archives: Three Walls of Manual Work, Memory, and Speed
Splitting a long stream archive into individual songs runs into three walls: manual timestamps, huge memory use, and waiting. How JamSlicer removes each one, with the design and code behind it.
[Read More]
How JamSlicer Splits Long Videos with FFmpeg silencedetect
A technical tour of JamSlicer: one FFmpeg pass for volume analysis, segment calculation in the browser, and stream-copy cutting. Why each layer is split the way it is, with code from the repository.
[Read More]
A LINE Butler That Keeps Its Character While Recommending Products
A characterful bot turns into a salesperson when you add product links, and a sales bot loses the character. How turtle-buttler split voice, catalogue, and judgement across a prompt, an API, and the model.
[Read More]
Building a LINE Butler Bot with Gemini Function Calling and MCP
How turtle-buttler, a LINE bot that speaks as a Kansai-dialect butler, is put together: Gemini 2.0 Flash function calling, MCP servers backed by Firebase Realtime Database, and an HTTP fallback for Cloud Run.
[Read More]
When Your Likes Get Buried: Designing Capture and Rediscovery for X Favorites
Likes on X sink into the timeline the moment you tap them, and there is no way to search by what an image looks like. How Refav Gallery split capture, storage, and rediscovery to fix that, including the tradeoffs and measured results.
[Read More]
Rediscovering Likes with Similar Images: Next.js, pgvector, DINOv2, and an MV3 Extension
How Refav Gallery turns X likes into a searchable image gallery: passive interception in a Manifest V3 extension, DINOv2 embeddings on Modal, a pgvector HNSW/exact-KNN split, and MMR for diversity, with code and measured numbers from the repository.
[Read More]
Preventing Missed Sequels by Design: Stop Relying on Remembering at the Bookstore
Missing the next volume is not a memory problem, it is a problem of where the state lives. How tsuzukan records owned volumes by ISBN and reports what the series is missing, including the cases a machine cannot decide.
[Read More]
How tsuzukan Tracks Manga Sequels: Go, React 19, and Three Bibliographic APIs
A tour of tsuzukan’s internals: a Go 1.26 + Gin API, a React 19 frontend, PostgreSQL, Firebase Auth, and three bibliographic sources with very different personalities — openBD, NDL Search, and Rakuten Books — folded behind one interface.
[Read More]