PROJECT MAMBO
Change colour theme

MamboMeme

MamboMeme is a local ranked-search application for meme images and quotes. Rust owns ingestion and the terminal interface; Python owns index construction, ranked retrieval, the long-lived worker, and evaluation.

Project boundary

  • A query such as john cena returns ranked stored assets for the user to inspect and choose.

  • Phase 1 builds a deterministic local SQLite/FTS corpus from a cleared fixture.

  • Phase 2 implements BM25, a measured LSA dense experiment, deterministic hybrid fusion, evaluation, and the worker. BM25 remains the default because fusion did not pass the confidence rule.

  • Phase 3 implements a Rust TUI connected to one long-lived Python retrieval worker through a versioned local protocol.

  • Phase 4 is paused before release. Its design is one finite, human-reviewed Wikimedia Commons page-ID plan through the same corpus contract, not a general crawler.

  • Selection feedback is optional offline evidence, never live self-training.

  • Context-aware suggestions reuse retrieval only after the prompt-search product works.

  • MMTS-Search-v1, component metrics, and contract tests justify each implemented stage.

Understand the system and corpus

Validate and plan delivery

Project status

MamboMeme is paused after the complete Phase 3 local source release. Phase 4's reviewed Commons source, deletion handling, stale-process protection, controlled network tests, and source report are designed but not released. MMTS-Search-v1 remains unscored until an independent human-labelled hidden benchmark with its safety subset exists; no context assistant, public API, packaged distribution, or deployment target is implemented.