JMFTS — Tutorials¶
Action × Acquisition
Guided builds for JMFTS, from a first index to a tuned hybrid query.
Tutorials are read in order. The quickstart is the first one — it gets something running so the rest have somewhere to start from.
- Quickstart — stand the appliance up, put three documents in it, and run vector, BM25 and hybrid queries over them. Teaches what the tree is for and why only one of those three scores can be thresholded. Runs on the 0.1.0 release.
- Ingest a document — upload a real file and watch a queue
turn it into a tree. Teaches that ingestion is asynchronous, what
settledmeans, the three ways to see a pipeline's plan before running it, and what deduplication does to an upload you thought was new. Needs thefeat/ingest-lifecyclebranch.
Once both are done, the Cookbook is the place to go for one specific job — customizing chunking, choosing a summarization model, adding your own task type, or routing work to a GPU pool.