infino append logs
Start without Parquet
Append JSON over REST and Infino writes hydrated Parquet in your bucket with the columns and BM25 index in one file. Embeddings can come from the table’s ingest model, from vectors you send, or from a later hydration pass.
infino append logs # indexed on write
Ingest is the integration
Write documents to a table and every batch lands as a standard Parquet file with the indexes inside.
- 01 Append JSON REST, single documents or batches
- 02 Indexed on write BM25; embeddings generated or supplied
- 03 Parquet in your bucket Readable by the tools already using the data
infino migrate # the old system keeps serving
Backfill the history in batches
- Export from the system you have: scroll out of an Elasticsearch cluster, dump from a database, or replay from the original source.
- Append them batch by batch. Each batch is indexed on write and lands as Parquet, and what has landed is queryable while the rest loads.
- Embeddings you already have ride along with the documents. For text-only documents, Infino can generate them on ingest using the model selected for the table.
infino hydrate s3://your-bucket/logs
Enrich everything later, in batches
- Ingest text now and add embeddings later. Hydration is a batch pass that rewrites the landed files with the embedding column and its index inside.
- A pass covers the whole table or the prefix you point it at.
infino contact