infino-ai/infino

How infino-ai/infino works

Infino is an embedded retrieval library for a single logical table, exposing full-text, vector, hybrid, and SQL queries over Parquet files on local or object storage. The principal workflow is connect, create a table with schema and indexes, append rows, then search or query it. Internally, supertable coordinates immutable-file writing, manifest state, query execution, and caching. README statements establish the end-to-end workflow and storage capabilities; source-index references support several internal relationships, but many implementation files were not sampled, so uncertain call edges are omitted.

Main entry points: query/mod.rs Query Execution · catalog/mod.rs Connection API · writer.rs Table Writer · reader_cache/mod.rs Reader Cache

Components 21

API and Catalog

Connection API src/catalog/mod.rs
Language Bindings infino-python/src/lib.rs
Engine Configuration src/config/mod.rs

Table Lifecycle

Schema and Indexes src/supertable/schema/mod.rs
Dynamic Rows src/dynamic/mod.rs
Table Writer src/supertable/writer.rs
Write-Ahead Log src/supertable/wal/pipeline.rs
Table Manifest src/supertable/manifest/mod.rs
Parquet Superfiles src/superfile/builder.rs
Table Handle src/supertable/handle.rs

Retrieval and Query

Full-Text Index src/superfile/fts/builder.rs
Text Analysis src/superfile/fts/tokenize.rs
Vector Index src/superfile/vector/hnsw.rs
Query Execution src/supertable/query/mod.rs
SQL Planning src/supertable/query/sql.rs

Persistence and Cache

Storage Providers src/storage/mod.rs
S3 Backend src/storage/s3.rs
Reader Cache src/supertable/reader_cache/mod.rs
Cached Block Reads src/supertable/reader_cache/block_source.rs
Memory Budget src/memory/mod.rs

Other components

Application

Connections 24

Application → Connection API: connects README.md
Application → Language Bindings: uses README.md
Connection API → Schema and Indexes: creates tables README.md
Connection API → Table Handle: opens tables src/catalog/mod.rs
Table Handle → Table Writer: appends rows
Table Writer → Parquet Superfiles: writes batches README.md
Table Writer → Write-Ahead Log: records writes
Table Writer → Table Manifest: commits files
Table Writer → Dynamic Rows: accepts JSON rows
Parquet Superfiles → Full-Text Index: embeds index README.md
Parquet Superfiles → Vector Index: embeds index README.md
Full-Text Index → Text Analysis: analyzes text src/superfile/fts/builder.rs
Table Handle → Query Execution: searches table
Query Execution → SQL Planning: plans SQL
Query Execution → Full-Text Index: runs full-text search README.md
Query Execution → Vector Index: runs vector search README.md
Query Execution → Table Manifest: reads snapshot
Query Execution → Reader Cache: loads readers
Connection API → Storage Providers: selects storage README.md
Storage Providers → S3 Backend: dispatches backend src/storage/mod.rs
Reader Cache → Cached Block Reads: fetches blocks src/supertable/reader_cache/mod.rs
Reader Cache → Storage Providers: reads bytes src/supertable/reader_cache/mod.rs
Reader Cache → Memory Budget: budgets residency
Connection API → Engine Configuration: loads settings

Drawn by GitDiagram on November 12, 2026 from the repository's file tree, README and source files. Read as Markdown ·