Use Cases

The use-case docs follow the same branching structure as the clickable use-case map on the home page. Start with the conceptual overview for your use-case type, then drill down into writer or reader specifics.

Not sure which extras to install? See the Extras matrix for a visual map from every use case to the pip install / uv sync --extra target you need.

Shared Snapshot Workflow

All use-case families publish through the same manifest + _CURRENT pointer model. See Shared Snapshot Workflow for the project-wide flow, or Manifest & _CURRENT and Manifest stores for implementation details.

Sharded KV Storage

The foundational use case: write key-value pairs into sharded immutable snapshots, read them back with routed lookups.

• Overview — KV sharding, manifests, two-phase publish, safety properties (start here)
• Build
• Choosing a writer
• Python
• Spark
• Dask
• Ray
• Read
• Choosing a reader
• Sync SlateDB
• Sync SQLite
• Async SlateDB
• Async SQLite

Sharded KV Storage with Vector Search

Write both KV pairs and vector embeddings in the same snapshot. Query by key or by nearest-neighbor search.

• Overview — shared manifest model, vector metadata, composite vs unified backends
• Build
• Composite (SlateDB + LanceDB)
• Unified (sqlite-vec)
• Read
• Sync — UnifiedShardedReader
• Async — AsyncUnifiedShardedReader

Sharded Vector Search

Vector-only: write embeddings into sharded indices, query by approximate nearest-neighbor search.

• Overview — manifest-backed routing strategies, scatter-gather flow
• Build
• LanceDB
• sqlite-vec
• Spark
• Dask
• Ray
• Read
• Sync — ShardedVectorReader
• Async — AsyncShardedVectorReader

Operations

Operational tasks for all use-case types:

• CLI
• History & rollback
• Prometheus metrics
• OTel metrics
• Production guide
• Cloud testing
• Tox & dependency matrix