haiku.rag/docs
Yiorgis Gozadinos 6f976ef2a9
Share the LanceDB session across connections
Every `Store` built its own connection with its own caches and discarded them on
close, so the index a vector query loads was refetched by the next connection.
On object storage that first fetch dominates: measured on a ~500k-chunk 2560-dim
corpus over a ~200ms link, the first query cost ~41s and the second ~3s, and a
new connection reusing the session cost ~7s instead of ~47s.

`connect_lancedb` now passes a process-wide session, keyed on the configured
cache sizes so a caller asking for different sizes gets its own.

Also sets `read_consistency_interval`, defaulting to 30s. It was None, meaning a
connection never re-checked for other processes' writes. Per-call connections hid
that; a shared session makes connections long-lived enough for a reader to go
stale against the ingester.

All three settings reject negatives at the config boundary. A negative cache size
raises OverflowError and a negative interval panics inside Lance, so neither is
catchable further in. Zero stays valid for both: no cache, and check on every
read.

The routing tests now assert the kwargs they care about rather than the full call
signature, since every connection carries the two new kwargs.
2026-08-18 14:58:17 +03:00
..
capabilities Refuse to compact evidence the host kept no record of 2026-08-13 15:46:23 +03:00
configuration Share the LanceDB session across connections 2026-08-18 14:58:17 +03:00
img Add ingester screenshot 2026-05-27 15:53:01 +03:00
stylesheets Landing page hero with TUI screen recording 2026-05-20 16:56:25 +03:00
apps.md Bind app example backend to loopback and document its lack of auth 2026-07-08 11:37:03 +03:00
benchmarks.md Lead the benchmarks page with results 2026-08-18 14:37:40 +03:00
changelog.md
chat.md Update caps documentation 2026-08-13 14:08:14 +03:00
cli.md Accept images on CLI ask/analyze and MCP tools 2026-07-25 10:35:25 +03:00
custom-pipelines.md Always use the Store-owned embedder 2026-05-29 11:36:53 +03:00
development.md Docs 2026-04-24 14:42:52 +03:00
index.md Fix the MCP registry entry and fill in package and docs metadata 2026-08-18 14:37:05 +03:00
ingester.md Classify obstore config errors as permanent 2026-08-06 12:53:45 +03:00
installation.md Fix packaging and settings facts in the docs 2026-07-28 18:22:05 +03:00
mcp.md Accept images on CLI ask/analyze and MCP tools 2026-07-25 10:35:25 +03:00
overview.md Register the optional capabilities where agents are composed 2026-08-13 15:04:05 +03:00
python.md Load document blobs only when asked 2026-08-06 13:04:51 +03:00
remote-processing.md Store docling-serve chunk bodies via raw_text 2026-06-23 16:35:21 +03:00
tools.md replace haiku.skills with native Pydantic AI capabilities 2026-07-24 15:26:17 +03:00
tuning.md Remove the wix evaluation dataset 2026-08-14 14:46:51 +03:00
tutorial.md replace haiku.skills with native Pydantic AI capabilities 2026-07-24 15:26:17 +03:00