Ten merges and renames, all in one direction: fewer pages, each covering a whole
subject.
Table absorbs schema evolution and versioning. Vector search absorbs
multi-vector. Explain absorbs explain analyze. Reranking was seven pages -- an
overview and six rerankers of a few dozen lines each -- and is now one. The
vector index page absorbs quantization, which was only ever a parameter of it.
Indexing moves to Manage, where the other catalog objects are. The operational
half stays in Operate as one Indexing page, absorbing GPU indexing -- which was
filed in sophon despite importing lancedb's snippets and being mostly
open-source content with one Enterprise-badged section.
Performance was a single page aggregating tips for four unrelated operations,
which is hard to keep current. Each section now sits with the operation it is
about: ingestion tips on Insert, index tips on Indexing, compaction tips on
Compaction, query tips on Search.
Data Types leaves Concepts. The type reference belongs in the SQL reference,
which already described the same types from the query side; Table now carries a
short summary that points there. The multimodal material it had absorbed is its
own page again, since storing blobs is a task rather than a type definition.
Query becomes Semantic Search, Branches becomes Branch, and the Enterprise
Security page becomes Compliance.
Redirects cover every moved URL, and 30 files were repointed at the merged
anchors.
The function catalogue is now real rather than a pointer. Scalar functions are
listed by taxonomy -- math, string, unicode, regex, date/time, array, map,
struct, crypto, type -- and the names come from what datafusion-functions 54
actually registers, not from what DataFusion documents on main. Aggregate and
window pages gain the same treatment.
Two things the audit corrected rather than copied.
SHOW TABLES does not work. DataFusion implements it over information_schema,
which this session does not enable, so it fails at plan time. It was documented
as working here, on the SHOW page and in three Manage tables; all four are
fixed, and the SQL overview now lists the absence alongside SHOW COLUMNS.
Prepared statements are a Flight SQL protocol feature, not a PREPARE statement.
A script carrying PREPARE ... AS SELECT will not run, while the same intent
expressed through a client will. The new page says so rather than copying
DataFusion's syntax.
Every section of the Documentation tab that has a SQL counterpart now links to
it -- 27 of 76 non-SQL pages, each pointing at the specific statement rather
than at the reference index.
Namespace and Table were the only objects behind a nested group. Both are now a
single page like Database, Materialized view, View and Function, so Manage is a
flat list of the things a catalog holds.
Namespace merges its two pages, which were a concept page and a usage page for
the same object.
Table is a new page. The old tables/index was a 672-line basic-usage tutorial
whose every section had acquired a better home during the restructure --
connecting, ingesting, searching, filtering, adding and dropping columns -- to
the point that its own Next steps cards pointed at the pages that superseded it.
It is replaced by a table object page: what a table carries, create, open, list,
rename, clone, drop, and links out.
Schema evolution, versioning and branching stay as siblings rather than folding
in. Merging them would have made one 1,800-line page, which is not what the
other object pages look like.
Every group now answers exactly one question. Get started and Concepts cover
what the system is; the verb groups cover doing something with it; Operate,
Secure and Deploy cover running it; SQL reference is lookup. Maintenance becomes
Operate, Governance becomes Secure, and the rest take verbs to match.
Three new pages in Concepts and Get started:
Guarantees states what a write promises when it returns, which version a reader
sees, what survives a failure, and -- the half usually missing -- what is not
guaranteed: no multi-table transactions, no read-your-writes across processes
without an interval, unenforced primary keys that do not enforce, no background
compaction.
Limits collects bounds that were stated in passing across a dozen pages, with
the numbers verified against the code rather than inferred: name lengths, job
event caps, SHOW ERRORS bounds, the sampling range that is refused rather than
clamped.
Architecture illustrates embedded against distributed as the one deployment
decision, and is written to cover both rather than only Enterprise.
Indexing splits by question. Choosing an index is a Query decision and moves
there; building, repairing and GPU-building it stay in Operate.
SQL becomes a top-level reference of fifteen pages instead of a sub-group of
Query. Auditing sophon-sql turned up a large undocumented surface: ALTER TABLE
ADD COLUMN in its plain, computed and function forms with FILTER, IGNORE NULLS
and OVER modifiers; SET UNENFORCED PRIMARY KEY; SET and UNSET WAL with sharding;
CREATE FUNCTION with its dependency options; REFRESH COLUMN and REFRESH
MATERIALIZED VIEW; and the SHOW family. None of it was written down.
Both tabs held one group of the same name as the tab, so every entry sat a level
deeper than it needed to and the group heading said nothing the tab had not
already said. The overview now sits above the real groups, which the tab can do
once it uses pages rather than groups.
Demos has no categories at all, so its four pages are simply flat.
The embedding overview is renamed from Overview, which would otherwise have been
the second entry with that label in the same sidebar.
The three gallery pages held thirteen examples as rows in a table, reachable
only by opening the gallery first. Each is now its own page, so the sidebar
lists them and search finds them by name.
A page for an example that lives in vectordb-recipes says so and links to the
notebook and its source, rather than pretending the tutorial is here.
Overview moves above the groups -- the tab uses pages rather than groups, which
is what allows an entry to sit outside every group. Stable World Model joins
Training, since that is what it is doing. The agent plugin page is now Using
LanceDB Agent Skill.
Mintlify cannot link outward from a sidebar: an {anchor, href} entry is ignored
and a bare URL renders with its scheme mangled to a single slash, so a real page
per example is the only way to list them.
The tab is Use Cases now, grouped General, Search, RAG, Feature Engineering,
Training, World Models -- one page per example, named for the example rather
than for its position in a tree.
The gallery pages that listed external notebooks become the More Examples entry
of the category they belong to, instead of standing in front of the real
tutorials. The top-level overview stops linking to itself through three levels
of index page and names the categories directly.
Branch experiments move to Feature Engineering, which is what they are an
experiment in. The agent plugin leads General, since it applies to everything
else in the tab.
LeRobotDataset moves to the Datasets tab, where the LeRobot datasets already
are, and Robotics moves to the top of that tab.
SELECT and Enterprise SQL sat apart in Query while documenting the same engine.
They are now one SQL group: overview, Select, Aggregation, Join, Window,
Sampling, Functions -- one page per construct, with the Enterprise client-setup
and FTS pages merged in rather than parallel to them.
Sampling is new. TABLESAMPLE is implemented and was undocumented: BERNOULLI
picks rows, SYSTEM picks Lance fragments, REPEATABLE fixes the draw so a sample
works as a stable evaluation set. Its refusals are documented too -- a
percentage over 100, a fractional row count, BUCKET, the SAMPLE spelling and
the OFFSET modifier are all errors, because a sampling clause that were read and
not applied would be a wrong answer rather than a missing feature.
Schema & Data Types is one page again, with the multimodal content merged into
its blob section rather than sitting beside it.
Agentic Skills stops being a tab. Both its pages are tutorials, so they join the
Tutorials tab; the Basics page that introduces them stays.
Basics gains a Schema & Data Types section -- the Arrow types a table can hold,
the extension types Lance adds on top (lance.json, lance.blob.v2,
lance.bfloat16), and where geospatial actually stands. The multimodal page moves
in beside it. Consistency joins Basics too, after Query & Job.
Query & Job now leads with the thing the _async convention actually gives you: a
handle. execute_query_async returns a query, everything else returns a job, and
both answer the same four questions -- which one, how is it going, give me the
result, stop. Only the third differs, because a query produces rows to stream
and a job produces an outcome to wait for.
Management gains a Materialized View page. Creating one was documented only on
the refresh page, and the stable-row-ids requirement is a creation-time
decision that cannot be undone.
Embeddings move to Integrations as the Embedding Registry, which is what they
are -- a client-side feature that wraps provider models, sitting above the
providers it registers.
The demo gallery was three apps on one page; it is now three pages. Use Cases
folds into Tutorials as Tutorials and Use Cases.
Versioning and branching are properties of a table, not steps in getting data
in, so they join the Table group. Consistency stays with Ingestion, where the
question it answers -- when do other writers' rows become visible -- is asked.
The Ingesting Data page was in the Table group while an Ingestion overview
existed separately, so the two are merged. The large-dataset section was
duplicated between it and the insert page; the insert page keeps it.
Schema and Data Evolution covered two different things under one title.
Changing a table's shape stays in Table management; deriving values from data
already there becomes a Feature engineering page, next to computed columns and
functions, which are the same job at larger scale.
Feature engineering is now the overview plus one page per operation: add a
computed column, refresh a column, refresh a materialized view. Computed
columns covered declaring and filling on one page, which are separate decisions
with separate failure modes.
Functions move to Management, where the other catalog objects are, and absorb
the environment-mismatch guidance from the Enterprise page -- a symptom table
worth having beside the parameters that cause it.
The Enterprise feature-engineering page is merged into the overview. Two of its
claims were stale: create_materialized_view is implemented for db:// connections
in remote/db.py, so views are not local-only, and the tier is Enterprise rather
than 'Cloud and Enterprise'.
The REST reference leaves the sidebar. Mintlify only generates per-endpoint
pages from a navigation group -- hidden, dropdowns, global anchors and a
top-level openapi key were all tried -- so removing the group removes those
pages. The overview survives as an unlisted page, linked from SDKs & SQL, and
the spec still ships for anything that wants to read it directly.
Basics now reads outward from the format: Lance Format, Lakehouse & Catalog,
Query & Job, SDKs & SQL, Agentic Skills.
Jobs becomes Query & Job. Async means two unrelated things here -- whether your
thread blocks, and whether the work outlives the call -- and nothing said so.
The API Reference tab is gone. Its SDK index is folded into SDKs & SQL, where
the surfaces are already compared; the generated REST pages move into Basics
beside it.
Connecting was described in passing on the catalog page. It is now a Management
page covering the URI forms and what each changes, deployment options, storage
options, and read consistency, since which one you pass decides what exists.
Quickstart ran out at multimodal storage. It now continues through a computed
column and its refresh into a PyTorch loader, which is the arc the rest of the
docs assume.
The Documentation tab now reads as a manual over Python, TypeScript and SQL
rather than a Python guide with SQL bolted on.
Basics replaces the Catalog section: what the catalog holds, what each API
surface can do, jobs, agentic skills, and what Lance is. Management is new and
is the CRUD and listing reference for every catalog object, with a column per
language so the gaps are visible -- TypeScript has no function or SQL methods,
SQL needs a deployment.
Search becomes Query and gains a SELECT reference, since an analytical query is
a query whether it came from a builder or a statement.
Jobs is new. Every long-running operation already returned the same handle from
an _async method; nothing documented it.
The SQL function reference gains what Lance registers on top of DataFusion:
contains_tokens, backed by the FTS index, and the JSON accessors, backed by a
JSON scalar index. Geospatial st_* functions are behind a build feature LanceDB
does not ship, which is now stated rather than left to a confusing error.
Removes the Geneva pages' orphaned snippets, merges FTS Examples into Full-Text
Search, and splits Query Optimization into Explain and Explain Analyze.
The tab opened on Quickstart, buried "What is LanceDB?" inside it, and put
everything else under a single Guides group that had to be expanded before any
section was reachable. Model training sat outside it for no reason.
Reorganize around what a user does, in order: Introduction, Catalog, Ingestion,
Search, Feature engineering, Training, Maintenance, SQL. Guides is gone as a
level, so each section is one click.
New pages: catalog overview and database, and a compaction page under
Maintenance. tables/update covered four separate operations, so it becomes the
Ingestion section — insert, merge insert, update, delete — with insert drawing
on the large-dataset material from tables/create. Its old URL redirects.
Agentic skills move to their own tab after API Reference.
DataFrame support is not going ahead, so the pages documenting it come out:
the relational operations and the expression system, both of which described
only that API.
The manual was framed around having two query surfaces and choosing between
them. With one, that framing is gone — the overview now leads with SQL over
Arrow Flight SQL and what the dialect is, rather than with a comparison to
something that will not exist.
The function reference keeps everything that was about SQL: the DataFusion
version, how Lance columns map onto its type system, the pseudo-columns, blob
handling, and `fts()`. Only the section on reaching functions through the
Python wrapper is replaced, since the wrapper is going too.
Mentions of pandas and polars DataFrames elsewhere in the documentation are
unrelated and untouched.
The manual said the SQL dialect's function and type reference was out of scope.
It is not: the engine is DataFusion 54 with its default function set registered,
so the dialect is documented already and the useful thing is to say which
version and point at it.
The types are worth being concrete about rather than deferring. Lance's
DataFusion table provider returns the dataset's Arrow schema directly — there is
no SQL type layer over storage — so DataFusion's type reference describes Lance
columns exactly, and a vector column is a FixedSizeList that behaves like one.
Two consequences of that mapping are documented because they surprise people.
The pseudo-columns the provider can append are real and typed — `_rowid`,
`_rowaddr` and `_rowoffset` are nullable UInt64, and the nullability is a reused
selection bitmap rather than a missing value, so a null there means nothing.
And a blob column's Arrow type depends on how it is read: descriptions by
default, binary on request, which changes the schema a query sees.
`fts()` is recorded as what it is, a DataFusion TableProvider over the dataset.
That explains the parts that otherwise look arbitrary: why it is a table
function rather than a predicate, why every column of the table is available in
its result, and why projections push into the scan.
Functions and materialized views are objects in the Lance Multimodal Lakehouse,
alongside tables and indexes — created, versioned and refreshed like anything
else in it. The pages introduced them as client methods, which undersold what
they are and left the capability sounding like a helper rather than part of the
storage layer.
The materialized view page also framed Enterprise availability around which
client you are holding. A reader does not care about that; they care whether
views exist where their data lives. They do, refreshed across the cluster, and
the page now says so and points at the Enterprise documentation rather than
implying the feature is missing.
Both the lancedb client and the geneva client expose
`create_materialized_view`, and they are not the same feature. The page said
materialized views are local only, which is true of this client and misleading
for anyone on Enterprise, where views are created through geneva and refreshed
on the cluster.
Found by assembling the combined site and reading the Geneva pages against the
new ones: they described the same call as distributed and Enterprise while this
one described it as raising. Both were accurate about their own client and
neither said which client it meant.
The feature engineering documentation describes the geneva package, a separate
SDK. The capability it covers now lives in this client — materialized views,
computed columns, and registered functions are all shipped API with no pages
describing them.
Four pages cover the three mechanisms and, more usefully, when each applies.
The overview leads with that choice rather than with concepts, because the
mechanisms overlap and picking the wrong one is the expensive mistake.
Two things they say plainly that the API reference cannot:
Declaring and filling are separate everywhere, and filling never revisits a row
it has already done. That makes refreshes idempotent and appends cheap, and it
means a filled value does not track its inputs — surprising if you expect a
spreadsheet, so it is stated rather than left to be discovered.
The availability split runs the opposite way to expectation. Materialized views
are local only and raise on a remote connection; functions are Cloud and
Enterprise only and raise on a local one. Someone reading about both in one
sitting will otherwise assume the usual direction and be wrong in one of them.
Examples are the tested ones from the docstrings, so they cannot drift from the
behaviour they describe.
The training and tutorial pages that pointed into the geneva documentation now
point here.
The README that carried this sat inside docs/web, which is the published site
root, so it would have shipped as a page. Contributor guidance belongs in the
contributing guide, which is also where someone changing behaviour will look for
whether the page describing it is theirs to update.
The pages describing this code lived in lancedb/docs, a repository with ten
stars that neither contributing guide mentioned. A user who hit a wrong page
while reading the source had to find a separate, near-invisible repository to
fix it. These 102 pages now sit beside the code, and a documentation fix is a
pull request in the repository the reader already had open.
`docs/web` is a complete Mintlify site: `cd docs/web && mint dev` serves it with
no other checkout. Enterprise pages stay private and are merged in when the
published site is assembled, so links into them do not resolve here — every
open-source page does.
`docs/web-tests` holds the tests the code examples are extracted from. They ran
against a published wheel before; here they can run against this repository's
own SDK, so an example that stops compiling fails the pull request that broke
it rather than being noticed later.
Both directories sit beside `docs/src` and are invisible to it: mkdocs reads
`docs_dir: src`, so the SDK reference builds and deploys exactly as before.
Merge insert has always taken a list of columns to match on, and local
tables have always joined on all of them. Remote tables did not: any
list longer than one was rejected with `MergeInsertBuilder only supports
a single 'on' column`, so a composite-key upsert was impossible against
LanceDB Cloud and Enterprise from Rust, Python or TypeScript.
The remote request now carries `on` as a list and sends it as one
repeated query parameter per column — `?on=shard_key&on=id`. That is how
the lance-namespace spec encodes an array-valued `on`, so the server
receives a composite key in the shape it expects. A single column still
serializes to `?on=id`, exactly what clients sent before, so existing
callers are unaffected. A column repeated within `on` is now rejected
client-side rather than sent for the server to reject with a 400.
No binding changes were needed: `Table.merge_insert` in Python and
`Table.mergeInsert` in TypeScript already accepted a list, it just could
not reach a remote table. Both gain a test for composite keys, and the
doc comments now say what passing several columns means.
Part of
[ENT-2084](https://linear.app/lancedb/issue/ENT-2084/mergeinsertintotablerequest-support-multiple-columns-for-the).
## Example
```python
table.merge_insert(["shard_key", "id"]) \
.when_matched_update_all() \
.when_not_matched_insert_all() \
.execute(new_data)
```
A row whose `id` matches an existing row but whose `shard_key` differs
is an insert, not an update.
## Not included
Java. Java callers reach merge insert through
`org.lance.namespace.LanceNamespace`, whose
`MergeInsertIntoTableRequest.on` is a single string until
lance-namespace 0.12
([lance-namespace#363](https://github.com/lance-format/lance-namespace/pull/363),
[lance#8915](https://github.com/lance-format/lance/pull/8915)). There is
nothing in this repo's Java SDK to change until the `lance-core` pin can
move.
Sending more than one column requires a server that accepts the repeated
parameter ([sophon#7571](https://github.com/lancedb/sophon/pull/7571));
an older server returns a 400 rather than silently merging on one
column.
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Remote create-index requests already expose `replace` on the builder,
but the remote client did not consistently forward an explicit
`replace=false` over REST. That meant create-only intent could be lost
before it reached a remote server, even though local builders and Python
APIs can express it. This PR forwards `replace=false` on the existing
`create_index` endpoint and keeps the current default behavior unchanged
for compatibility.
This was accomplished with the following changes:
- Serialize `replace: false` into the existing remote create-index
request body when the builder is configured with `.replace(false)`.
- Forward `replace` through the synchronous Python remote `create_index`
wrapper so `RemoteTable.create_index(..., replace=False)` reaches the
repaired path.
- Continue omitting `replace` for the default path so existing remote
create-index requests keep their current semantics.
- Document `name` and `replace` on the existing OpenAPI create-index
request schema.
- Add coverage that verifies the remote client uses the existing
`/create_index/` route and forwards `replace=false`, including the
synchronous Python unified API.
### Testing
- `cargo fmt --all --check`
- `cargo test -p lancedb --features remote
test_create_index_forwards_replace_false_on_existing_route --locked`
- `uv tool run maturin develop --extras tests,dev,embeddings`
- `uv run --frozen pytest
python/tests/test_remote_db.py::test_remote_create_index_new_api`
- `uv run ruff format --check python/lancedb/remote/table.py
python/tests/test_remote_db.py`
- `cargo build -p lancedb --features remote --locked`
- `cargo clippy -p lancedb --features remote --all-targets --locked --
-D warnings`
A view definition recorded its source by bare name and refresh resolved
that name at the root, so declaring a view over a namespaced source was
refused outright -- materialized views were root-only for every caller.
The definition now carries `source_namespace`, and refresh opens the
source at that coordinate. `plan` takes the namespace too: refresh
re-plans the stored definition and persists the result when it migrates,
so defaulting it there would strand the view on its next rebuild.
The stored kind is the version boundary. Root definitions keep the
`select` form byte-for-byte, so everything written before this change
reads exactly as it always did. A namespaced source is stored as
`namespaced_select`: released readers drop unknown fields and resolve a
`select` source at the root, so keeping the old kind would let a
rolled-back worker refresh a view from a same-name root table -- the new
kind routes them to their existing unrecognized-kind refusal instead.
The Python and Node definition parsers learn the new kind alongside the
Rust core.
The bindings are built, installed and published with pnpm everywhere,
but a parallel npm dependency graph was still being maintained beside
it. This removes it, raises the supported Node floor to the versions we
actually test, and gives Dependabot the npm coverage it was missing.
## Dropping npm
`nodejs/package-lock.json` was regenerated by `ci/update_lockfiles.sh`
on every release commit and read by nothing — no workflow runs `npm ci`
or `npm install` in `nodejs/`, and npm never publishes a lockfile in a
package tarball. It could not even agree with the real install, since
npm does not see pnpm's `overrides`. Because GitHub's dependency graph
parses `package-lock.json`, it was also reporting vulnerabilities for a
tree we neither install nor ship.
`docs/package.json`, `docs/package-lock.json` and `docs/tsconfig.json`
go too. They depend on `file:../node` and
`file:../node/node_modules/apache-arrow` — the `node/` directory was
removed long ago — the tsconfig compiles `src/*.ts` where no TypeScript
files exist, and nothing installs any of it. `docs.yml` only referenced
the lockfile to configure an npm cache for an install it never ran.
Two `workflow_dispatch` workflows for regenerating those lockfiles are
removed as well. Both were already broken: they `uses:` composite
actions at `.github/workflows/update_package_lock{,_nodejs}` that do not
exist, so dispatching either failed immediately.
The remaining `npx` calls become direct `node_modules/.bin/...`
invocations. These were already running locally installed binaries
rather than resolving anything, but naming the binary removes the npm
CLI from the loop and does not depend on which Node version is active.
`dev.yml`'s commitlint check was the last place doing real npm
dependency resolution — an unpinned `npm install
@commitlint/config-conventional` that also bypassed the
`minimumReleaseAge` hold configured for `nodejs/` — and is now a pinned
`pnpm dlx`.
## Node support
Node 18 and 20 both reached end-of-life, in April 2025 and April 2026.
The matrix moves to 22, 24 and 26, and `engines` rises from `>= 18` to
`>= 22` so the declared floor is one the matrix actually covers. Node 22
is LTS until April 2027; 24 is LTS; 26 is Current and becomes LTS in
October 2026.
This also removes the reason the workflows reached for `npx` in the
first place: pnpm 11 requires Node >= 22.13, which every matrix version
now satisfies.
The prebuilt-binary smoke test in `npm-publish.yml` moves from Node 20
to Node 22 — the floor, where a napi ABI problem would surface first —
rather than fanning out across all three, to keep the publish matrix
from tripling.
## Dependabot
There were no npm-ecosystem entries at all, which is why the advisories
behind #4073 went unnoticed. Both pnpm lockfiles are now watched —
`nodejs/` and `nodejs/examples/`, which is a separate install — using
the same `lockfile-only` strategy as the existing cargo and pip entries,
so version ranges in `package.json` are left alone.
## Pre-commit biome
The hook ran `npx @biomejs/biome@1.8.3` while `nodejs/package.json`
resolved 1.9.4. The two disagree about formatting, so the hook rejected
code that `pnpm lint` accepts, and failed on unmodified `main` for
anyone touching `nodejs/`. It now uses the pnpm-managed biome, which
fixes the drift with no source changes.
## Testing
`dev.yml`'s commitlint job does not check out the repo, so it runs in an
empty workspace, and I could not verify `pnpm/action-setup` there
locally. It triggers on `pull_request_target`, so this PR exercises it
directly — worth confirming green before merge. I did verify the `pnpm
dlx` invocation itself locally: it accepts a conventional title and
rejects a non-conventional one with exit 1.
Node 26 is new enough that the examples job may surface gaps in prebuilt
native binaries (`onnxruntime-node`, `sharp`) before their maintainers
publish for it.
## Not included
`nodejs/examples/` still pins `sharp: "0.33.5"` and has its own audit
findings. Raising the Node floor unblocks that work — sharp 0.35
requires Node >= 20.9, which the matrix now satisfies — but it is a
dependency bump rather than tooling cleanup, so it is left separate.
## Breaking changes
`@lancedb/lancedb` now requires Node >= 22; previously >= 18. The
`@types/node` peer range moves from `>=18` to `>=22` to match. Users on
Node 18 or 20 must upgrade their runtime; both have been end-of-life for
some time. Existing installs are unaffected, since `engines` is only
checked on install.
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
In LanceDB Enterprise, we've adopted these conventions to give some
"canonical" metadata paths. This lets us display them in a certain way
in the UI or let agents standardize on them, to assume they'll find info
in a certain place. This PR (only comments/docs) just documents those
choices.
## Summary
LanceDB could not request Lance's list-element FTS document granularity
through Python or Remote APIs, and generic nested-field resolution
exposed Arrow's internal `item` segment instead of the public field
path.
This exposes typed `row | list_element` configuration for Python FTS
index creation and match/phrase queries, preserves `_doc_index`, and
keeps nested FTS paths public (for example, `docs.content`). Remote
list-element requests require server API version 0.6.0 so older servers
cannot silently execute them with row semantics; explicit row requests
remain compatible.
## Compatibility
Omitted index and query parameters retain row granularity. Remote row
index creation omits the new wire field.
## Tracking
[ENT-2342](https://linear.app/lancedb/issue/ENT-2342/expose-list-element-fts-document-granularity-end-to-end)
## Summary
- lazily initialize built-in OpenAI and Hugging Face providers when
consumers call the public embedding registry API
- choose automatic vector versus FTS search from embedding metadata on a
fresh pinned table revision for every execution
- expose automatic string searches as an `AutoQuery` with only
operations common to both native query families
- keep the registry shared and built-in registration safe across
duplicated module graphs
## Root cause
Nitro treats dependency modules as side-effect-free and removes the bare
OpenAI provider import from its generated route. Registration therefore
never runs, so `getRegistry().get("openai")` remains undefined even when
the registry itself is shared globally. Bundlers may also duplicate the
provider and registry module graphs.
The public embedding entry point now initializes built-in providers only
when `getRegistry()` is explicitly called, keeping initialization on a
live path that Nitro retains. Each terminal automatic-search execution
pins the exact table revision visible at dispatch, reads embedding
metadata and computes an embedding from that snapshot, replays the
builder operations, and constructs and executes the selected native
query against the same snapshot. Pinned native snapshots execute locally
when namespace pushdown cannot carry their revision, while remote
snapshots are seeded directly from one version-and-schema response. The
public `AutoQuery` builder exposes only the operations shared by FTS and
vector search, so runtime class narrowing cannot expose invalid
vector-only methods. Repeated built-in registration replaces stale
constructors from duplicated module graphs while public `register()`
retains its duplicate-alias error.
## Validation
- `cargo fmt --all`
- `cargo check --quiet --features remote --tests --examples`
- `cargo clippy --quiet --features remote --tests --examples`
- `pnpm build`
- `pnpm lint`
- `pnpm run docs`
- `pnpm test --runInBand` (783 passed, 5 skipped)
- serial examples suite with a local OpenAI mock (11 passed), including
`sentence-transformers.test.ts`
- packaged Nitro 2.13.4 server route using the reported imports returned
`{"registered":true}`
- fresh-process FTS fixture initialized both public built-ins and
confirmed automatic string search still returned the indexed row
- schema-consistency regressions cover read-consistency refresh,
checkout, checkoutLatest, restore, runtime class narrowing, concurrent
overwrite during embedding computation, and reused automatic-search
builders
- focused regressions confirm pinned native snapshots bypass unversioned
namespace pushdown and remote snapshots use one describe request
Fixes#2429
<!-- lance-gatekeeper-fix:v1 agent=2adf0f21b8bfb634606ed8897a849e30
generation=1 -->
---------
Co-authored-by: Gatefixer <313497061+lancedb-gatefixer[bot]@users.noreply.github.com>
Co-authored-by: Xuanwo <github@xuanwo.io>
`table_names` is being replaced by `list_tables` across the SDKs, but
TypeScript only had `tableNames`. This PR adds `listTables`, which
returns a page of table names together with the token that resumes after
it, and marks `tableNames` and `TableNamesOptions` deprecated in favor
of it.
It binds the `Connection::list_tables` that already exists, so nothing
in the Rust API changes and nothing existing breaks. `pageToken` is
documented as opaque rather than as a table name, since what resumes a
listing is the database's to decide — that keeps callers off a detail
that is going to change.
Stacked on #4040, which fixes a table being dropped at every page
boundary. The page-walking test here needs that fix to pass. Review the
last commit only until #4040 lands.
## Example
```ts
const names = [];
let pageToken = undefined;
do {
const page = await conn.listTables({ pageToken, limit: 100 });
names.push(...page.tables);
pageToken = page.pageToken;
} while (pageToken);
```
A namespace can be listed by passing its path first, mirroring
`tableNames`:
```ts
const page = await conn.listTables(["analytics"], { limit: 100 });
```
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
## Summary
- add `StreamingDataLoader`, which transports worker snapshots with
prefetched batches and commits them to the parent dataset only when the
trainer receives each batch
- preserve exact non-uniform per-split progress and resume lagging
splits without replaying already-consumed rows
- reject stale parent checkpoints after a standard multi-process
`DataLoader` has started, with guidance to use the consumer-aware loader
- document the new public loader and merge non-uniform state across
ranks
## Root cause
PyTorch runs `StreamingDataset.__iter__` in private worker-process
copies, while callers invoke `state_dict()` on the parent dataset.
Sharing producer counters would still be incorrect because DataLoader
prefetch can advance workers beyond batches returned to the trainer.
## Validation
- `uv run --extra tests pytest python/tests/test_elastic_dataloader.py
-q` (154 passed)
- focused non-uniform merge regression (1 passed)
- `uv run --project python --extra tests --extra dev ruff format .`
- `uv run --project python --extra tests --extra dev ruff check .`
- `cd docs && PYTHONPATH=. ../python/.venv/bin/mkdocs build`
Fixes#3967
<!-- lance-gatekeeper-fix:v1 agent=572be272619660b97e87fd5c85188341
generation=1 -->
---------
Co-authored-by: Gatefixer <313497061+lancedb-gatefixer[bot]@users.noreply.github.com>
Co-authored-by: Xuanwo <github@xuanwo.io>
Exposes materialized views to TypeScript: createMaterializedView,
openMaterializedView and listMaterializedViews on Connection, and a
MaterializedView handle carrying the parsed definition and
refresh({full, sourceVersion}), which returns the typed refresh result.
select accepts column names, [alias, expression] pairs, or a record of
the
same; the definition reads back off the stored schema, so a reopened
handle
needs no side channel. Remote connections surface the core's
not-supported
error up front.
The napi crate needed the same recursion-limit raise as the core crate:
the
refresh future's type graph overflows the default trait-recursion depth.
<sub>Stack created with <a
href="https://github.com/github/gh-stack">GitHub Stacks CLI</a> • <a
href="https://gh.io/stacks-feedback">Give Feedback 💬</a></sub>
This PR is a **breaking** rename of #3686.
merge reads like git merge w/ three-way, replay history, combine two
lines of work. That is not this API.
This call takes one additive change on a branch and lands it on main.
New column, including a blob column. Main's existing columns are not
rewritten. If it cannot land, you get `status="failed"` and
`diff.errors`, not a merge conflict to resolve.
Cherry-pick is terminology that aligns more with that.
```python
table = db.open_table("images")
table.branches.create("exp")
exp = table.branches.checkout("exp")
exp.add_columns({"tag": "cast('draft' as string)"})
diff = table.branches.diff("exp")
preview = table.branches.cherry_pick("exp", dry_run=True)
result = table.branches.cherry_pick("exp")
if result["status"] == "cherryPicked":
print("landed at", result["mainVersionAfter"])
elif result["status"] == "failed":
print(result["diff"]["errors"])
```
### Behavior
- Remote / Enterprise only. Local still NotSupported.
- HTTP 409 is not an exception. It is Ok with status="failed" and
diff.errors (CherryPickError).
- Unknown error / status codes still parse as Unknown.
- Requests are not retried. 409 is final and carries the body.
- Endpoint is POST /v1/table/{id}/branches/cherry_pick/.
- merge_insert and Table.merge are unchanged.
### Testing
- `cargo test -p lancedb --features remote diff_branch`
- `cargo test -p lancedb --features remote cherry_pick`
- `pytest python/python/tests/test_remote_db.py -k cherry_pick`
- node `remote.test.ts` diffs / cherry-picks path
Exposes materialized views to Python in both the async and sync clients:
create_materialized_view / open_materialized_view /
list_materialized_views
on the connections, and MaterializedView / AsyncMaterializedView handles
carrying the parsed definition and refresh(full=, source_version=),
which
returns the typed refresh result. select accepts column names, (alias,
expression) pairs, or a dict of the same; the definition reads back off
the
stored schema, so a reopened handle needs no side channel. Remote
connections raise NotImplementedError up front rather than failing deep
in
a request, matching the computed-column convention.
<sub>Stack created with <a
href="https://github.com/github/gh-stack">GitHub Stacks CLI</a> • <a
href="https://gh.io/stacks-feedback">Give Feedback 💬</a></sub>
Jack Ye
Lance Release
Will JonesandClaude Opus 5
Bruno Ramirez
Wyatt Alt
Drew
Dan Tasse
Xuanwo
lancedb-gatefixer[bot]andXuanwo