Merge main into blob concurrency repair

docs/src/java/java.md

@@ -14,7 +14,7 @@ Add the following dependency to your `pom.xml`:
 <dependency>
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-core</artifactId>
-    <version>0.40.0-beta.8</version>
+    <version>0.40.0-beta.11</version>
 </dependency>
 ```
 

docs/src/js/classes/Connection.md

@@ -557,9 +557,10 @@ on the returned job to know when cleanup has finished.
 abstract dropView(name, namespacePath?): Promise<void>
 ```
 
-Drop the view named `name`.
+Drop the view named `name` and wait for its definition to be deleted.
 
-The tables it reads are untouched: a view holds no rows of its own.
+The tables it reads are untouched: a view holds no rows of its own. Use
+[dropViewAsync](Connection.md#dropviewasync) to retain the cleanup job instead of waiting on it.
 
 #### Parameters
 
@@ -573,6 +574,30 @@ The tables it reads are untouched: a view holds no rows of its own.
 
 ***
 
+### dropViewAsync()
+
+```ts
+abstract dropViewAsync(name, namespacePath?): Promise<Job>
+```
+
+Start dropping the view named `name` and return the job deleting its
+definition, without waiting for completion.
+
+The name is free before this resolves. When nothing was bound to it, the
+returned job is already finished and has no id.
+
+#### Parameters
+
+* **name**: `string`
+
+* **namespacePath?**: `string`[]
+
+#### Returns
+
+`Promise`&lt;[`Job`](Job.md)&gt;
+
+***
+
 ### isOpen()
 
 ```ts
@@ -798,6 +823,28 @@ abstract openTable(
 
 ***
 
+### pauseJob()
+
+```ts
+abstract pauseJob(jobId): Promise<string>
+```
+
+Pause a server-side job by id.
+
+The job's workers drain and it stays parked until resumed. Resolves to
+"pausing", "already_paused", or "committing" -- a job finalizing its
+results cannot be parked; retry shortly.
+
+#### Parameters
+
+* **jobId**: `string`
+
+#### Returns
+
+`Promise`&lt;`string`&gt;
+
+***
+
 ### renameTable()
 
 ```ts
@@ -831,6 +878,28 @@ a "not supported" error.
 
 ***
 
+### resumeJob()
+
+```ts
+abstract resumeJob(jobId): Promise<string>
+```
+
+Resume a paused server-side job by id.
+
+Its workers pick their work back up from checkpoints. Resolves to
+"resumed", "still_pausing" -- the pause's worker drain is not confirmed
+yet; retry shortly -- or "not_paused".
+
+#### Parameters
+
+* **jobId**: `string`
+
+#### Returns
+
+`Promise`&lt;`string`&gt;
+
+***
+
 ### ~~tableNames()~~
 
 #### tableNames(options)

docs/src/js/classes/Table.md

@@ -548,6 +548,10 @@ Bytes for `column` at row IDs from [Query.withRowId](Query.md#withrowid).
 Reads the table's current checkout. IDs from another version can fail after
 compaction unless stable row ids are enabled. Results keep input order and
 duplicates. Null blobs are `null`. Empty blobs are empty buffers.
+Remote servers limit each request to 1024 row IDs and 64 MiB of blob bytes.
+The client splits requests automatically and reads an individual larger
+blob through the Range route. This method still materializes all bytes in
+memory; use [Table.fetchBlobFiles](Table.md#fetchblobfiles) for large values.
 
 #### Parameters
 

docs/src/js/interfaces/AddDataOptions.md

@@ -10,6 +10,22 @@ Options for adding data to a table.
 
 ## Properties
 
+### allowExternalBlobOutsideBases?
+
+```ts
+optional allowExternalBlobOutsideBases: boolean;
+```
+
+Whether blob URIs outside registered external bases may be written.
+
+Defaults to `false`. This option is supported only for local/native
+tables; remote tables return an error when it is enabled. An enabled write
+stores the absolute URI reference without registering a base or copying
+the external object into the database. The object must remain accessible
+when the blob is read later.
+
+***
+
 ### mode
 
 ```ts

docs/src/python/python.md

@@ -198,6 +198,33 @@ listing a storage directory.
 
 ::: lancedb.materialized_view.MaterializedViewDefinition
 
+::: lancedb.vector_dedup
+
+::: lancedb.materialized_view.VectorDedupSource
+
+For an existing indexed source, declare and materialize a dedup result without
+writing SQL. Creation captures the source version once and waits for the complete
+result; remote connections use the service's MV jobs.
+
+```python
+from lancedb import vector_dedup
+
+view = db.create_materialized_view(
+    "images_clean",
+    vector_dedup("images", column="phash", distance_threshold=4),
+)
+cleaned = view.table
+```
+
+To submit without waiting, call `db.create_materialized_view_async(...)` with the
+same source expression and wait on the returned job. To declare only, pass
+`with_no_data=True`, then use `view.refresh_async().wait()` when ready. Pass
+`dataset_version=...` to `vector_dedup` for an explicit snapshot. The default
+policy retains direct representatives (A-B-C without A-C keeps A and C), preserves
+the source, and materializes its original columns. Only within-segment/partition
+pairs are covered, using the index's distance representation.
+
+
 ## Views
 
 ::: lancedb.view.ViewDescription
@@ -379,6 +406,29 @@ still work. Queries return descriptors. Call
 
 ## Reranking
 
+`TypeSafeReranker` supports opt-in request batching:
+
+```python
+from lancedb.rerankers import TypeSafeReranker
+
+reranker = TypeSafeReranker(batch_size=40, max_concurrency=8)
+```
+
+The default `batch_size=1` keeps the query and document in request state and
+sends one request per non-null candidate. With `batch_size=40`, 80 non-null
+candidates require two requests. `max_concurrency` still limits simultaneous
+requests, and the SDK handles retries.
+
+In batched mode, state contains only `{"query": query}`. Each independent
+question contains `{"question": instructions, "document": document}` in its
+structured instructions, so it sees only its own document and the shared query.
+Custom instructions and criteria are kept verbatim: adapt prompts that explicitly
+reference request-state fields such as `state.document` before enabling batching.
+Null documents retain a zero score without an API call; empty strings are scored.
+API failures, mismatched answer IDs, and invalid probabilities raise errors.
+Batching changes the payload and can affect model scores; compare quality and
+latency on your workload before opting in.
+
 ::: lancedb.rerankers
     options:
       show_root_heading: false