Download Latest Version v1.2.0 source code.zip (26.9 MB)
Home / v1.1.0
Name Modified Size InfoDownloads / Week
Parent folder
README.md 2026-11-01 13.1 kB
v1.1.0 source code.tar.gz 2026-11-01 24.6 MB
v1.1.0 source code.zip 2026-11-01 25.3 MB
Totals: 3 Items   49.8 MB 0

What's Changed

Connector key fixes: no more collisions or lost writes

Every source item and target state has a key. CocoIndex uses the key to match an object with its previous run, and to decide what to create, update or delete. Prompted by the Google Drive fix in [#2485], this release audits how every Python connector builds its keys. It fixes each case where the connector gave two different objects one key, or where a key that changed for the same object made CocoIndex delete that object. Two keys you choose can also reach one object, such as overlapping directories or two ContextKeys for one database. CocoIndex can't detect that, so it stays your responsibility; the docs now point out where it can happen.

Two different objects shared one key, so one overwrote or deleted the other:

One object's key changed between runs, and the delete for the old key removed what the new key had just written:

A delete missed its object, leaving stale data behind:

A change went undetected:

Docs:

When you upgrade:

  • Deprecated, still working (with a DeprecationWarning, and with the same keys as before):
  • Neo4j and FalkorDB: table_target, declare_table_target, mount_table_target, relation_target, declare_relation_target, mount_relation_target and RelationTarget.declare_relation. Use node_target, declare_node_target, mount_node_target (they return a NodeTarget), relationship_target, declare_relationship_target, mount_relationship_target and RelationshipTarget.declare_relationship, with the same parameters. Switching an existing table takes two updates, described in the connector docs under "Switching an existing table"; skipping the first one deletes the table's nodes or edges. On the new APIs, Neo4j labels and relationship types can't contain __ or end with _, and derived edge ids need int or str endpoint ids (#2612)
  • Kafka: topic_as_map(consumer, [...]) with a list of topics. With one topic, pass it as a string; the keys don't change. With several, use one feed and one consumer per topic, each under its own subpath. That changes the item keys, so follow the replay steps in the Kafka docs (#2597)
  • Postgres: TableTarget(provider, table_schema). Pass the table name: TableTarget(provider, table_name, table_schema) (#2603, [#2615])
  • Behavior change: for S3, Azure Blob and OCI, prefix names a folder. prefix="docs" lists docs/..., but no longer docsa.md or docs-old/x, and target states from those objects are removed on the next update. To select objects by how their names start, use path_matcher (#2601)
  • New errors, for declarations that could never work correctly: DirTarget names that are absolute or leave the directory (#2598); an Any, object, union or bytes key column in a new zvec collection (#2608); LanceDB key column types that deletes can't filter on, such as dates, timestamps, float16 and lists (#2600); Postgres vector indexes whose truncated names collide (#2603); the Valkey field name vector (#2606). FilePaths of different types no longer compare equal, and ordering them raises TypeError (#2601)
  • One-time effects on the first update: Google Drive components get new keys and are reprocessed (#2485); objects that shared a key become separate items (#2601); SurrealDB relations declared without an explicit id are re-created under a new record ID (#2599); DirTarget files declared under a name that now normalizes (such as ./a.txt) are written once under the new name (#2598). To remove duplicate graph edges left by earlier versions, run one update with --full-reprocess (#2611)

Other changes

  • feat(fn): coco.LogicDep and coco.use_logic_dep() let a call declare a logic dependency for itself only, such as the version of the per-language extractor it used. Bumping that version re-runs only the calls that used it, and their memoized callers. Also fixed: a function-memo hit lost the logic deps its entry recorded, so a caller re-stored on that hit missed later changes to deeper callees by @georgeh0 in https://github.com/cocoindex-io/cocoindex/pull/2590
  • feat(batching): @coco.fn.as_async(batching=True, ...) takes max_in_flight, which bounds how many batches run at once, and rate_limiter, which charges one token per batch instead of per call. For a limiter kept on the instance, pass a callable that receives self by @georgeh0 in https://github.com/cocoindex-io/cocoindex/pull/2588 and https://github.com/cocoindex-io/cocoindex/pull/2589
  • feat(engine): live components share the max_inflight_components limit fairly, so one live component's large update (e.g. a full re-index) no longer holds back another's small refresh. An update that runs alone still gets the whole pool by @georgeh0 in https://github.com/cocoindex-io/cocoindex/pull/2624
  • perf(engine): moving many target states from one component to another in one commit is no longer quadratic (5,000 targets: 243 s before, 1 to 2 s now) by @georgeh0 in https://github.com/cocoindex-io/cocoindex/pull/2621
  • fix(memo): memo keys and fingerprints depend only on values. A structure that references one object twice fingerprints the same as one holding equal copies, and a temporary object freed during fingerprinting can no longer cause a wrong memo hit when Python reuses its id. One-time effect: for values with shared sub-objects, fingerprints change. Memoized work may run once more, generate_id / generate_uuid / IdGenerator / UuidGenerator may return new IDs (which changes row keys built from them), and rows tracked by fingerprint_object are written again by @hatemadi15 in https://github.com/cocoindex-io/cocoindex/pull/2411
  • fix(gpu_pool): a GPU released to a cancelled acquire goes back to the pool. Before, every later acquire of that GPU hung by @georgeh0 in https://github.com/cocoindex-io/cocoindex/pull/2595
  • fix(runtime): coco.stop() / coco.stop_blocking() release the lifespan-provided resources that registered apps held, instead of keeping them until each app's next update by @hatemadi15 in https://github.com/cocoindex-io/cocoindex/pull/2425
  • fix(environment): opening a db_path that is still open in the process raises a RuntimeError that names the path and the way out (Environment.close()), instead of a misleading message about "different options" by @Siddesh-bype in https://github.com/cocoindex-io/cocoindex/pull/2561
  • fix(connectorkits): cancelling a pending read of a synchronous source listing (localfs, Google Drive) no longer hangs asyncio.run() at shutdown by @Sujit-1509 in https://github.com/cocoindex-io/cocoindex/pull/2619
  • fix(serde): VectorSchema(dtype=np.float32) (the class, not a dtype) now reloads on the second run. Other classes the state reader can't load now fail when first written, with a hint to use coco.unpickle_safe. Before, the error came on the next run by @rtmalikian in https://github.com/cocoindex-io/cocoindex/pull/2536
  • fix(connectors): the repr of Snowflake and Doris connection configs leaves out the password by @jerryqu-coco in https://github.com/cocoindex-io/cocoindex/pull/2620
  • fix(oci_object_storage): zero-length reads, and partial reads of empty objects, return empty bytes instead of failing with HTTP 416 by @pranav440 in https://github.com/cocoindex-io/cocoindex/pull/2466
  • fix(code_ast): C# file-scoped namespaces (namespace X;) now show in the render_match context frames of the declarations after them by @georgeh0 in https://github.com/cocoindex-io/cocoindex/pull/2629
  • Docs fixes: examples that failed when run as written
  • docs(ops): the SeparatorSplitter usage example now runs as written by @March-7 in https://github.com/cocoindex-io/cocoindex/pull/2476
  • docs(localfs): the lifespan example uses coco.EnvironmentBuilder; coco.EnvBuilder doesn't exist by @Siddesh-bype in https://github.com/cocoindex-io/cocoindex/pull/2397
  • docs: the get_context example no longer calls the async coco.default_env() as if it were synchronous, and the App guide's component path tree matches the subpaths mount() derives by @hardness1020 in https://github.com/cocoindex-io/cocoindex/pull/2416

New Contributors

Full Changelog: https://github.com/cocoindex-io/cocoindex/compare/v1.0.26...v1.1.0

Source: README.md, updated 2026-11-01