milvus-io/milvus

Is there an existing issue for this? I have searched the existing issues. No directly matching open issue was identified in the catalog duplicate check. Recheck the target branch and in-flight PRs with a maintainer before claiming this task. Project Catalog ID: F01 Difficulty: easy Estimated incremental effort: 3–5 hours Make a bounded Lance dataset import runnable through the existing Parquet importer. Implementation scope Open an explicit dataset version and stream live rows/selected fields; map basic scalars and fixed-width FP32 vectors; reuse prepared upload/task-submission scaffolding and report source version/count/status. Acceptance criteria A fixture with deletions imports only live rows from the chosen version; IDs/vectors agree, dimension errors fail before submission and memory remains batch-bounded. Dependencies No numbered project prerequisite. The preparation conditions below still apply. Boundaries and preparation Existing Lance/Parquet tooling and upload scaffold r

Is there an existing issue for this? I have searched the existing issues. No directly matching open issue was identified in the catalog duplicate check. Recheck the target branch and in-flight PRs with a maintainer before claiming this task. Project Catalog ID: G02 Difficulty: easy Estimated incremental effort: 3–5 hours Connect a compatible rerank service without losing candidate/score alignment. Implementation scope Pin one supported request/response protocol; add a thin adapter or configuration reuse; demonstrate actual calls, timeout and cancellation. Acceptance criteria Returned indexes map to the correct candidates; missing, duplicate and out-of-range indexes are rejected. Dependencies No numbered project prerequisite. The preparation conditions below still apply. Boundaries and preparation Existing protocol and rerank infrastructure must be reusable. New protocol design is a separate task. The estimate assumes familiarity with the implementation language, a working build/te

Is there an existing issue for this? I have searched the existing issues. No directly matching open issue was identified in the catalog duplicate check. Recheck the target branch and in-flight PRs with a maintainer before claiming this task. Project Catalog ID: G01 Difficulty: easy Estimated incremental effort: 3–5 hours Provide a runnable text-to-embedding-to-search workflow using an existing compatible provider. Implementation scope Pin a deployed model/version; reuse existing provider and Function configuration; deliver startup settings, health checks and an end-to-end script. Acceptance criteria A real service embeds and searches text successfully; dimension mismatch and service unavailability fail clearly. Dependencies No numbered project prerequisite. The preparation conditions below still apply. Boundaries and preparation Model/service must already be ready. This is a thin integration/example, not a new provider platform. The estimate assumes familiarity with the implementa

Is there an existing issue for this? I have searched the existing issues. No directly matching open issue was identified in the catalog duplicate check. Recheck the target branch and in-flight PRs with a maintainer before claiming this task. Project Catalog ID: E02 Difficulty: easy Estimated incremental effort: 3–5 hours Make the approved plan explanation available through a usable client entry point. Implementation scope Reuse E01 and its approved response schema; connect one public API and a Milvus CLI entry point with text/JSON output, authorization and database scope. Acceptance criteria CLI and API agree for the same request. Invalid expressions and unauthorized requests preserve existing error behavior; no search is executed. Dependencies #53829 Boundaries and preparation Requires E01 and ready API/CLI extension points. No new cross-component protocol or runtime profiler within this estimate. The estimate assumes familiarity with the implementation language, a working buil

Is there an existing issue for this? I have searched the existing issues Environment - Milvus version:v3.0.0 - Deployment mode(standalone or cluster):cluster - MQ type(rocksmq, pulsar or kafka): kafka - SDK version(e.g. pymilvus v2.0.0rc2): - OS(Ubuntu or CentOS): - CPU/Memory: - GPU: - Others: Current Behavior 我的 Milvus 版本是 3.0。使用 Kafka 作为消息队列，同时还使用了第三方的对象存储服务。当我尝试将消息队列从 Kafka 切换到 Woodpecker 时，遇到了以下错误： Milvus 在通过 BulkPackWriterV2 将 Insert Log 写入对象存储时，关闭 PackedWriter 失败。对象存储在执行 CompleteMultipartUpload 操作时返回 HTTP 400，错误码为 EntityTooSmall。 错误信息表明：Multipart Upload 中存在小于对象存储最小限制（5 MB）的非最后一个分片，导致整个上传任务无法完成。 完整报错如下 [2026/08/18 03:31:52.229 +00:00] [WARN] [syncmgr/pack_writer_v2.go:178] ["failed to write inserts