Marqo

New Features
- Global Score Modifiers for Hybrid Search ([1082](https://github.com/marqo-ai/marqo/pull/1082)). Introduce global score modifiers for hybrid search, allowing fine-tuned adjustments to RRF scores in combined result lists. This enhancement provides better control over returned results in hybrid search scenarios. Use the `rerankDepth` parameter to control the number of hits to rerank. For detailed usage, check [here](https://docs.marqo.ai/latest/reference/api/search/search/#rerank-depth).

- Custom LanguageBind Model ([1072](https://github.com/marqo-ai/marqo/pull/1072)). Marqo now supports loading custom LanguageBind models from S3 buckets, URLs, or HuggingFace model cards. Fine-tune your own LanguageBind model and integrate it with Marqo to achieve better in-domain results. For more details, see [here](https://docs.marqo.ai/2.15/models/marqo/bring-your-own-model/#bring-your-own-languagebind-models).

- Add Marqtune models to the model registry ([1063](https://github.com/marqo-ai/marqo/pull/1063)). Marqtune models have been added to the model registry, with some models renamed to align with the Marqtune naming convention. This update improves consistency and makes it easier to identify models Marqo and Marqtune. The changes are fully backwards compatible.

Bug Fixes and Minor Changes
- Fix a bug where Marqo incorrectly inferred the modality of a field, even when the field is not a tensor field in an unstructured index ([1086](https://github.com/marqo-ai/marqo/pull/1086)).
- Resolve an issue where Languagebind models could only handle a single video or audio in a weighted search query. You can now provide multiple videos and audios in a weighted query seamlessly.([1072](https://github.com/marqo-ai/marqo/pull/1072)).
- Improve memory usage when indexing image documents with LanguageBind models, enabling more efficient handling of image data. ([1072](https://github.com/marqo-ai/marqo/pull/1072)).

Contributor Shout-Outs
- A huge thanks to our 4.7k stargazers for your continued support!
- Thanks a lot for the discussion and suggestions in our community. Join us on [Slack](https://join.slack.com/t/marqo-community/shared_invite/zt-2jm456s90-1pFxdE5kDQt5imqddXUIcw) and our [forum](https://community.marqo.ai/) today!

New features

- Add support for [hf_transfer](https://github.com/huggingface/hf_transfer) to accelerate the downloading speed of HuggingFace models by 10 to 30 times. See [here](https://docs.marqo.ai/latest/other-resources/guides/advanced-usage/configuration/#third-party-environment-variables) for details about how to enable it ([1066](https://github.com/marqo-ai/marqo/pull/1066)).
- Add `/healthz` endpoint for Marqo container liveness checks, which performs a status check for CUDA devices and returns a 500 error if any existing CUDA devices become unavailable or run out of memory ([1068](https://github.com/marqo-ai/marqo/pull/1068)).

Bug fixes and minor changes

- Fix a bug where numeric map fields are not returned when searching with `attributes_to_retrieve` parameter for unstructured indexes created prior to Marqo 2.13 ([1062](https://github.com/marqo-ai/marqo/pull/1064)).
- Fix a bug where numeric fields, numeric map fields, boolean fields and string array fields are not returned when searching with `attributes_to_retrieve` parameter for unstructured indexes created with Marqo 2.13 or later ([1062](https://github.com/marqo-ai/marqo/pull/1064)).
- Fix a bug where `document-processing` element is removed from the `services.xml` config file when bootstrapping the vector store ([1075](https://github.com/marqo-ai/marqo/pull/1079)).

New features

- FFmpeg-CUDA Support ([1030](https://github.com/marqo-ai/marqo/pull/1030)). Add GPU acceleration for video decoding by integrating FFmpeg with CUDA support.
This feature significantly improves video processing performance, making video handling up to 5 times faster. Check [here](https://docs.marqo.ai/2.14/other-resources/guides/advanced-usage/configuration/#marqo-video-gpu-acceleration-configuration) for guidance and requirements.

- Video and audio file size limits ([1012](https://github.com/marqo-ai/marqo/pull/1012)). Introduce configurable size limits for video and audio files in the add_documents, search, and embed endpoints. This enhancement allows users to manage and optimize resource usage effectively, ensuring smoother processing of multimedia content. Check [here](https://docs.marqo.ai/2.14/other-resources/guides/advanced-usage/configuration/#configuring-usage-limits) for more details.

- Upgrade to Python 3.9 ([1006](https://github.com/marqo-ai/marqo/pull/1006)).
Upgrade the Marqo Docker image to use Python 3.9. With Python 3.8 reaching its End of Life (EOL), we have upgraded our platform to Python 3.9 to maintain security, compatibility, and access to ongoing support.


Bug fixes and minor changes

- Move NLTK resource downloads to Marqo's startup process and remove the unsafe punkt package. This avoids potential code start issues and enhances security ([1040](https://github.com/marqo-ai/marqo/pull/1040)).

- Fix model serialization in OpenAPI specifications. This community-contributed PR resolves issues with OpenAPI spec generation and SwaggerUI by fixing incorrect type hints in the API definition, ensuring accurate model serialization and improving API documentation accessibility ([986](https://github.com/marqo-ai/marqo/pull/986)).

- Add brief description for each endpoint in the OpenAPI specifications. This improves API documentation clarity for users ([1042](https://github.com/marqo-ai/marqo/pull/1042)).

Contributor shout-out
- Shoutouts to our valuable 4.7k stargazers!
- Thanks a lot for the heated discussion and suggestions in our community. We love to hear your thoughts and requests. Join our [Slack channel](https://join.slack.com/t/marqo-community/shared_invite/zt-2jm456s90-1pFxdE5kDQt5imqddXUIcw) and [forum](https://community.marqo.ai/) now.
- Special thanks to community contributor [gabauer](https://github.com/gabauer) for their impactful PR, helping improve Marqo for everyone!

Bug fixes and minor changes

- Fix a bug where `document-processing` element is removed from the `services.xml` config file when bootstrapping the vector store ([1075](https://github.com/marqo-ai/marqo/pull/1075)).

Bug fixes and minor changes

- Fix a bug where numeric map fields are not returned when searching with `attributes_to_retrieve` parameter for unstructured indexes created prior to Marqo 2.13 ([1062](https://github.com/marqo-ai/marqo/pull/1062)).
- Fix a bug where numeric fields, numeric map fields, boolean fields and string array fields are not returned when searching with `attributes_to_retrieve` parameter for unstructured indexes created with Marqo 2.13 or later ([1062](https://github.com/marqo-ai/marqo/pull/1062)).

Bug fixes and minor changes

- Fix a bug where adding documents with numeric lists to an unstructured index results in a 500 error. Now, Marqo successfully processes the document batch, and returns a 400 error only for individual documents that contain numeric lists([1034](https://github.com/marqo-ai/marqo/pull/1034)).
- Fix validation of custom vector fields. Custom vector fields were silently ignored when not specified as tensor fields for an unstructured index. This will now trigger a 400 error. This helps guide users to properly define the field as a tensor field([1034](https://github.com/marqo-ai/marqo/pull/1034)).
- Improve the bootstrapping process to prevent Marqo from crashing during startup when the vector store takes longer to converge, especially with multiple indexes. This ensures a smoother startup process even if the vector store takes time to fully initialize([1036](https://github.com/marqo-ai/marqo/pull/1036)).