589 Commits

Author SHA1 Message Date
George
0892291f75 new: add inference free splade (#652) 2026-07-22 22:32:30 +07:00
Dylan Couzon
8108e4467c add nomic-embed-vision-v1.5 (#651) 2026-07-20 17:55:48 -04:00
Esteban Yusunguaira
8a8ea4f42f ci: update all actions to node 24 (#636) 2026-05-25 09:52:43 -05:00
Esteban Yusunguaira
a499c313af ci: fix remaining actions on node 20 (#634) 2026-05-22 11:30:12 +07:00
Esteban Yusunguaira
fde1e0b361 ci: update github actions to node 24 (#633) 2026-05-20 16:33:49 +07:00
George
adfc1aef59 fix: add timeout for download from gcs (#629) 2026-04-21 15:32:52 +07:00
George
1ec283c744 fix: fix license (#625) 2026-04-15 14:59:45 +07:00
George
a6a4e375ca fix: use original jina de model instead of fp16 due to onnxruntime up… (#623)
* fix: use original jina de model instead of fp16 due to onnxruntime updates

* fix: update size in description
2026-04-15 12:53:10 +07:00
George
2b069d2fbd fix: check if model files exist before returning cached path (#624) 2026-04-10 14:11:02 +07:00
estebany-qd
21df54e3a4 ci: Pin all gh actions to commit SHAs (#617) 2026-03-30 20:31:46 +02:00
George
87678dd784 new: add gemmaembedding-300m (#592)
* new: add gemma embed

* refactor: rename builtin pooling normalized embedding to builtin sentence embedding
2026-03-25 15:22:16 +07:00
George Panchuk
6fa442b960 bump version to 0.8.0 v0.8.0 2026-03-23 23:30:02 +07:00
Alexey Masolov
52ebfba27c fix: respect HF_HUB_OFFLINE in download_model to avoid network calls (#614)
When HF_HUB_OFFLINE is set to a truthy value (1, true, yes, on),
download_model() should treat local_files_only=True to avoid any
network calls. Currently, even with the local-cache-first pass (which
may fail due to missing metadata), the retry loop still calls
download_files_from_huggingface() without local_files_only, which
triggers model_info() — a network API call that immediately fails in
offline mode. This causes an unnecessary fallback to GCS download from
storage.googleapis.com.

By setting local_files_only=True when HF_HUB_OFFLINE is enabled:

1. The HF local cache pass works if the model is cached
2. The retry loop skips the network-dependent HF path entirely
3. retrieve_model_gcs() only checks for local fast-* directories
4. No network calls are attempted at all

The truthy value check aligns with huggingface_hub's own parsing of
HF_HUB_OFFLINE, which accepts "1", "true", "yes", "on" (case-insensitive).

This is critical for air-gapped / restricted environments where both
HuggingFace and Google Cloud Storage are unreachable.

Made-with: Cursor
2026-03-23 22:40:03 +07:00
George
ea55268e01 fix: fix onnxruntime 1.24, uncap pillow (#611)
* fix: fix onnxruntime 1.24, uncap pillow

* fix: fix python3.10 onnxruntime version

* fix: fix onnxruntime for 3.14, update onnx dep
2026-03-13 00:50:10 +07:00
Kacper Łukawski
800f3887b7 Model: ModernVBERT/colmodernvbert (#588)
* Add ColModernVBERT to LateInteractionMultimodalEmbedding registry

* Implement image processing based on Idefics3ImageProcessor logic

* Fix padding support

* Implement ColModernVBERT logic

* Remove TODOs

* Handle empty pixel values with proper image_size

* Add ColModernVBERT tests

* Run pre-commit

* mypy fixes

* mypy fixes

* mypy fixes

* mypy fixes

* Fix typo in the class name

* Add processor_config.json to additional files

* Fix mypy errors

* Refactor onnx_embed_image

* Fix mypy errors

* fix: colmodernvbert tests and query processing

* fix: remove Union references

* fix: fix exit stack, update tests, implement token count

* fix: uncomment colpali in tests

* fix: lowercase models to cache

* fix: fix models to cache

* refactor: move colmodernvbert related onnx embed to its class

---------

Co-authored-by: George Panchuk <george.panchuk@qdrant.tech>
2026-01-09 18:52:50 +07:00
Bastian Hofmann
020d535f9c Update logo and favicon (#589) 2025-12-18 15:49:31 +01:00
George
685fd9b5a1 new: use cuda if available (#537)
* new: use cuda if available

* fix: fix warning msg

* fix: add missing import
2025-12-10 20:23:34 +07:00
George
b304a2aff0 new: drop python3.9, replace optional and union with | (#574)
* new: drop python3.9, replace optional and union with |

* new: remove python 3.9 from pyproject

* refactor: replace remaining union and optional with |

* new: remove optional and union in dataclasses

* fix: add typealias to numpy type

* new: replace union with | in token count
2025-12-10 19:01:01 +07:00
George
c715416361 fix: update colbert description (#534) 2025-12-09 19:22:54 +07:00
George
428381cb04 bump version to v0.7.4 v0.7.4 2025-12-05 18:38:45 +07:00
George
3511b08831 fix: numpy dropped 3.10 as of 2.3.0 (#585)
* fix: numpy dropped 3.10 as of 2.3.0

* fix: update poetry lock
2025-12-05 18:08:20 +07:00
George
b718cc6a88 new: add token count method (#583)
* new: add token count method

* fix: fix mypy

* fix: load model in token_count

* fix: remove debug code
2025-12-04 11:29:36 +07:00
George
2ba8990260 new: try unlocking huggingface hub and pillow (#582)
* new: try unlocking huggingface hub and pillow

* new: adjust pillow version for 3.9

* fix: fix pillow for python3.13
2025-12-01 18:21:28 +07:00
George
dab185fd9d fix: fix onnx version for python3.13 (#580) 2025-11-25 21:42:10 +07:00
George
ec0e3128ee new: expose some onnx session options (#578)
* new: expose some onnx session options

* fix: fix extra session options is None case

* fix: fix missing params

* new: add tests
2025-11-25 17:49:02 +07:00
George
44e332999c new: try loading models from cache before making any network calls (#577) 2025-11-25 12:07:50 +07:00
George
533b54cee5 tests: introduce model cache to tests (#573)
* tests: introduce model cache to tests

* fix: fix not cached model deletion

* new: do not run CI tests on mac os and windows on python 3.10-3.12

* fix: lowercase cache keys, bm25 caching

* tests: do not run parallel processing on all cpus in sparse text embed

* fix: fix models to cache names, do not run parallel=0

* fix: fix sparse embedding tests

* fix: bm42 language by lower case model name
2025-11-12 18:54:33 +07:00
George Panchuk
ba1f6053bd bump version to v0.7.3 v0.7.3 2025-08-29 14:15:25 +03:00
George
4dc76e3859 fix: fix colbert query postprocessing (#557)
* fix: fix colbert query postprocessing

* fix: improve colbert single embedding tests
2025-08-29 13:25:43 +03:00
George
6efe06b172 new: decouple colbert query and document tokenizer (#556) 2025-08-29 13:25:18 +03:00
George Panchuk
887239239b bump version to v0.7.2 v0.7.2 2025-08-25 16:29:23 +03:00
Kacper Łukawski
ca023be0c0 feat: MUVERA embeddings (#542)
* Implement MuveraEmbedding

* Add random generator parameter for reproducibility in MuveraEmbedding

* Document random_seed parameter

* Remove unnecessary module docstring from muvera_embedding.py

* refactor: clean up constructor parameters and improve formatting in MuveraEmbedding

* refactor: rename muvera_embedding.py to muvera.py and update related references

* feat: enhance MuveraEmbedding with multi-vector model support and improve parameter defaults

* feat: add embedding_size property to MuveraEmbedding

* feat: update MuveraPostprocessor to use model description for embedding size and add Jupyter notebook for MUVERA usage

* fix: fix types, doctest, rename variables, refactor (#545)

* fix: fix types, doctest, rename variables, refactor

* fix: fix python3.9 compatibility

* fix: make get_output_dimension protected

* Optimize muvera (#551)

* vectorize operations

* fix: fill empty clusters with dataset vectors

* rollback get_output_dimension

* fix: fix type hints

* fix: review comments

* tests: add tests

---------

Co-authored-by: George <george.panchuk@qdrant.tech>
2025-08-21 02:59:02 +03:00
Anush
d1ddc8142d docs: Updated README.md (#550) 2025-08-11 13:29:45 +05:30
Andrey Vasnetsov
faf3d9fe18 batch inference should return same shape as individual inference (#547) 2025-08-06 22:44:28 +02:00
George
acec31277b fix: fix mypy colpali (#533) v0.7.1 2025-06-16 11:55:29 +03:00
George Panchuk
cb902149c8 bump version to 0.7.1 2025-06-16 12:23:35 +04:00
George
a260022ae6 fix: propagate local files only and specific model path into embed parallel (#524) 2025-05-20 14:46:42 +04:00
George
d5da56299a new: add embedding size property (#521)
* new: add embedding size property

* fix: format exception message

* new: replace embedding size property with get_embedding_size classmethod

* fix: fix missed parts

* chore: fix docstrings

* new: add embedding_size property
2025-05-20 14:46:32 +04:00
George
4e5575f4a7 fix: raise exception if pooling is incorrect in custom model (#522) 2025-05-19 13:49:22 +03:00
George
4736f46548 fix: check lowercase model name for warnings (#523) 2025-05-19 13:49:12 +03:00
Andrey Vasnetsov
c85e8c278f remove unused function (#516) v0.7.0 2025-05-13 17:04:51 +03:00
George
0df605fc92 fix: remove onnxruntime cap as not needed anymore (#517)
* fix: remove onnxruntime cap as not needed anymore

* fix: fix mypy complaints in python3.13
2025-05-13 17:54:54 +04:00
Andrey Vasnetsov
04bc7a3039 MiniCOIL v1 (#513)
* add token embeddings

* fix parallel worker init

* implement minicoil

* fix mypy

* fix mypy

* register minicoil

* rollback suggested "fix"

* add minicoil test

* some pr issues (#514)

* some pr issues

* revert query embed refactor

* test: add query embed tests

* nit

* Update tests/test_sparse_embeddings.py

---------

Co-authored-by: Andrey Vasnetsov <andrey@vasnetsov.com>

* review

* fix: revert change to colbert query_embed

---------

Co-authored-by: George <george.panchuk@qdrant.tech>
2025-05-13 14:22:50 +02:00
George
b785640bd5 fix: fix list of supported bm25 languages (#506) 2025-04-15 14:16:09 +03:00
Hossam Hagag
5568a62c2f ci: Unlock numpy in ci (#504) 2025-04-11 13:00:54 +03:00
George Panchuk
b34209dcfb bump version to 0.6.1 v0.6.1 2025-04-10 16:23:54 +03:00
George
c91d42dda7 Update setting jina v3 tasks (#503)
* new: improve task setter in jina v3

* refactor

* new: add hf_token secret

* fix: cross platform env propagation
2025-04-10 16:19:41 +03:00
George
aa0c475a1f fix: fix splade name (#499) 2025-03-17 11:07:14 +03:00
Dmitrii Ogn
4c239b11d5 Custom rerankers support (#496)
* Custom rerankers support

* Test for reranker_custom_model

* test fix

* Model description type fix

* Test fix

* fix: fix naming

* fix: remove redundant arg from tests

* new: update readme

---------

Co-authored-by: George Panchuk <george.panchuk@qdrant.tech>
2025-03-16 20:02:58 +03:00
Hossam Hagag
6acfb001fb Speedup ci (#489)
* chore: Trigger CI test

* chore: Trigger CI test

* chore: Trigger CI test

* chore: Trigger CI test

* chore: Trigger CI test

* chore: Trigger CI test

* chore: Trigger CI test

* chore: Trigger CI test

* chore: Trigger CI test

* Trigger CI

* Trigger CI

* Trigger CI

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* Trigger CI test

* new: Added on workflow dispatch

* tests: Updated tests

* fix: Fix CI

* fix: Fix CI

* fix: Fix CI

* improve: Prevent stop iteration error caused by next

* fix: Fix variable might be referenced before assignment

* refactor: Revised the way of getting models to test

* fix: Fix test in image model

* refactor: Call one model

* fix: Fix ci

* fix: Fix splade model name

* tests: Updated tests

* chore: Remove cache

* tests: Update multi task tests

* tests: Update multi task tests

* tests: Updated tests

* refactor: refactor utils func, add comments, conditions refactor

---------

Co-authored-by: George Panchuk <george.panchuk@qdrant.tech>
2025-03-06 04:39:58 +02:00