fix(deps): update all dependencies

UserDefinedSparseModel::new takes the SparseModel family as a third argument: new(onnx_file, tokenizer_files, model).
Image embeddings for non-square inputs differ from earlier releases across all image models.
Qwen3-VL image embeddings differ from 5.16.1 through 6.1.0.
The candle models (Qwen3TextEmbedding, Qwen3VLEmbedding, NomicV2MoeTextEmbedding) now cache under FASTEMBED_CACHE_DIR (default .fastembed_cache) instead of ~/.cache/huggingface/hub. Existing weights are re-downloaded unless HF_HOME is set.
fastembed::InitOptions is a deprecated alias of TextInitOptions and now emits a warning.

Bug fixes

Qwen3-VL: restore per-token MRoPE frequencies, with a regression test.
Image preprocessing: aspect-preserving shortest_edge resize, correct width/height order for exact sizes, and CHW layout for padded center crops.
Sparse models return Error::OutputKeyMissing instead of panicking when the expected output tensor is missing.
Nomic v2 MoE runs in F16 and BF16.
BGE-M3 try_new_from_user_defined and try_new_from_path apply disable_cpu_fallback and dimension_overrides. try_new_from_path accepts either a directory or the .onnx file.
Tokenizer loading accepts pad_token as an AddedToken object and resolves the pad id from the vocabulary when config.json has no pad_token_id.
Broken rustdoc links fixed.

Hugging Face access

HF_TOKEN, or the huggingface-cli login token, is sent with every request.
The candle models honour FASTEMBED_CACHE_DIR, HF_HOME, HF_ENDPOINT and HF_TOKEN like the ONNX models, and resolve sharded checkpoints through model.safetensors.index.json.

API additions

SparseTextEmbedding::try_new_from_user_defined, plus UserDefinedSparseModel::with_idf_file.
NomicBertModel is re-exported with a public constructor for loading local weights.

Performance

SPLADE post-processing max-pools in place instead of allocating two vocab-sized copies per batch.
Tokenization and input building share one code path across text, sparse, BGE-M3 and rerank.

Toolchain

Minimum supported Rust version is declared as 1.88.
CI lints all targets with the hf-hub feature, compiles the doctests, and builds docs with warnings denied.
Dependabot now tracks GitHub Actions.

dtolnay/thiserror (thiserror)

`v2.0.21`

Compare Source

Fix parsing of generic unit variants in display expressions (#459)

Configuration

📅 Schedule: (UTC)

Branch creation
- At any time (no schedule defined)
Automerge
- At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

👻 Immortal: This PR will be recreated if closed unmerged. Get config help if that's undesired.

If you want to rebase/retry this PR, check this box

This PR has been generated by Mend Renovate CLI.

This PR contains the following updates: | Package | Type | Update | Change | |---|---|---|---| | [clap](https://github.com/clap-rs/clap) | dependencies | patch | `4.6.6` → `4.6.7` | | [fastembed](https://crates.io/crates/fastembed) ([source](https://github.com/Anush008/fastembed-rs)) | dependencies | major | `6.0.0` → `7.0.0` | | [thiserror](https://github.com/dtolnay/thiserror) | dependencies | patch | `2.0.20` → `2.0.21` | --- ### Release Notes <details> <summary>clap-rs/clap (clap)</summary> ### [`v4.6.7`](https://github.com/clap-rs/clap/blob/HEAD/CHANGELOG.md#467---2026-09-14) [Compare Source](https://github.com/clap-rs/clap/compare/v4.6.6...v4.6.7) ##### Features - *(derive)* Add `#[command(defer = <bool>)]` attribute to opt-in to lazy initialisation of subcommands </details> <details> <summary>Anush008/fastembed-rs (fastembed)</summary> ### [`v7.1.0`](https://github.com/anush008/fastembed-rs/releases/tag/v7.1.0) [Compare Source](https://github.com/Anush008/fastembed-rs/compare/v7.0.1...v7.1.0) ##### 🍕 Features - Add WebGPU (`webgpu`) feature-flag ([#305](https://github.com/anush008/fastembed-rs/issues/305)) ([6463f3f](https://github.com/anush008/fastembed-rs/commit/6463f3fadd154ba1f4d3c960fac505703fc30cbd)) - By [@centuryofimage](https://github.com/centuryofimage) - Split image preprocessing from inference ([#306](https://github.com/anush008/fastembed-rs/issues/306)) ([40a38c8](https://github.com/anush008/fastembed-rs/commit/40a38c866406175cd60e934d7c58a3e4a1706f5e)) - By [@centuryofimage](https://github.com/centuryofimage) ##### 🧹 Chores - **deps:** bump jlumbroso/free-disk-space from 1.3.1 to 2.0.0 in the github-actions group ([#304](https://github.com/anush008/fastembed-rs/issues/304)) ([f4e9dc1](https://github.com/anush008/fastembed-rs/commit/f4e9dc11cd80ffa47bebd8aee956198e554c0343)) - By [@dependabot](https://github.com/dependabot) ### [`v7.0.1`](https://github.com/Anush008/fastembed-rs/compare/v7.0.0...v7.0.1) [Compare Source](https://github.com/Anush008/fastembed-rs/compare/v7.0.0...v7.0.1) ### [`v7.0.0`](https://github.com/anush008/fastembed-rs/releases/tag/v7.0.0) [Compare Source](https://github.com/Anush008/fastembed-rs/compare/v6.1.0...v7.0.0) #### ⚠ BREAKING changes - `UserDefinedSparseModel::new` takes the `SparseModel` family as a third argument: `new(onnx_file, tokenizer_files, model)`. - Image embeddings for non-square inputs differ from earlier releases across all image models. - Qwen3-VL image embeddings differ from 5.16.1 through 6.1.0. - The candle models (`Qwen3TextEmbedding`, `Qwen3VLEmbedding`, `NomicV2MoeTextEmbedding`) now cache under `FASTEMBED_CACHE_DIR` (default `.fastembed_cache`) instead of `~/.cache/huggingface/hub`. Existing weights are re-downloaded unless `HF_HOME` is set. - `fastembed::InitOptions` is a deprecated alias of `TextInitOptions` and now emits a warning. #### Bug fixes - Qwen3-VL: restore per-token MRoPE frequencies, with a regression test. - Image preprocessing: aspect-preserving `shortest_edge` resize, correct width/height order for exact sizes, and CHW layout for padded center crops. - Sparse models return `Error::OutputKeyMissing` instead of panicking when the expected output tensor is missing. - Nomic v2 MoE runs in F16 and BF16. - BGE-M3 `try_new_from_user_defined` and `try_new_from_path` apply `disable_cpu_fallback` and `dimension_overrides`. `try_new_from_path` accepts either a directory or the `.onnx` file. - Tokenizer loading accepts `pad_token` as an `AddedToken` object and resolves the pad id from the vocabulary when `config.json` has no `pad_token_id`. - Broken rustdoc links fixed. #### Hugging Face access - `HF_TOKEN`, or the `huggingface-cli login` token, is sent with every request. - The candle models honour `FASTEMBED_CACHE_DIR`, `HF_HOME`, `HF_ENDPOINT` and `HF_TOKEN` like the ONNX models, and resolve sharded checkpoints through `model.safetensors.index.json`. #### API additions - `SparseTextEmbedding::try_new_from_user_defined`, plus `UserDefinedSparseModel::with_idf_file`. - `NomicBertModel` is re-exported with a public constructor for loading local weights. #### Performance - SPLADE post-processing max-pools in place instead of allocating two vocab-sized copies per batch. - Tokenization and input building share one code path across text, sparse, BGE-M3 and rerank. #### Toolchain - Minimum supported Rust version is declared as 1.88. - CI lints all targets with the hf-hub feature, compiles the doctests, and builds docs with warnings denied. - Dependabot now tracks GitHub Actions. </details> <details> <summary>dtolnay/thiserror (thiserror)</summary> ### [`v2.0.21`](https://github.com/dtolnay/thiserror/releases/tag/2.0.21) [Compare Source](https://github.com/dtolnay/thiserror/compare/2.0.20...2.0.21) - Fix parsing of generic unit variants in display expressions ([#459](https://github.com/dtolnay/thiserror/issues/459)) </details> --- ### Configuration 📅 **Schedule**: (UTC) - Branch creation - At any time (no schedule defined) - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox. 👻 **Immortal**: This PR will be recreated if closed unmerged. Get [config help](https://github.com/renovatebot/renovate/discussions) if that's undesired. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR has been generated by [Mend Renovate CLI](https://github.com/renovatebot/renovate).

Renovate force-pushed renovate/all from 66806dc78e

renovate/stability-days Updates have met minimum release age requirement

Details

to 9eba2809db

renovate/stability-days Updates have met minimum release age requirement

Details

renovate/artifacts Artifact file update failure

2026-09-24 05:13:01 +00:00

Compare

Renovate changed title from ~~chore(deps): update rust crate clap to v4.6.7~~ to fix(deps): update all dependencies

2026-09-24 05:13:02 +00:00

Renovate force-pushed renovate/all from 9eba2809db

renovate/stability-days Updates have met minimum release age requirement

Details

renovate/artifacts Artifact file update failure

to dd1aa7f341

renovate/stability-days Updates have met minimum release age requirement

Details

2026-10-01 05:12:17 +00:00

Compare

azer added 2 commits

2026-10-02 06:08:19 +00:00

Merge remote-tracking branch 'origin/master' into renovate/all 2e7dc8df00

fix: use TextInitOptions for fastembed 7 6eae3afd7f

fastembed 7 deprecates the InitOptions alias, which -D warnings turns into
an error.

azer merged commit 40d9c3807e into master

2026-10-02 06:08:21 +00:00

azer deleted branch renovate/all

2026-10-02 06:08:22 +00:00

azer referenced this pull request from a commit

2026-10-02 06:08:23 +00:00

Merge pull request 'fix(deps): update all dependencies' (#15) from renovate/all into master

azer referenced this pull request from a commit

2026-10-11 18:38:28 +00:00

Merge #15: update all dependencies

Rows
Columns

fix(deps): update all dependencies #15

Release Notes

v4.6.7

Features

v7.1.0

🍕 Features

🧹 Chores

v7.0.1

v7.0.0

⚠ BREAKING changes

Bug fixes

Hugging Face access

API additions

Performance

Toolchain

v2.0.21

Configuration

`v4.6.7`

`v7.1.0`

`v7.0.1`

`v7.0.0`

`v2.0.21`