v0.7.0

## Highlights
- **New CLI:** `mistralrs-cli`
- **Prefix Caching:** We have implemented Prefix Caching for PagedAttention (#1750). This significantly accelerates multi-turn conversations and RAG workflows by reusing KV cache for shared prompt prefixes.
- **Major model expanstion:** Support for the Embedding Gemma, Qwen 3 Embedding, Gemma 3n, GLM-4, Granite Hybrid MoE, GLM-4 MoE, GLM-4 MoE Lite
- **Dynamic model loading:** Dynamic Model Loading: The server now supports loading and unloading models at runtime (#1828)
- **Performance:** Added support for CUDA 13.0/13.1 (#1767) and introduced highly optimized fused kernels (GEMV, GLU) and blockwise FP8 kernels for significant speedups on NVIDIA GPUs.
- **`candle` 0.9.2:** We have migrated to the official crates.io release of candle 0.9.2, stabilizing our backend dependencies!

## New Models & Architectures
- **Embedding models:** Qwen 3 Embedding, Embedding Gemma
- **Text models:** GLM-4, GLM-4.7 Flash, Granite Hybrid, GPT-OSS
- **Vision models:** Gemma 3n, Qwen 3 VL & Qwen 3 VL MoE

## What's Changed
* Improve automatic tool call by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1460
* chore: `Dockerfile.cuda-all` configurable threads by @polarathene in https://github.com/EricLBuehler/mistral.rs/pull/1458
* chore: `Dockerfile.cuda-all` - Merge `RUN` for `apt-get install` by @polarathene in https://github.com/EricLBuehler/mistral.rs/pull/1459
* Add fallback definition for metal::isnan by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1463
* chore: `Dockerfile` - Drop runtime rayon thread ENV by @polarathene in https://github.com/EricLBuehler/mistral.rs/pull/1465
* Remove duplicate calls for api_dir_list by @guoqingbao in https://github.com/EricLBuehler/mistral.rs/pull/1474
* Fix transient pyo3 dep in `mistralrs_mcp` by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1478
* Fix objc dep when non macos by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1480
* Fix phi4 mini + nccl cache issue by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1481
* Fix phi3.5 moe (#1447) by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1482
* Support GLM4 model by @guoqingbao in https://github.com/EricLBuehler/mistral.rs/pull/1437
* Refactor distributed backend by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1484
* Cap metal paged attn kv allocation by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1485
* Better paged attn metal cap by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1486
* Server core: consolidate and unify route handlers and API surface by @matthewhaynesonline in https://github.com/EricLBuehler/mistral.rs/pull/1423
* Support qwen3 gguf by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1488
* Make bos/eos token IDs optional by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1493
* Remove python deps from CUDA dockerfiles by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1487
* Handle noncontiguous v in naive_sdpa by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1499
* Server Core: refactor Paged Attention configuration (breaking change) by @matthewhaynesonline in https://github.com/EricLBuehler/mistral.rs/pull/1500
* Use `StorageModePrivate` for Metal PA kv cache by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1506
* fix(stream): emit  field in tool-call deltas for schema compliance by @Sbargaoui in https://github.com/EricLBuehler/mistral.rs/pull/1507
* PagedAttention kv-cache quantization (F8E4M3) by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1400
* Validate model name in OpenAI API by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1509
* Fix mcp import in doc string by @GaetanLepage in https://github.com/EricLBuehler/mistral.rs/pull/1510
* Add multi-model support by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1512
* Add stars label to readme by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1513
* Handle base_model.model case in lora by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1514
* Add thread_local! for engine-specific const/static by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1517
* Fix MCP doc test by @GaetanLepage in https://github.com/EricLBuehler/mistral.rs/pull/1511
* Allow disabling metal precompilation by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1518
* Rust 1.88 clippy by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1522
* Fix cuda warnings by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1526
* Fix panic on error decoding tokens by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1527
* Split Marlin and Paged Attention kernels for faster build by @guoqingbao in https://github.com/EricLBuehler/mistral.rs/pull/1525
* chore: update llguidance by @ammar-elsabe in https://github.com/EricLBuehler/mistral.rs/pull/1535
* Add the SmolLM3 model by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1501
* Add Gemma 3n support by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1519
* Fix sequence length check by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1546
* update candle version by @AlpineVibrations in https://github.com/EricLBuehler/mistral.rs/pull/1545
* Add the capability to build for ios by @rubiktubik in https://github.com/EricLBuehler/mistral.rs/pull/1548
* chore: Dockerfile - Remove redundant symlink creation + ENV by @polarathene in https://github.com/EricLBuehler/mistral.rs/pull/1504
* Apply changes from Gemma 3n weight & config reupload by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1553
* Faster multimodal merging for gemma3n by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1558
* Improved Rust UQFF api by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1563
* Uqff minor tweaks and optimizations by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1565
* Update Candle backend, support new dtypes, CUDA 12.9 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1566
* Fix nccl regression by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1569
* Fix metal regression by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1575
* Initial support for OpenAI Responses API by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1580
* Sanitize returned server errors by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1581
* Fix invalid HTML warning by @szepeviktor in https://github.com/EricLBuehler/mistral.rs/pull/1583
* Make typos configuration stricter by @szepeviktor in https://github.com/EricLBuehler/mistral.rs/pull/1582
* fix: use try_init when initialize tracing by @christer-eriksson in https://github.com/EricLBuehler/mistral.rs/pull/1588
* Add blockwise fp8 quantize kernels by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1586
* Support old gemma3n intermediate size config by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1594
* Handle when there are invalid vision tower weights by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1595
* Fix needless bool warning by @jncraton in https://github.com/EricLBuehler/mistral.rs/pull/1599
* Use smaller model in streaming example by @jncraton in https://github.com/EricLBuehler/mistral.rs/pull/1598
* Simplify streaming example output by @jncraton in https://github.com/EricLBuehler/mistral.rs/pull/1597
* Add vector 1xK fp8 kernels by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1600
* Target generic CPUs when building Python package by @jncraton in https://github.com/EricLBuehler/mistral.rs/pull/1596
* Fix tests in CI by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1603
* Add tiktoken -> `Tokenizer` conversion utilities by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1604
* Reworked attention chunking by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1591
* Fix cuda ring compilation by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1608
* mistralrs-quant: Fix build when `feature=+cuda,-ring`. by @ryanli in https://github.com/EricLBuehler/mistral.rs/pull/1611
* chore: Dockerfile - Add `SHELL` instruction by @polarathene in https://github.com/EricLBuehler/mistral.rs/pull/1612
* Send the mcp server an initialization notification to let it know the client is done initializing plus finish off making all 3 supported mcp transports properly increment their request ids by @sonicrules1234 in https://github.com/EricLBuehler/mistral.rs/pull/1614
* Add Claude Code GitHub Workflow by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1616
* Add MXFP4 gather gemm support by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1615
* Update ISQ Python example by @jncraton in https://github.com/EricLBuehler/mistral.rs/pull/1617
* Rust 1.89 clippy by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1621
* Support Qwen3 MoE GGUF model with fast MoE kernel by @guoqingbao in https://github.com/EricLBuehler/mistral.rs/pull/1622
* Disable CUDA event tracking, use cudarc 0.17 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1623
* Enforce workspace msrv by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1631
* Fix bench async crash by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1632
* Bump tracing-subscriber from 0.3.19 to 0.3.20 by @dependabot[bot] in https://github.com/EricLBuehler/mistral.rs/pull/1633
* Update MCP server guide source by @emmanuel-ferdman in https://github.com/EricLBuehler/mistral.rs/pull/1635
* Batch of bugfixes by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1634
* Add `repetition_penalty` to `SamplingParams`. by @ryanli in https://github.com/EricLBuehler/mistral.rs/pull/1638
* Rust 1.90 clippy by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1639
* Drain dummy run reciever to fix sender dropping by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1645
* Handle cpu dtype requirements for conv by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1650
* Audio processing functions have been added: normalize, apply_fade, an… by @semihpolat in https://github.com/EricLBuehler/mistral.rs/pull/1572
* Handle case where gemma 3n q != (k=v) devices by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1653
* Refactor/simplify paged attn modules by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1654
* Fix engine busyloop by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1655
* Include stubs in maturin source builds by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1656
* Implement Qwen 3 VL! by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1657
* Refactor flash attn dispatch to handle CPU by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1671
* Fix cpu flash attn mask case by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1672
* Remove restriction on qwen vl batch size by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1673
* Fix gemma3 example by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1674
* feat: Remove ram limits for CPU by @anchpop in https://github.com/EricLBuehler/mistral.rs/pull/1675
* Fix hang and performance drop with Metal by @parasyte in https://github.com/EricLBuehler/mistral.rs/pull/1662
* Fix inverted logic bug in DRY sampler sequence breaker encoding by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1679
* Fix panic in prompt token truncation logic by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1678
* Slightly faster mistralrs-vision normalize path by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1681
* Correctly handle tied embeddings in qwen3_vl config by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1682
* Improved flexibility for Topology by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1683
* Temporarily disable fast sampler by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1685
* Support embedding models: EmbeddingGemma by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1684
* Implement Qwen 3 Embedding by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1686
* Fix apply chat template tool call case by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1689
* Fixes for qwen 2.5 vl by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1690
* Remove `once_cell` dependency from multiple crates for improved performance and consistency. by @htiennv in https://github.com/EricLBuehler/mistral.rs/pull/1691
* Ensure paged attn is disabled for embedding models by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1693
* Add example for batching in embedding models by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1694
* Expose max seqlen api on Engine by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1695
* Support cuda 13.0 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1697
* Refactor search embedding to use EmbeddingGemma by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1698
* Fix flash attn on cuda 13.0 build by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1704
* Fix embedding inputs processor in search by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1705
* Fix auto loader confusion by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1707
* Fixes for qwen 2.5 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1708
* Fix server hang with mcp by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1706
* Cuda clippy fixes by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1709
* Docs: Clarify correct GPU arch for FlashAttention  by @llamaha in https://github.com/EricLBuehler/mistral.rs/pull/1712
* Add custom logging example by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1719
* Respect silent setting in progress bar by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1720
* Fix overcounting on nonmapped params in device mapping by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1721
* Misc bugfixes by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1730
* Support the IBM Granite Hybrid MoE models! by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1731
* Fix isnan metal build failure by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1732
* Fix toplevel quantization config location by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1733
* Switch candle source to HF main! by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1722
* Optimize FP8 KV Cache with dedicated scale_update kernel by @guoqingbao in https://github.com/EricLBuehler/mistral.rs/pull/1651
* Support Fused MoE for unquantized Qwen3 Models by @guoqingbao in https://github.com/EricLBuehler/mistral.rs/pull/1723
* Add metal kv scale update kernels by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1736
* Update cudarc version 0.17 -> 0.18 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1745
* Bucket and preempt paged attn sequences by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1746
* Optimized cpu-side sampling routines by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1747
* Support qwen 3 vl moe by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1748
* Implement prefix caching for paged attn by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1750
* Webchat various improvements, fixes by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1751
* Add cudnn ubi9 dockerfile for rhel based systems by @lmzuccarelli in https://github.com/EricLBuehler/mistral.rs/pull/1742
* Fix FP8 handling for stacked format MoE experts by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1755
* No longer getting cache issue from HF runs by @Cooksey99 in https://github.com/EricLBuehler/mistral.rs/pull/1752
* Use mmap for more cases during loading by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1756
* Add blockwise fp8 gemm and grouped gemm by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1759
* Update candle dep by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1764
* [Granite Hybrid] Fix Mamba softplus NaNs, GPU mismatches, paged_attn by @haricot in https://github.com/EricLBuehler/mistral.rs/pull/1761
* Support cuda 13.1 (cudarc 0.18.2) by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1767
* Fix cc 75 moe kernel compilation by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1768
* Gemma3n audio encoder various fixes by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1769
* Support optimized AFQ on CUDA by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1770
* Use candle 0.9.2-alpha.2 from crates.io! by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1775
* Fix candle dep, flash-attn doesnt build on alpha.2 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1779
* Support GPT-OSS and harmony! by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1760
* Add specialized GEMV cuda kernels by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1788
* Add fused GLU kernels by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1789
* Add GB10 gpu runner by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1790
* Support cuda integrated memory systems by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1791
* Fix is_prefill case for MoEExperts by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1792
* Update candle dep by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1795
* Update reqwest 0.12 -> 0.13 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1796
* Add utils.metal include to fused_glu.metal by @vigsterkr in https://github.com/EricLBuehler/mistral.rs/pull/1793
* fix #1794 by @vigsterkr in https://github.com/EricLBuehler/mistral.rs/pull/1797
* Add cuda moe gemv kernel by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1798
* Integrate custom gguf indexed_moe kernels by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1800
* Fix shared memory mismatch on v100 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1802
* chore(deps): update candle revision by @vigsterkr in https://github.com/EricLBuehler/mistral.rs/pull/1801
* More complete docs by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1803
* Add builtin agentic loop by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1807
* Automatic Reference Counting for PagedAttention Blocks by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1808
* Fixes for x-lora models by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1812
* Bump cudarc to 0.18.2, support cuda 13.1 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1813
* Support cuda 13.1 in mistralrs-quant by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1814
* Support Ministral 3 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1816
* Fix UQFF generation on CUDA by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1817
* Fix cublaslt device management on multi-gpu by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1819
* Fix nccl vision model case by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1820
* Hotfix: Remove fastmath from candle-kernels by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1821
* Support the OpenResponses specification by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1822
* Fix missing default for gemma 2 tie word embeddings by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1823
* Attempt to improve metal paged attn stability by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1824
* Fix stack buffer overflow in cuda topk_softmax by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1825
* Metal normalization accumulate in f32 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1826
* Support model loading/unloading by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1828
* Clarify paged attn prefix caching/sequence prefix caching by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1829
* Support the GLM 4 Flash model! by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1830
* Replace serde_yaml with serde_saphyr by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1832
* Switch to cudarc 0.19, fix d2d cuda copies by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1833
* Improve mistralrs-mcp API with connevtion/tool/lifecycle management, update MCP schema to  by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1834
* Update most deps to latest versions by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1835
* reuses cuda devices instead of instantiating multiple main contexts by @krampenschiesser in https://github.com/EricLBuehler/mistral.rs/pull/1818
* Add MLA attention decode kernels for DSv2/v3 and GLM4.7 Flash by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1837
* Switch deepseek v2 and v3 to new expert implementation by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1838
* Use crates.io candle 0.9.2! by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1839
* Fix doc mistakes by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1840
* Support auto pipeline for diffusion and speech models by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1841
* Support aliasing for multi models by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1843
* v0.7.0 preparation: mistralrs-cli, revamped documentation, readme update by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1848
* Use feature flags to handle fp8 gating in CUDA by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1849
* Handle xcode 26+ default metal standard by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1846
* Handle missing metal toolchain in build script, troubleshooting docs by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1850
* Support fp8 in llama 4 models by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1851
* Simplify mistralrs rust sdk exposed items by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1852
* Final changes for v0.7.0 by @EricLBuehler in https://github.com/EricLBuehler/mistral.rs/pull/1853

## New Contributors
* @Sbargaoui made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1507
* @GaetanLepage made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1510
* @ammar-elsabe made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1535
* @AlpineVibrations made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1545
* @rubiktubik made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1548
* @christer-eriksson made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1588
* @ryanli made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1611
* @sonicrules1234 made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1614
* @emmanuel-ferdman made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1635
* @semihpolat made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1572
* @anchpop made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1675
* @parasyte made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1662
* @htiennv made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1691
* @llamaha made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1712
* @lmzuccarelli made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1742
* @Cooksey99 made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1752
* @vigsterkr made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1793
* @krampenschiesser made their first contribution in https://github.com/EricLBuehler/mistral.rs/pull/1818

**Full Changelog**: https://github.com/EricLBuehler/mistral.rs/compare/v0.6.0...v0.7.0