Releases: mudler/LocalAI
Releases · mudler/LocalAI
v2.24.2
v2.24.1
v2.24.0
LocalAI release v2.24.0!
🚀 Highlights
- Backend deprecation: We’ve removed
rwkv.cppandbert.cpp, replacing them with enhanced functionalities inllama.cppfor simpler installation and better performance. - New Backends Added: Introducing
bark.cppfor text-to-audio andstablediffusion.cppfor image generation, both powered by the ggml framework. - Voice Activity Detection (VAD): Added support for
silero-vadto detect speech in audio streams. - WebUI Improvements: Now supports API key authentication for enhanced security.
- Real-Time Token Usage: Monitor token consumption during streamed outputs.
- Expanded P2P Settings: Greater flexibility with new configuration options like
listen_maddrs,dht_announce_maddrs, andbootstrap_peers.
📤 Backends Deprecation
As part of our cleanup efforts, the rwkv.cpp and bert.cpp backends have been deprecated. Their functionalities are now integrated into llama.cpp, offering a more streamlined and efficient experience.
🆕 New Backends Introduced
-
bark.cppBackend: Transform text into realistic audio using Bark, a transformer-based text-to-audio model. Install it easily with:local-ai models install bark-cpp-small
Or start it directly:
local-ai run bark-cpp-small
-
stablediffusion.cppBackend: Create high-quality images from textual descriptions using the Stable Diffusion backend, now leveraging the ggml framework. -
Voice Activity Detection with
silero-vad: Introducing support for accurate speech segment detection in audio streams. Install via:local-ai models install silero-vad
Or configure it through the WebUI.
🔒 WebUI Access with API Keys
The WebUI now supports API key authentication. If one or more API Keys are configured, the WebUI will automatically display a page to authenticate with.
🏆 Enhancements and Features
- Real-Time Token Usage: Monitor token consumption dynamically during streamed outputs. This feature helps optimize performance and manage costs effectively.
- P2P Configuration: New settings for advanced peer-to-peer mode:
listen_maddrs: Define specific multiaddresses for your node.dht_announce_maddrs: Specify addresses to announce to the DHT network.bootstrap_peers: Set custom bootstrap peers for initial connectivity.
These options offer more control, especially in constrained networks or custom P2P environments.
🖼️ New Models in the Gallery
We've significantly expanded our model gallery with a variety of new models to cater to diverse AI applications. Among these:
- Calme-3 Qwen2.5 Series: Enhanced language models offering improved understanding and generation capabilities.
- Mistral-Nemo-Prism-12b: A powerful model designed for complex language tasks.
- Llama 3.1 and 3.2 Series: Upgraded versions of the Llama models with better performance and accuracy.
- Qwen2.5-Coder Series: Specialized models optimized for code generation and programming language understanding.
- Rombos-Coder Series: Advanced coder models for sophisticated code-related tasks.
- Silero-VAD: High-quality voice activity detection model for audio processing applications.
- Bark-Cpp-Small: Lightweight audio generation model suitable for quick and efficient audio synthesis.
Explore these models and more in our updated model gallery to find the perfect fit for your project needs.
🐞 Bug Fixes and Improvements
- Performance Enhancements: Resolved issues with AVX flags and optimized binaries for accelerated performance, especially on macOS systems.
- Dependency Updates: Upgraded various dependencies to ensure compatibility, security, and performance improvements across the board.
- Parsing Corrections: Fixed parsing issues related to maddr and ExtraLLamaCPPArgs in P2P configurations.
📚 Documentation and Examples
- Updated Guides: Refreshed documentation with new configuration examples, making it easier to get started and integrate the latest features.
📥 How to Upgrade
To upgrade to LocalAI v2.24.0:
- Download the Latest Release: Get the binaries from our GitHub Releases page.
- Update Docker Image: Pull the latest Docker image using:
docker pull localai/localai:latest
See also the Documentation at: https://localai.io/basics/container/#standard-container-images
Happy hacking!
What's Changed
Breaking Changes 🛠
- feat(models): use rwkv from llama.cpp by @mudler in #4264
- feat(backends): Drop bert.cpp by @mudler in #4272
Bug fixes 🐛
- fix(hipblas): disable avx flags when accellerated bins are used by @mudler in #4167
- chore(deps): bump sycl intel image by @mudler in #4201
- fix(go.mod): add urfave/cli v2 by @mudler in #4206
- chore(go.mod): add valyala/fasttemplate by @mudler in #4207
- fix(p2p): parse maddr correctly by @mudler in #4219
- fix(p2p): parse correctly ExtraLLamaCPPArgs by @mudler in #4220
- fix(llama.cpp): embed metal file into result binary for darwin by @mudler in #4279
Exciting New Features 🎉
- feat: add WebUI API token authorization by @mintyleaf in #4197
- feat(p2p): add support for configuration of edgevpn listen_maddrs, dht_announce_maddrs and bootstrap_peers by @mintyleaf in #4200
- feat(silero): add Silero-vad backend by @mudler in #4204
- feat: include tokens usage for streamed output by @mintyleaf in #4282
- feat(bark-cpp): add new bark.cpp backend by @mudler in #4287
- feat(backend): add stablediffusion-ggml by @mudler in #4289
🧠 Models
- models(gallery): add calme-3 qwen2.5 series by @mudler in #4107
- models(gallery): add calme-3 qwenloi series by @mudler in #4108
- models(gallery): add calme-3 llamaloi series by @mudler in #4109
- models(gallery): add mn-tiramisu-12b by @mudler in #4110
- models(gallery): add qwen2.5-coder-14b by @mudler in #4125
- models(gallery): add qwen2.5-coder-3b-instruct by @mudler in #4126
- models(gallery): add qwen2.5-coder-32b-instruct by @mudler in #4127
- models(gallery): add qwen2.5-coder-14b-instruct by @mudler in #4128
- models(gallery): add qwen2.5-coder-1.5b-instruct by @mudler in #4129
- models(gallery): add qwen2.5-coder-7b-instruct by @mudler in #4130
- models(gallery): add qwen2.5-coder-7b-3x-instruct-ties-v1.2-i1 by @mudler in #4131
- models(gallery): add qwen2.5-coder-7b-instruct-abliterated-i1 by @mudler in #4132
- models(gallery): add rombos-coder-v2.5-qwen-7b by @mudler in #4133
- models(gallery): add rombos-coder-v2.5-qwen-32b by @mudler in #4134
- models(gallery): add rombos-coder-v2.5-qwen-14b by @mudler in #4135
- models(gallery): add eva-qwen2.5-72b-v0.1-i1 by @mudler in #4136
- models(gallery): add mistral-nemo-prism-12b by @mudler in #4141
- models(gallery): add tess-3-llama-3.1-70b by @mudler in #4143
- models(gallery): add celestial-harmony-14b-v1.0-experimental-1016-i1 by @mudler in #4145
- models(gallery): add llama3.1-8b-enigma by @mudler in #4146
- chore(model): add llama3.1-8b-cobalt to the gallery by @mudler in #4147
- chore(model): add qwen2.5-32b-arliai-rpmax-v1.3 to the gallery by @mudler in #4148
- chore(model): add llama3.2-3b-enigma to the gallery by @mudler in #4149
- chore(model): add llama-3.1-8b-arliai-rpmax-v1.3 to the gallery by @mudler in #4150
- chore(model): add magnum-12b-v2.5-kto-i1 to the gallery by @mudler in #4151
- chore(model): add l3.1-8b-slush-i1 to the gallery by @mudler in #4152
- models(gallery): add q2.5-ms-mistoria-72b-i1 by @mudler in #4158
- chore(model): add l3.1-ms-astoria-70b-v2 to the gallery by @mudler in #4159
- chore(model): add magnum-v2-4b-i1 to the gallery by @mudler in #4160
- chore(model): add athene-v2-agent to the gallery by @mudler in #4161
- chore(model): add athene-v2-chat to the gallery by @mudler in #4162
- chore(model-gallery): ⬆️ update checksum by @localai-bot in #4165
- chore(model): add qwen2.5-7b-nerd-uncensor...
Contributors
meonkeys, mudler, and 7 other contributors
Assets 11
v2.23.0
What's Changed
Breaking Changes 🛠
Bug fixes 🐛
- fix(parler-tts): use latest audiotools by @mudler in #3954
- fix(parler-tts): pin grpcio-tools by @mudler in #3960
- fix(gallery): overrides for parler-tts in the gallery by @mudler in #3962
- fix(grpc): pass by modelpath by @mudler in #4023
- fix(tts): correctly pass backend config when generating model options by @mudler in #4091
Exciting New Features 🎉
- feat(vllm): expose 'load_format' by @mudler in #3943
- feat(tts): Implement naive response_format for tts endpoint by @n-Arno in #4035
- feat(diffusers): allow multiple lora adapters by @mudler in #4081
- feat(ui): move model detailed info to a modal by @mudler in #4086
- feat: allow to disable '/metrics' endpoints for local stats by @mudler in #3945
- feat: add flux single file support by @sozercan in #3959
🧠 Models
- fix(phi3-vision): add multimodal template by @mudler in #3944
- models(gallery): add l3.1-moe-2x8b-v0.2 by @mudler in #3969
- models(gallery): add llama3.1-darkstorm-aspire-8b by @mudler in #3970
- models(gallery): add llama-3.2-sun-2.5b-chat by @mudler in #3971
- models(gallery): add darkest-muse-v1 by @mudler in #3973
- models(gallery): add llama-3.2-3b-instruct-uncensored by @mudler in #3974
- models(gallery): add thebeagle-v2beta-32b-mgs by @mudler in #3975
- models(gallery): add l3.1-70blivion-v0.1-rc1-70b-i1 by @mudler in #3977
- models(gallery): add llama-3.1-hawkish-8b by @mudler in #3978
- models(gallery): add quill-v1 by @mudler in #3980
- models(gallery): add delirium-v1 by @mudler in #3981
- models(gallery): add magnum-v4-9b by @mudler in #3983
- models(gallery): add llama-3-whiterabbitneo-8b-v2.0 by @mudler in #3984
- models(gallery): add l3-nymeria-maid-8b by @mudler in #3985
- models(gallery): add meraj-mini by @mudler in #3987
- models(gallery): add granite-3.0-1b-a400m-instruct by @mudler in #3994
- models(gallery): add moe-girl-800ma-3bt by @mudler in #3995
- models(gallery): add spiral-da-hyah-qwen2.5-72b-i1 by @mudler in #4022
- models(gallery): add llama3.1-bestmix-chem-einstein-8b by @mudler in #4028
- models(gallery): add starcannon-unleashed-12b-v1.0 by @mudler in #4032
- models(gallery): add smollm2-1.7b-instruct by @mudler in #4033
- models(gallery): add control-8b-v1.1 by @mudler in #4039
- models(gallery): add whiterabbitneo-2.5-qwen-2.5-coder-7b by @mudler in #4042
- models(gallery): add llama-3.1-whiterabbitneo-2-8b by @mudler in #4043
- models(gallery): add g2-9b-aletheia-v1 by @mudler in #4056
- models(gallery): add cybertron-v4-qw7b-mgs by @mudler in #4063
- models(gallery): add g2-9b-sugarquill-v0 by @mudler in #4073
- chore(model-gallery): ⬆️ update checksum by @localai-bot in #4080
- models(gallery): add q25-1.5b-veolu by @mudler in #4088
- models(gallery): add valor-7b-v0.1 by @mudler in #4089
- models(gallery): add tess-r1-limerick-llama-3.1-70b by @mudler in #4095
- models(gallery): add llenn-v0.75-qwen2.5-72b-i1 by @mudler in #4098
- models(gallery): add eva-qwen2.5-14b-v0.2 by @mudler in #4099
- models(gallery): add opencoder-8b instruct and base by @mudler in #4101
- models(gallery): add opencoder-1.5b instruct and base by @mudler in #4102
- models(gallery): add tissint-14b-128k-rp by @mudler in #4103
- models(gallery): add tq2.5-14b-sugarquill-v1 by @mudler in #4104
📖 Documentation and examples
- docs: add Homebrew as an option to install on MacOS by @mauromorales in #3946
- chore: drop examples folder now that LocalAI-examples has been created by @dave-gray101 in #4017
👒 Dependencies
- chore(deps): Bump llama-index from 0.11.17 to 0.11.19 in /examples/chainlit by @dependabot in #3893
- chore(deps): Bump weaviate-client from 4.8.1 to 4.9.0 in /examples/chainlit by @dependabot in #3894
- chore(deps): Bump langchain from 0.3.3 to 0.3.4 in /examples/functions by @dependabot in #3900
- chore(deps): Bump langchain-community from 0.3.2 to 0.3.3 in /examples/langchain/langchainpy-localai-example by @dependabot in #3923
- chore(deps): Bump docs/themes/hugo-theme-relearn from
007cc20to06e70daby @dependabot in #3932 - chore(deps): Bump sqlalchemy from 2.0.35 to 2.0.36 in /examples/langchain/langchainpy-localai-example by @dependabot in #3920
- chore(deps): Bump yarl from 1.15.2 to 1.15.5 in /examples/langchain/langchainpy-localai-example by @dependabot in #3921
- chore(deps): Bump openai from 1.51.2 to 1.52.0 in /examples/langchain-chroma by @dependabot in #3908
- chore(deps): Bump llama-index from 0.11.17 to 0.11.19 in /examples/langchain-chroma by @dependabot in #3907
- chore(deps): Bump marshmallow from 3.22.0 to 3.23.0 in /examples/langchain/langchainpy-localai-example by @dependabot in #3917
- chore(deps): Bump openai from 1.51.2 to 1.52.0 in /examples/functions by @dependabot in #3901
- chore(deps): Bump yarl from 1.15.5 to 1.16.0 in /examples/langchain/langchainpy-localai-example by @dependabot in #3938
- chore(deps): Bump openai from 1.51.2 to 1.52.2 in /examples/langchain/langchainpy-localai-example by @dependabot in #3993
- chore(deps): Bump torchvision from 0.18.1+rocm6.0 to 0.20.0+cu118 in /backend/python/diffusers by @dependabot in #3997
- chore(deps): Bump docs/themes/hugo-theme-relearn from
06e70dato28fce6bby @dependabot in #3986 - chore(deps): Bump llama-index from 0.11.19 to 0.11.20 in /examples/langchain-chroma by @dependabot in #3988
- chore(deps): Bump openai from 1.52.0 to 1.52.2 in /examples/langchain-chroma by @dependabot in #3989
- chore(deps): Bump llama-index from 0.11.19 to 0.11.20 in /examples/chainlit by @dependabot in #3990
- chore(deps): Bump tqdm from 4.66.5 to 4.66.6 in /examples/langchain/langchainpy-localai-example by @dependabot in #3991
- chore(deps): Bump frozenlist from 1.4.1 to 1.5.0 in /examples/langchain/langchainpy-localai-example by @dependabot in #3992
- chore(deps): Bump openai from 1.52.0 to 1.52.2 in /examples/functions by @dependabot in #4000
- chore(deps): bump grpcio to 1.67.1 by @mudler in #4009
- chore(deps): bump llama-cpp to 8f275a7c4593aa34147595a90282cf950a853690 by @mudler in #4016
Other Changes
- docs: ⬆️ update docs version mudler/LocalAI by @localai-bot in #3936
- chore: ⬆️ Update ggerganov/llama.cpp to
e01c67affe450638162a1a457e2e57859ef6ebf0by @localai-bot in #3937 - chore: update go-piper to latest by @dave-gray101 in #3939
- chore: ⬆️ Update ggerganov/llama.cpp to
c8c07d658a6cefc5a50cfdf6be7d726503612303by @localai-bot in #3940 - chore: ⬆️ Update ggerganov/whisper.cpp to
0fbaac9c891055796456df7b9122a70c220f9ca1by @localai-bot in #3950 - chore: ⬆️ Update ggerganov/llama.cpp to
0a1c750c80147687df267114c81956757cc14382by @localai-bot in #3948 - chore...
Contributors
mauromorales, sozercan, and 5 other contributors
Assets 11
v2.22.1
What's Changed
Bug fixes 🐛
- fix(vllm): images and videos are base64 by default by @mudler in #3867
- fix(dependencies): pin pytorch version by @mudler in #3872
- fix(dependencies): move deps that brings pytorch by @mudler in #3873
- fix(vllm): do not set videos if we don't have any by @mudler in #3885
Exciting New Features 🎉
- feat(templates): extract text from multimodal requests by @mudler in #3866
- feat(templates): add sprig to multimodal templates by @mudler in #3868
🧠 Models
- models(gallery): add llama-3_8b_unaligned_beta by @mudler in #3818
- models(gallery): add llama3.1-flammades-70b by @mudler in #3819
- models(gallery): add llama3.1-gutenberg-doppel-70b by @mudler in #3820
- models(gallery): add llama-3.1-8b-arliai-formax-v1.0-iq-arm-imatrix by @mudler in #3821
- models(gallery): add supernova-medius by @mudler in #3822
- models(gallery): add hermes-3-llama-3.1-70b-lorablated by @mudler in #3823
- models(gallery): add hermes-3-llama-3.1-8b-lorablated by @mudler in #3824
- models(gallery): add eva-qwen2.5-14b-v0.1-i1 by @mudler in #3825
- models(gallery): add cursorcore-qw2.5-7b-i1 by @mudler in #3826
- models(gallery): add cursorcore-qw2.5-1.5b-lc-i1 by @mudler in #3827
- models(gallery): add cursorcore-ds-6.7b-i1 by @mudler in #3828
- models(gallery): add cursorcore-yi-9b by @mudler in #3829
- models(gallery): add edgerunner-command-nested-i1 by @mudler in #3830
- models(gallery): add llama-3.2-chibi-3b by @mudler in #3843
- models(gallery): add llama-3.2-3b-reasoning-time by @mudler in #3844
- models(gallery): add ml-ms-etheris-123b by @mudler in #3845
- models(gallery): add doctoraifinetune-3.1-8b-i1 by @mudler in #3846
- models(gallery): add astral-fusion-neural-happy-l3.1-8b by @mudler in #3848
- models(gallery): add tsunami-0.5x-7b-instruct-i1 by @mudler in #3849
- models(gallery): add mahou-1.5-llama3.1-70b-i1 by @mudler in #3850
- models(gallery): add llama-3.1-nemotron-70b-instruct-hf by @mudler in #3854
- models(gallery): add qevacot-7b-v2 by @mudler in #3855
- models(gallery): add l3.1-etherealrainbow-v1.0-rc1-8b by @mudler in #3856
- models(gallery): add phi-3.5-mini-titanfusion-0.2 by @mudler in #3857
- models(gallery): add mn-lulanum-12b-fix-i1 by @mudler in #3859
- models(gallery): add apollo2-9b by @mudler in #3860
- models(gallery): add theia-llama-3.1-8b-v1 by @mudler in #3861
- models(gallery): add tor-8b by @mudler in #3862
- models(gallery): add darkens-8b by @mudler in #3863
- models(gallery): add baldur-8b by @mudler in #3864
- models(gallery): add meissa-qwen2.5-7b-instruct by @mudler in #3865
- models(gallery): add phi-3 vision by @mudler in #3890
👒 Dependencies
- chore(deps): Bump docs/themes/hugo-theme-relearn from
d5a0ee0toe1a1f01by @dependabot in #3798 - chore(deps): Bump mxschmitt/action-tmate from 3.18 to 3.19 by @dependabot in #3799
- chore(deps): Bump sentence-transformers from 3.1.1 to 3.2.0 in /backend/python/sentencetransformers by @dependabot in #3801
- chore(deps): Bump langchain from 0.3.2 to 0.3.3 in /examples/langchain/langchainpy-localai-example by @dependabot in #3803
- chore(deps): Bump llama-index from 0.11.16 to 0.11.17 in /examples/langchain-chroma by @dependabot in #3804
- chore(deps): Bump python from 3.12-bullseye to 3.13-bullseye in /examples/langchain by @dependabot in #3805
- chore(deps): Bump openai from 1.51.1 to 1.51.2 in /examples/functions by @dependabot in #3806
- chore(deps): Bump llama-index from 0.11.16 to 0.11.17 in /examples/chainlit by @dependabot in #3807
- chore(deps): Bump langchain from 0.3.1 to 0.3.3 in /examples/langchain-chroma by @dependabot in #3809
- chore(deps): Bump openai from 1.51.1 to 1.51.2 in /examples/langchain/langchainpy-localai-example by @dependabot in #3808
- chore(deps): Bump yarl from 1.13.1 to 1.15.1 in /examples/langchain/langchainpy-localai-example by @dependabot in #3816
- chore(deps): Bump chromadb from 0.5.11 to 0.5.13 in /examples/langchain-chroma by @dependabot in #3811
- chore(deps): Bump langchain from 0.3.2 to 0.3.3 in /examples/functions by @dependabot in #3802
- chore(deps): Bump debugpy from 1.8.6 to 1.8.7 in /examples/langchain/langchainpy-localai-example by @dependabot in #3814
- chore(deps): Bump aiohttp from 3.10.9 to 3.10.10 in /examples/langchain/langchainpy-localai-example by @dependabot in #3812
- chore(deps): Bump openai from 1.51.1 to 1.51.2 in /examples/langchain-chroma by @dependabot in #3810
- chore(deps): Bump charset-normalizer from 3.3.2 to 3.4.0 in /examples/langchain/langchainpy-localai-example by @dependabot in #3834
- chore(deps): Bump langchain-community from 0.3.1 to 0.3.2 in /examples/langchain/langchainpy-localai-example by @dependabot in #3831
- chore(deps): Bump yarl from 1.15.1 to 1.15.2 in /examples/langchain/langchainpy-localai-example by @dependabot in #3832
- chore(deps): Bump numpy from 2.1.1 to 2.1.2 in /examples/langchain/langchainpy-localai-example by @dependabot in #3833
- chore(deps): Bump docs/themes/hugo-theme-relearn from
e1a1f01to007cc20by @dependabot in #3835 - chore(deps): Bump gradio from 3.48.0 to 5.0.0 in /backend/python/openvoice in the pip group by @dependabot in #3880
- chore(deps): bump llama-cpp to cda0e4b648dde8fac162b3430b14a99597d3d74f by @mudler in #3884
Other Changes
- docs: ⬆️ update docs version mudler/LocalAI by @localai-bot in #3796
- chore: dependabot ignore generated grpc go package by @dave-gray101 in #3795
- chore: ⬆️ Update ggerganov/llama.cpp to
edc265661cd707327297b6ec4d83423c43cb50a5by @localai-bot in #3797 - chore: ⬆️ Update ggerganov/llama.cpp to
d4c19c0f5cdb1e512573e8c86c79e8d0238c73c4by @localai-bot in #3817 - chore: ⬆️ Update ggerganov/llama.cpp to
a89f75e1b7b90cb2d4d4c52ca53ef9e9b466aa45by @localai-bot in #3837 - chore: ⬆️ Update ggerganov/whisper.cpp to
06a1da9daff94c1bf1b1d38950628264fe443f76by @localai-bot in #3836 - Update integrations.md with LLPhant by @f-lombardo in #3838
- fix(llama.cpp): consider also native builds by @mudler in #3839
- chore: ⬆️ Update ggerganov/whisper.cpp to
b6049060dd2341b7816d2bce7dc7451c1665828eby @localai-bot in #3842 - chore: ⬆️ Update ggerganov/llama.cpp to
755a9b2bf00fbae988e03a47e852b66eaddd113aby @localai-bot in #3841 - chore(deps): bump grpcio to 1.67.0 by @mudler in #3851
- chore: ⬆️ Update ggerganov/llama.cpp to
9e041024481f6b249ab8918e18b9477f873b5a5eby @localai-bot in #3853 - chore: ⬆️ Update ggerganov/whisper.cpp to
d3f7137cc9befa6d74dc4085de2b664b97b7c8bbby @localai-bot in #3852 - fix(mamba): pin torch version by @mudler in #3871
- chore: ⬆️ Update ggerganov/llama.cpp to
99bd4ac28c32cd17c0e337ff5601393b033dc5fcby @localai-bot in #3869 - chore: ⬆️ Update ggerganov/whisper.cpp to
a5abfe6a90495f7bf19fe70d016ecc255e97359cby @localai-bot in #3870 - chore(deps): pin packaging by @mudler ...
Contributors
f-lombardo, mudler, and 3 other contributors
Assets 11
v2.22.0
LocalAI v2.22.0 is out 🥳
💡 Highlights
- Image-to-Text and Video-to-Text Support: The VLLM backend now supports both image-to-text and video-to-text processing.
- Enhanced Multimodal Support: Template placeholders are now available, offering more flexibility in multimodal applications
- Model Management Made Easy: List all your loaded models directly via the /system endpoint for seamless management.
- Various bugfixes and improvements: Fixed issues with dangling processes to ensure proper resource management and resolved channel closure issues in the base GRPC server.
🖼️ Multimodal vLLM
To use multimodal models with vLLM simply specify the model in the YAML file. Models however can differ if support multiple images or single images, along how they process internally placeholders for images.
Some models/libraries have different way to express images, videos or audio placeholders. For example, llama.cpp backend expects images within an [img-ID] tag, but other backends/models (e.g. vLLM) use a different notation ( <|image_|>).
For example, to override defaults, now it is possible to set in the model configuration the following:
template:
video: "<|video_{{.ID}}|> {{.Text}}"
image: "<|image_{{.ID}}|> {{.Text}}"
audio: "<|audio_{{.ID}}|> {{.Text}}"📹 Video and Audio understanding
Some libraries might support both Video and Audio. Currently only vLLM supports Video understanding, and can be used in the API by "extending" the OpenAI API with audio and video type along images:
curl http://localhost:8080/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-4o",
"messages": [
{
"role": "user",
"content": [
{
"type": "text",
"text": "What'\''s in this video?"
},
{
"type": "video_url",
"video_url": {
"url": "https://video-image-url"
}
}
]
}
],
"max_tokens": 300
}'
🧑🏭 Work in progress
- Realtime API is work in progress , tracked in #3714. thumbs up if you want to see it supported in LocalAI!
What's Changed
Bug fixes 🐛
- chore: simplify model loading by @mudler in #3715
- fix(initializer): correctly reap dangling processes by @mudler in #3717
- fix(base-grpc): close channel in base grpc server by @mudler in #3734
- fix(vllm): bump cmake - vllm requires it by @mudler in #3744
- fix(llama-cpp): consistently select fallback by @mudler in #3789
- fix(welcome): do not list model twice if we have a config by @mudler in #3790
- fix: listmodelservice / welcome endpoint use LOOSE_ONLY by @dave-gray101 in #3791
Exciting New Features 🎉
- feat(api): list loaded models in
/systemby @mudler in #3661 - feat: Add Get Token Metrics to GRPC server by @siddimore in #3687
- refactor: ListModels Filtering Upgrade by @dave-gray101 in #2773
- feat: track internally started models by ID by @mudler in #3693
- feat: tokenization endpoint by @shraddhazpy in #3710
- feat(multimodal): allow to template placeholders by @mudler in #3728
- feat(vllm): add support for image-to-text and video-to-text by @mudler in #3729
- feat(shutdown): allow force shutdown of backends by @mudler in #3733
- feat(transformers): Use downloaded model for Transformers backend if it already exists. by @joshbtn in #3777
- fix: roll out bluemonday Sanitize more widely by @dave-gray101 in #3794
🧠 Models
- models(gallery): add llama-3.2 3B and 1B by @mudler in #3671
- chore(model-gallery): ⬆️ update checksum by @localai-bot in #3675
- models(gallery): add magnusintellectus-12b-v1-i1 by @mudler in #3678
- models(gallery): add bigqwen2.5-52b-instruct by @mudler in #3679
- feat(api): add correlationID to Track Chat requests by @siddimore in #3668
- models(gallery): add replete-llm-v2.5-qwen-14b by @mudler in #3688
- models(gallery): add replete-llm-v2.5-qwen-7b by @mudler in #3689
- models(gallery): add calme-2.2-qwen2.5-72b-i1 by @mudler in #3691
- models(gallery): add salamandra-7b-instruct by @mudler in #3726
- models(gallery): add mn-backyardai-party-12b-v1-iq-arm-imatrix by @mudler in #3740
- models(gallery): add t.e-8.1-iq-imatrix-request by @mudler in #3741
- models(gallery): add violet_twilight-v0.2-iq-imatrix by @mudler in #3742
- models(gallery): add gemma-2-9b-it-abliterated by @mudler in #3743
- models(gallery): add moe-girl-1ba-7bt-i1 by @mudler in #3766
- models(gallery): add archfunctions models by @mudler in #3767
- models(gallery): add versatillama-llama-3.2-3b-instruct-abliterated by @mudler in #3771
- models(gallery): add llama3.2-3b-enigma by @mudler in #3772
- models(gallery): add llama3.2-3b-esper2 by @mudler in #3773
- models(gallery): add llama-3.1-swallow-70b-v0.1-i1 by @mudler in #3774
- models(gallery): add rombos-llm-v2.5.1-qwen-3b by @mudler in #3778
- models(gallery): add qwen2.5-7b-ins-v3 by @mudler in #3779
- models(gallery): add dans-personalityengine-v1.0.0-8b by @mudler in #3780
- models(gallery): add llama-3.2-3b-agent007 by @mudler in #3781
- models(gallery): add nihappy-l3.1-8b-v0.09 by @mudler in #3782
- models(gallery): add llama-3.2-3b-agent007-coder by @mudler in #3783
- models(gallery): add fireball-meta-llama-3.2-8b-instruct-agent-003-128k-code-dpo by @mudler in #3784
- models(gallery): add gemma-2-ataraxy-v3i-9b by @mudler in #3785
📖 Documentation and examples
👒 Dependencies
- chore: ⬆️ Update ggerganov/llama.cpp to
ea9c32be71b91b42ecc538bd902e93cbb5fb36cbby @localai-bot in #3667 - chore: ⬆️ Update ggerganov/whisper.cpp to
69339af2d104802f3f201fd419163defba52890eby @localai-bot in #3666 - chore: ⬆️ Update ggerganov/llama.cpp to
95bc82fbc0df6d48cf66c857a4dda3d044f45ca2by @localai-bot in #3674 - chore: ⬆️ Update ggerganov/llama.cpp to
b5de3b74a595cbfefab7eeb5a567425c6a9690cfby @localai-bot in #3681 - chore: ⬆️ Update ggerganov/whisper.cpp to
8feb375fbdf0277ad36958c218c6bf48fa0ba75aby @localai-bot in #3680 - chore: ⬆️ Update ggerganov/llama.cpp to
c919d5db39c8a7fcb64737f008e4b105ee0acd20by @localai-bot in #3686 - chore(deps): bump grpcio to 1.66.2 by @mudler in #3690
- chore(deps): Bump openai from 1.47.1 to 1.50.2 in /examples/langchain-chroma by @dependabot in #3697
- chore(deps): Bump chromadb from 0.5.7 to 0.5.11 in /examples/langchain-chroma by @dependabot in #3696
- chore(deps): Bump langchain from 0.3.0 to 0.3.1 in /examples/langchain-chroma by @dependabot in #3694
- chore: ⬆️ Update ggerganov/llama.cpp to
6f1d9d71f4c568778a7637ff6582e6f6ba5fb9d3by @localai-bot in #3708 - chore(deps): Bump securego/gosec from 2.21.0 to 2.21.4 by @dependabot in #3698
- chore(deps): Bump openai from 1.47.1 to 1.50.2 in /examples/functions by @dependabot in #3699
- chore(deps): Bump langchain from 0.3.0 to 0.3.1 in /examples/langchain/langchainpy-localai-example by @dependabot in #3704
- chore(deps): Bump greenlet from 3.1.0 to 3.1.1 in /examples/langchain/langchainpy-localai-example by @dependabot in #3703
- chore(deps): Bump langchain from 0.3.0 to 0.3.1 in /examples/functions by @dependabot in #3700
- chore(deps): Bump langchain-community from 0.2.16 to 0.3.1 in /examples/langchain/langchainpy-localai-example by ...
Contributors
joshbtn, jjasghar, and 6 other contributors
Assets 11
v2.21.1
What's Changed
Bug fixes 🐛
👒 Dependencies
- chore(deps): Bump sentence-transformers from 3.1.0 to 3.1.1 in /backend/python/sentencetransformers by @dependabot in #3651
- chore(deps): Bump pydantic from 2.8.2 to 2.9.2 in /examples/langchain/langchainpy-localai-example by @dependabot in #3648
- chore(deps): Bump openai from 1.45.1 to 1.47.1 in /examples/functions by @dependabot in #3645
- chore: ⬆️ Update ggerganov/llama.cpp to
70392f1f81470607ba3afef04aa56c9f65587664by @localai-bot in #3659 - chore(deps): Bump llama-index from 0.11.7 to 0.11.12 in /examples/langchain-chroma by @dependabot in #3639
- chore(deps): Bump openai from 1.45.1 to 1.47.1 in /examples/langchain-chroma by @dependabot in #3641
- chore(deps): Bump llama-index from 0.11.9 to 0.11.12 in /examples/chainlit by @dependabot in #3642
- chore: ⬆️ Update ggerganov/whisper.cpp to
0d2e2aed80109e8696791083bde3b58e190b7812by @localai-bot in #3658 - chore(deps): Bump chromadb from 0.5.5 to 0.5.7 in /examples/langchain-chroma by @dependabot in #3640
Other Changes
- docs: ⬆️ update docs version mudler/LocalAI by @localai-bot in #3657
Full Changelog: v2.21.0...v2.21.1
Contributors
mudler, dependabot, and localai-bot
Assets 11
v2.21.0
💡 Highlights!
LocalAI v2.21 release is out!
- Deprecation of the
exllamabackend - AIO images now have
gpt-4oinstead ofgpt-4-vision-previewfor Vision API - vLLM backend now supports embeddings
- New endpoint to list system information (
/system) trust_remote_codeis now respected bysentencetransformers- Auto warm-up and load models on start
coquibackend switched to the community-maintained fork
What's Changed
Breaking Changes 🛠
- chore(exllama): drop exllama backend by @mudler in #3536
- chore(aio): rename gpt-4-vision-preview to gpt-4o by @mudler in #3597
Exciting New Features 🎉
- feat: elevenlabs
sound-generationapi by @dave-gray101 in #3355 - feat(vllm): add support for embeddings by @mudler in #3440
- feat: add endpoint to list system informations by @mudler in #3449
- feat: extract output with regexes from LLMs by @mudler in #3491
- feat: allow setting trust_remote_code for sentencetransformers backend by @Nyralei in #3552
- feat(api): allow to pass videos to backends by @mudler in #3601
- feat(api): allow to pass audios to backends by @mudler in #3603
- feat: auto load into memory on startup by @sozercan in #3627
- feat(coqui): switch to maintained community fork by @mudler in #3625
Bug fixes 🐛
- fix(p2p): correctly allow to pass extra args to llama.cpp by @mudler in #3368
- fix(model-loading): keep track of open GRPC Clients by @mudler in #3377
- fix(tts): check error before inspecting result by @mudler in #3415
- fix(shutdown): do not shutdown immediately busy backends by @mudler in #3543
- fix(parler-tts): fix install with sycl by @mudler in #3624
- fix(ci): fixup checksum scanning pipeline by @mudler in #3631
- fix(hipblas): do not push all variants to hipblas builds by @mudler in #3630
🧠 Models
- chore(model-gallery): add more quants for popular models by @mudler in #3365
- models(gallery): add phi-3.5 by @mudler in #3376
- models(gallery): add calme-2.1-phi3.5-4b-i1 by @mudler in #3383
- models(gallery): add magnum-v3-34b by @mudler in #3384
- models(gallery): add phi-3.5-vision by @mudler in #3421
- Revert "models(gallery): add phi-3.5-vision" by @mudler in #3422
- chore(model-gallery): ⬆️ update checksum by @localai-bot in #3425
- feat: Added Piper voice it-paola-medium by @fakezeta in #3434
- chore(model-gallery): ⬆️ update checksum by @localai-bot in #3442
- models(gallery): add hubble-4b-v1 by @mudler in #3444
- chore(model-gallery): ⬆️ update checksum by @localai-bot in #3446
- models(gallery): add yi-coder (and variants) by @mudler in #3482
- chore(model-gallery): ⬆️ update checksum by @localai-bot in #3486
- models(gallery): add reflection-llama-3.1-70b by @mudler in #3487
- models(gallery): add athena-codegemma-2-2b-it by @mudler in #3490
- models(gallery): add azure_dusk-v0.2-iq-imatrix by @mudler in #3538
- models(gallery): add mn-12b-lyra-v4-iq-imatrix by @mudler in #3539
- models(gallery): add datagemma models by @mudler in #3540
- models(gallery): add l3.1-8b-niitama-v1.1-iq-imatrix by @mudler in #3550
- models(gallery): add llama-3.1-8b-stheno-v3.4-iq-imatrix by @mudler in #3551
- fix:
gallery/index.yamlcomment spacing by @dave-gray101 in #3585 - models(gallery): add qwen2.5-14b-instruct by @mudler in #3607
- models(gallery): add qwen2.5-math-7b-instruct by @mudler in #3609
- models(gallery): add qwen2.5-14b_uncencored by @mudler in #3610
- models(gallery): add qwen2.5-coder-7b-instruct by @mudler in #3611
- models(gallery): add qwen2.5-math-72b-instruct by @mudler in #3612
- models(gallery): add qwen2.5-0.5b-instruct, qwen2.5-1.5b-instruct by @mudler in #3613
- models(gallery): add qwen2.5 32B, 72B, 32B Instruct by @mudler in #3614
- models(gallery): add llama-3.1-supernova-lite-reflection-v1.0-i1 by @mudler in #3615
- models(gallery): add llama-3.1-supernova-lite by @mudler in #3616
- models(gallery): add llama3.1-8b-shiningvaliant2 by @mudler in #3617
- models(gallery): add buddy2 by @mudler in #3618
- models(gallery): add llama-3.1-8b-arliai-rpmax-v1.1 by @mudler in #3619
- Fix NeuralDaredevil URL by @nyx4ris in #3621
- models(gallery): add nightygurps-14b-v1.1 by @mudler in #3633
- models(gallery): add gemma-2-9b-arliai-rpmax-v1.1 by @mudler in #3634
- models(gallery): add gemma-2-2b-arliai-rpmax-v1.1 by @mudler in #3635
- models(gallery): add acolyte-22b-i1 by @mudler in #3636
📖 Documentation and examples
- docs: ⬆️ update docs version mudler/LocalAI by @localai-bot in #3366
- chore(docs): add Vulkan images links by @mudler in #3620
👒 Dependencies
- chore: ⬆️ Update ggerganov/llama.cpp to
3ba780e2a8f0ffe13f571b27f0bbf2ca5a199efcby @localai-bot in #3361 - chore(deps): Bump openai from 1.41.1 to 1.42.0 in /examples/functions by @dependabot in #3390
- chore(deps): Bump docs/themes/hugo-theme-relearn from
82a5e98to3a0ae52by @dependabot in #3391 - chore(deps): Bump idna from 3.7 to 3.8 in /examples/langchain/langchainpy-localai-example by @dependabot in #3399
- chore(deps): Bump llama-index from 0.10.65 to 0.11.1 in /examples/chainlit by @dependabot in #3404
- chore(deps): Bump llama-index from 0.10.67.post1 to 0.11.1 in /examples/langchain-chroma by @dependabot in #3406
- chore(deps): Bump marshmallow from 3.21.3 to 3.22.0 in /examples/langchain/langchainpy-localai-example by @dependabot in #3400
- chore(deps): Bump openai from 1.40.5 to 1.42.0 in /examples/langchain-chroma by @dependabot in #3405
- chore(deps): Bump openai from 1.41.1 to 1.42.0 in /examples/langchain/langchainpy-localai-example by @dependabot in #3401
- chore(deps): update edgevpn to v0.28 by @mudler in #3412
- chore(deps): Bump langchain from 0.2.14 to 0.2.15 in /examples/functions by @dependabot in #3453
- chore(deps): Bump certifi from 2024.7.4 to 2024.8.30 in /examples/langchain/langchainpy-localai-example by @dependabot in #3457
- chore(deps): Bump yarl from 1.9.4 to 1.9.7 in /examples/langchain/langchainpy-localai-example by @dependabot in #3459
- chore(deps): Bump langchain-community from 0.2.12 to 0.2.15 in /examples/langchain/langchainpy-localai-example by @dependabot in #3461
- chore(deps): Bump llama-index from 0.11.1 to 0.11.4 in /examples/chainlit by @dependabot in #3462
- chore(deps): Bump llama-index from 0.11.1 to 0.11.4 in /examples/langchain-chroma by @dependabot in #3467
- chore(deps): Bump docs/themes/hugo-theme-relearn from
3a0ae52to550a6eeby @dependabot in #3472 - chore(deps): Bump openai from 1.42.0 to 1.43.0 in /examples/functions by @dependabot in #3452
- chore(deps): Bump langchain from 0.2.14 to 0.2.15 in /examples/langchain/langchainpy-localai-example by @dependabot in #3460
- chore(deps): Bump openai from 1.42.0 to 1.43.0 in /examples/langchain-chroma by @dependabot in #3468
- chore(deps): Bump langchain from 0.2.14 to 0.2.15 in /examples/langchain-chroma by ...
Contributors
sozercan, mudler, and 7 other contributors
Assets 11
v2.20.1
It's that time again—I’m excited (and honestly, a bit proud) to announce the release of LocalAI v2.20! This one’s a biggie, with some of the most requested features and enhancements, all designed to make your self-hosted AI journey even smoother and more powerful.
TL;DR
- 🌍 Explorer & Community: Explore global community pools at explorer.localai.io
- 👀 Demo instance available: Test out LocalAI at demo.localai.io
- 🤗 Integration: Hugging Face Local apps now include LocalAI
- 🐛 Bug Fixes: Diffusers and hipblas issues resolved
- 🎨 New Feature: FLUX-1 image generation support
- 🏎️ Strict Mode: Stay compliant with OpenAI’s latest API changes
- 💪 Multiple P2P Clusters: Run multiple clusters within the same network
- 🧪 Deprecation Notice:
gpt4all.cppandpetalsbackends deprecated
🌍 Explorer and Global Community Pools
Now you can share your LocalAI instance with the global community or explore available instances by visiting explorer.localai.io. This decentralized network powers our demo instance, creating a truly collaborative AI experience.
How It Works
Using the Explorer, you can easily share or connect to clusters. For detailed instructions on creating new clusters or connecting to existing ones, check out our documentation.
👀 Demo Instance Now Available
Curious about what LocalAI can do? Dive right in with our live demo at demo.localai.io! Thanks to our generous sponsors, this instance is publicly available and configured via peer-to-peer (P2P) networks. If you'd like to connect, follow the instructions here.
🤗 Hugging Face Integration
I am excited to announce that LocalAI is now integrated within Hugging Face’s local apps! This means you can select LocalAI directly within Hugging Face to build and deploy models with the power and flexibility of our platform. Experience seamless integration with a single click!
This integration was made possible through this PR.
🎨 FLUX-1 Image Generation Support
FLUX-1 lands in LocalAI! With this update, LocalAI can now generate stunning images using FLUX-1, even in federated mode. Whether you're experimenting with new designs or creating production-quality visuals, FLUX-1 has you covered.
Try it out at demo.localai.io and see what LocalAI + FLUX-1 can do!
🐛 Diffusers and hipblas Fixes
Great news for AMD users! If you’ve encountered issues with the Diffusers backend or hipblas, those bugs have been resolved. We’ve transitioned to uv for managing Python dependencies, ensuring a smoother experience. For more details, check out Issue #1592.
🏎️ Strict Mode for API Compliance
To stay up to date with OpenAI’s latest changes, now LocalAI have support as well for Strict Mode ( https://openai.com/index/introducing-structured-outputs-in-the-api/ ). This new feature ensures compatibility with the most recent API updates, enforcing stricter JSON outputs using BNF grammar rules.
To activate, simply set strict: true in your API calls, even if it’s disabled in your configuration.
Key Notes:
- Setting
strict: trueenables grammar enforcement, even if disabled in your config. - If
format_typeis set tojson_schema, BNF grammars will be automatically generated from the schema.
🛑 Disable Gallery
Need to streamline your setup? You can now disable the gallery endpoint using LOCALAI_DISABLE_GALLERY_ENDPOINT. For more options, check out the full list of commands with --help.
🌞 P2P and Federation Enhancements
Several enhancements have been made to improve your experience with P2P and federated clusters:
- Load Balancing by Default: This feature is now enabled by default (disable it with
LOCALAI_RANDOM_WORKERif needed). - Target Specific Workers: Directly target workers in federated mode using
LOCALAI_TARGET_WORKER.
💪 Run Multiple P2P Clusters in the Same Network
You can now run multiple clusters within the same network by specifying a network ID via CLI. This allows you to logically separate clusters while using the same shared token. Just set LOCALAI_P2P_NETWORK_ID to a UUID that matches across instances.
Please note, while this offers segmentation, it’s not fully secure—anyone with the network token can view available services within the network.
🧪 Deprecation Notice: gpt4all.cpp and petals Backends
As we continue to evolve, we are officially deprecating the gpt4all.cpp and petals backends. The newer llama.cpp offers a superior set of features and better performance, making it the preferred choice moving forward.
From this release onward, gpt4all models in ggml format are no longer compatible. Additionally, the petals backend has been deprecated as well. LocalAI’s new P2P capabilities now offer a comprehensive replacement for these features.
What's Changed
Breaking Changes 🛠
Bug fixes 🐛
- fix(ui): do not show duplicate entries if not installed by gallery by @mudler in #3107
- fix: be consistent in downloading files, check for scanner errors by @mudler in #3108
- fix: ensure correct version of torch is always installed based on BUI… by @cryptk in #2890
- fix(python): move accelerate and GPU-specific libs to build-type by @mudler in #3194
- fix(apple): disable BUILD_TYPE metal on fallback by @mudler in #3199
- fix(vall-e-x): pin hipblas deps by @mudler in #3201
- fix(diffusers): use nightly rocm for hipblas builds by @mudler in #3202
- fix(explorer): reset counter when network is active by @mudler in #3213
- fix(p2p): allocate tunnels only when needed by @mudler in #3259
- fix(gallery): be consistent and disable UI routes as well by @mudler in #3262
- fix(parler-tts): bump and require after build type deps by @mudler in #3272
- fix: add llvm to extra images by @mudler in #3321
- fix(p2p): re-use p2p host when running federated mode by @mudler in #3341
- fix(ci): pin to llvmlite 0.43 by @mudler in #3342
- fix(p2p): avoid starting the node twice by @mudler in #3349
- fix(chat): re-generated uuid, created, and text on each request by @mudler in #3359
Exciting New Features 🎉
- feat(guesser): add gemma2 by @sozercan in #3118
- feat(venv): shared env by @mudler in #3195
- feat(openai): add
json_schemaformat type and strict mode by @mudler in #3193 - feat(p2p): allow to run multiple clusters in the same p2p network by @mudler in #3128
- feat(p2p): add network explorer and community pools by @mudler in #3125
- feat(explorer): relax token deletion with error threshold by @mudler in #3211
- feat(diffusers): support flux models by @mudler in #3129
- feat(explorer): make possible to run sync in a separate process by @mudler in #3224
- feat(federated): allow to pickup a specific worker, improve loadbalancing by @mudler in #3243
- feat: Initial Version of vscode DevContainer by @dave-gray101 in #3217
- feat(explorer): visual improvements by @mudler in #3247
- feat(gallery): lazy load images by @mudler in #3246
- chore(explorer): add join instructions by @mudler in #3255
- chore: allow to disable gallery endpoints, improve p2p connection handling by @mudler in #3256
- chore(ux): add animated header with anime.js in p2p sections by @mudler in #3271
- chore(p2p): make commands easier to copy-paste by @mudler in #3273
- chore(ux): ...
v2.20.0
TL;DR
- 🌍 Explorer & Community: Explore global community pools at explorer.localai.io
- 👀 Demo instance available: Test out LocalAI at demo.localai.io
- 🤗 Integration: Hugging Face Local apps now include LocalAI
- 🐛 Bug Fixes: Diffusers and hipblas issues resolved
- 🎨 New Feature: FLUX-1 image generation support
- 🏎️ Strict Mode: Stay compliant with OpenAI’s latest API changes
- 💪 Multiple P2P Clusters: Run multiple clusters within the same network
- 🧪 Deprecation Notice:
gpt4all.cppandpetalsbackends deprecated
🌍 Explorer and Global Community Pools
Now you can share your LocalAI instance with the global community or explore available instances by visiting explorer.localai.io. This decentralized network powers our demo instance, creating a truly collaborative AI experience.
How It Works
Using the Explorer, you can easily share or connect to clusters. For detailed instructions on creating new clusters or connecting to existing ones, check out our documentation.
👀 Demo Instance Now Available
Curious about what LocalAI can do? Dive right in with our live demo at demo.localai.io! Thanks to our generous sponsors, this instance is publicly available and configured via peer-to-peer (P2P) networks. If you'd like to connect, follow the instructions here.
🤗 Hugging Face Integration
I am excited to announce that LocalAI is now integrated within Hugging Face’s local apps! This means you can select LocalAI directly within Hugging Face to build and deploy models with the power and flexibility of our platform. Experience seamless integration with a single click!
This integration was made possible through this PR.
🎨 FLUX-1 Image Generation Support
FLUX-1 lands in LocalAI! With this update, LocalAI can now generate stunning images using FLUX-1, even in federated mode. Whether you're experimenting with new designs or creating production-quality visuals, FLUX-1 has you covered.
Try it out at demo.localai.io and see what LocalAI + FLUX-1 can do!
🐛 Diffusers and hipblas Fixes
Great news for AMD users! If you’ve encountered issues with the Diffusers backend or hipblas, those bugs have been resolved. We’ve transitioned to uv for managing Python dependencies, ensuring a smoother experience. For more details, check out Issue #1592.
🏎️ Strict Mode for API Compliance
To stay up to date with OpenAI’s latest changes, now LocalAI have support as well for Strict Mode ( https://openai.com/index/introducing-structured-outputs-in-the-api/ ). This new feature ensures compatibility with the most recent API updates, enforcing stricter JSON outputs using BNF grammar rules.
To activate, simply set strict: true in your API calls, even if it’s disabled in your configuration.
Key Notes:
- Setting
strict: trueenables grammar enforcement, even if disabled in your config. - If
format_typeis set tojson_schema, BNF grammars will be automatically generated from the schema.
🛑 Disable Gallery
Need to streamline your setup? You can now disable the gallery endpoint using LOCALAI_DISABLE_GALLERY_ENDPOINT. For more options, check out the full list of commands with --help.
🌞 P2P and Federation Enhancements
Several enhancements have been made to improve your experience with P2P and federated clusters:
- Load Balancing by Default: This feature is now enabled by default (disable it with
LOCALAI_RANDOM_WORKERif needed). - Target Specific Workers: Directly target workers in federated mode using
LOCALAI_TARGET_WORKER.
💪 Run Multiple P2P Clusters in the Same Network
You can now run multiple clusters within the same network by specifying a network ID via CLI. This allows you to logically separate clusters while using the same shared token. Just set LOCALAI_P2P_NETWORK_ID to a UUID that matches across instances.
Please note, while this offers segmentation, it’s not fully secure—anyone with the network token can view available services within the network.
🧪 Deprecation Notice: gpt4all.cpp and petals Backends
As we continue to evolve, we are officially deprecating the gpt4all.cpp and petals backends. The newer llama.cpp offers a superior set of features and better performance, making it the preferred choice moving forward.
From this release onward, gpt4all models in ggml format are no longer compatible. Additionally, the petals backend has been deprecated as well. LocalAI’s new P2P capabilities now offer a comprehensive replacement for these features.
What's Changed
Breaking Changes 🛠
Bug fixes 🐛
- fix(ui): do not show duplicate entries if not installed by gallery by @mudler in #3107
- fix: be consistent in downloading files, check for scanner errors by @mudler in #3108
- fix: ensure correct version of torch is always installed based on BUI… by @cryptk in #2890
- fix(python): move accelerate and GPU-specific libs to build-type by @mudler in #3194
- fix(apple): disable BUILD_TYPE metal on fallback by @mudler in #3199
- fix(vall-e-x): pin hipblas deps by @mudler in #3201
- fix(diffusers): use nightly rocm for hipblas builds by @mudler in #3202
- fix(explorer): reset counter when network is active by @mudler in #3213
- fix(p2p): allocate tunnels only when needed by @mudler in #3259
- fix(gallery): be consistent and disable UI routes as well by @mudler in #3262
- fix(parler-tts): bump and require after build type deps by @mudler in #3272
- fix: add llvm to extra images by @mudler in #3321
- fix(p2p): re-use p2p host when running federated mode by @mudler in #3341
- fix(ci): pin to llvmlite 0.43 by @mudler in #3342
- fix(p2p): avoid starting the node twice by @mudler in #3349
- fix(chat): re-generated uuid, created, and text on each request by @mudler in #3359
Exciting New Features 🎉
- feat(guesser): add gemma2 by @sozercan in #3118
- feat(venv): shared env by @mudler in #3195
- feat(openai): add
json_schemaformat type and strict mode by @mudler in #3193 - feat(p2p): allow to run multiple clusters in the same p2p network by @mudler in #3128
- feat(p2p): add network explorer and community pools by @mudler in #3125
- feat(explorer): relax token deletion with error threshold by @mudler in #3211
- feat(diffusers): support flux models by @mudler in #3129
- feat(explorer): make possible to run sync in a separate process by @mudler in #3224
- feat(federated): allow to pickup a specific worker, improve loadbalancing by @mudler in #3243
- feat: Initial Version of vscode DevContainer by @dave-gray101 in #3217
- feat(explorer): visual improvements by @mudler in #3247
- feat(gallery): lazy load images by @mudler in #3246
- chore(explorer): add join instructions by @mudler in #3255
- chore: allow to disable gallery endpoints, improve p2p connection handling by @mudler in #3256
- chore(ux): add animated header with anime.js in p2p sections by @mudler in #3271
- chore(p2p): make commands easier to copy-paste by @mudler in #3273
- chore(ux): allow to create and drag dots in the animation by @mudler in #3287
- feat(federation): do not allocate local services for load balancing by @mudler in #3337
- feat(p2p): allow to set intervals b...
Contributors
cryptk, sozercan, and 5 other contributors