Releases: helixml/helix
1.0.11 - better text extraction in RAG
What's Changed
- Feature/readability filter to extract text from html by @rusenask in #488
- typesense retries on connect by @rusenask in #489
- go via localhost, otherwise keycloak cookies won't work by @lukemarsden in #490
- fix user prompt visibility by @rusenask in #491
Full Changelog: 1.0.9...1.0.10
1.0.9 - fix firewalled typesense; knowledge search + UI, image fine tuning
What's Changed
Makes RAG work in firewalled environments, adds a new knowledge search UI so you can see what chunks got indexed from knowledge:
Fixes regression in image fine-tuning.
- Feature/search api by @rusenask in #478
- Feature/search embeddings through UI by @rusenask in #479
- Feature/knowledge deletion + refresh by @rusenask in #480
- message by @rusenask in #481
- fix image fine tuning - missing import by @lukemarsden in #482
- cleanup unused funcs by @rusenask in #483
- add dockerfile for typesense baked weights; update refs by @lukemarsden in #486
Full Changelog: 1.0.8...1.0.9
1.0.8 - Fix runner overschedling
What's Changed
- Feature/knowledge name fix by @rusenask in #475
- Feature/knowledge status second try by @rusenask in #476
- fix(runner): prevent too many models being scheduled on a GPU by @philwinder in #477
Full Changelog: 1.0.7...1.0.8
1.0.7
What's Changed
- fixes UI crash on preview gen by @rusenask in #473
- fixes typesense crash due to high concurrency by defaulting to a concurrency of 1 in a16703e
Full Changelog: 1.0.6...1.0.7
1.0.6
What's Changed
- fix(runner): prevent panic on qa pair generation by @philwinder in #470
- feat(helm): upgrade helm chart to include typesense and tika by @philwinder in #471
- fix app model name parsing by @rusenask in #472
Full Changelog: 1.0.5...1.0.6
1.0.5 - model picker in app editor
What's Changed
Makes App Editor UI work when using external LLM providers
- model picker by @lukemarsden in #469
Full Changelog: 1.0.4...1.0.5
1.0.4 - installer improvements
What's Changed
- fix: explicitly set inference and qapair provider in installer by @lukemarsden in #467
- feature: autodetect ollama in installer by @lukemarsden in #468
Full Changelog: 1.0.3...1.0.4
1.0.3 - fix compose
Fix regression in compose file used for default installation
Full Changelog: 1.0.2...1.0.3
1.0.2 - knowledge uploads in CLI, cron app triggers, multi-provider, much smaller RAG
What's Changed
- Feature/helix filestore knowledge by @rusenask in #453
- Feature/multi provider support by @rusenask in #463
- Feature/cron (periodic) apps by @rusenask in #464
- Feature/rag with typesense by @rusenask in #465
- single url parsing fix with typesense and tika by @rusenask in #466
Full Changelog: 1.0.1...1.0.2
1.0.1 - Private GenAI Stack, OpenAI You Can Run Locally - helm + gptscript fixes
What's Changed
Today, we’re announcing Helix 1.0, because after a major refactor and months of hammering we are now declaring Helix stable and production-ready, ready for serious production use-cases.
Easy installation
Linux | macOS | Windows (WSL2)
Installation is now as simple as:
curl -sL -O https://get.helix.ml/install.sh && bash install.sh
Launch blog: https://blog.helix.ml/p/announcing-helix-10-secure-local
Demo: https://youtu.be/6QcOXq3VFpc
Installation Quickstart: https://docs.helix.ml/helix/private-deployment/controlplane/
Major features in 1.0
- Helix Apps, version controlled configuration for LLM-based applications
- Knowledge, continuously updated RAG from a URL
- API integrations so your app can call an API to get up to date information when needed
- New Helix App Editor UI
- New easy installer with support for Helix running on macOS (alongside Ollama) and Windows on an NVIDIA GPU
This release
Full Changelog: 0.10.18...1.0.1