Releases: kartikm7/llocal
Releases · kartikm7/llocal
v1.0.0-beta.6
Major Release
✨ Knowledge Base is here with File Upload Support!
- With the power of VectorDB's, LLocal now utilizes a persistent way to store vector embeddings. (Although only applicable to File Uploads at the moment).
- pdf, docx, pptx, csv and txt files are supported for file upload.
🌊 What's Changed
- Delete chat is now here. #5
- Fixed the overflow during code generation. #4
- The more button (+) near the text input, has an added file upload section.
- Settings has been revamped now with two sections, Settings (General) and Knowledge Base.
- You can delete the saved VectorDB's through the knowledge base section in settings.
- There's also an auto-complete pop-up for easy navigation through existing VectorDB's by using a / forward slash at the beginning of your prompt.
❄️ Knowledge Base (More information)
- The knowledge base uses a basic similarity search at the moment, which retrieves relevant chunks quickly unlike agentic solutions. Although the results are still not up-to the mark.
- Further, on every RAG based prompt the response also has a sources table attached to show where the retrieved chunks are in the actual file.
Full Changelog: v1.0.0-beta.5...v1.0.0-beta.6
v1.0.0-beta.5
Major Release
✨ Web Search, Image Upload & Linux Support is here!
- Experimental web search is here, you can submit a link as well as perform a duckduckgo search.
- Linux support is here, although the logo of llocal is not visible but the functionalities work just as on the other platforms.
- The auto-install of Ollama works only on debian distros. (Atleast only tested on debian systems)
- Image Upload is now available to be utilized by supported vision models (More Information Here)
🌊 What's Changed
- Copy code block is now available with Syntax Highlight for almost every language out there (with the atom theme haha)!
- Tabular format support is here.
- Stop Generating button is here, although the position is debatable.
- Fixed changing chats while streaming, now you can only switch between chats when the text is not streaming.
- More button (+) in the input for text, which allows to toggle on and off web search aswell as supports images.
❄️ Web Search (More information)
- Web search toggle, has two inbuilt modes a webscraper mode aswell as a duckduckgo search.
- Web scraper : You can simply paste a link in the prompt along with your query and llocal will handle the rest. (Eg. What are the 3 endings in Sekiro? https://en.wikipedia.org/wiki/Sekiro:_Shadows_Die_Twice )
- Web search : This performs a web search, based on a prompt. It returns the top 10 results and a very small summary of each result from the duckduckgo api.
⚡ New Contributors
- @FruitPnchSamuraiG made their first contribution in #3
Full Changelog: v1.0.0-beta.4...v1.0.0-beta.5
1.0.0-beta.4
Major Release
Official Mac Support
- Mac support is here. This has been one of the major goals for the project.
- Ollama, can be downloaded and prompted to be installed if detected as not installed.
What's Changed
- refactor: improve error handling in checkOllama function by @virajbhartiya in #1
- fix : partial mac setup by @dandonarahul2002 in #2
New Contributors
- @virajbhartiya made their first contribution in #1
- @dandonarahul2002 made their first contribution in #2
Full Changelog: v1.0.0-beta.2...v1.0.0-beta.4
1.0.0-beta.3
v1.0.0-beta.3 feat: autoupdater
1.0.0-beta.2
v1.0.0-beta.2 feat: autoupdater
1.0.0-beta.1
Features
Functional Features
- Ollama auto-downloader
- Ollama auto-installer
User experience features
- New logo (I think it might just count as a feature)
1.0.0-beta
Features
Functional Features
- Ollama support (both, pulling models and chatting)
- Local data storage (no cloud connectivity)
- Chat streaming
- Support to switch between models
UX Features:
- Theming options ( across 5 themes)
- Dark mode and light mode support
Note: It is still in a pre-release phase, so all features might not work as intended.