Skip to content

Latest commit

 

History

History
234 lines (174 loc) · 5.63 KB

CHANGELOG.md

File metadata and controls

234 lines (174 loc) · 5.63 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

[2.1.0] Importastic

Added

  • Added new deployment type: Custom
  • Added new port configuration
  • Added Groq

Fixed

  • Catch Exception when trying to access the OpenAI API Embedding endpoint to retrieve model names
  • Fixed reading empty string as environment variables
  • Fixed default Unstructed URL

[2.0.0] Importastic

Added

  • Async Ingestion with realtime logging

  • Migrated to Weaviate v4 Client

  • Added new File Selection Interface

  • Add Directory Upload

  • Control Settings per file/url individually

  • Import indivdual files or all

  • Overwrite existing files

  • Add multiple labels to documents

  • More configuration for readers, chunkers, and embedders

  • Improved Document Search UI

  • Add Config Validation

  • Add HTML Reader

  • Add Recursive Chunker

  • HTML Chunker

  • Markdown Chunker

  • Code Import

  • Code Chunking

  • Semantic Chunking

  • Label Filter

  • Document Filter (Add document to chat)

  • Add more themes

  • Reworked Admin Interface

  • Added Suggestion View

  • Reworked Suggestion logic

  • Added VoyageAI

  • Added custom metadata

  • Added DocumentExplorer with

    • Content View
    • Chunk View
    • Vector View
      • Visualize vectors of chunks of one or multiple documents
      • PCA

[1.0.3]

Added

  • Cancel Generation Button
  • Added .docx support
  • Added Documentation for JSON Files
  • Added GitLabReader (#151)
  • Improved HuggingFace Embedding Models thanks to @tomaarsen
  • MixedBreadEmbedder
  • AllMPNetEmbedder

Fixed

  • Check error logs coming from Ollama and send it to the frontend
  • Check If Chunks Are NoneType

[1.0.2]

Added

  • Readme Variable: OPENAI_BASE_URL

Fixed

[1.0.0] - Beautiful Verba Update

Added

  • Added DaisyUI
  • Optimized frontend codebase
  • Fully Reworked Verba Design
    • Fully Responsive, optimized for all screen sizes
  • Customization Capabilities
    • Added Default, Darkmode, Weaviate themes
    • Full text, color, image customization
  • Improve Chat Interface
    • Better formatting of markdown + code
    • Keep conversations saved in localBrowser storage
    • Better Debugging by providing more information about current states
  • Improve Document Viewer Interface
    • Add Pagination
    • Add Sorting
    • Use Aggregation for Filtering
  • Improve Status Overview
    • Reworked Frontend + Optimize Code
    • Sort status entries
    • Improve Loading Speed by using Aggregation
  • Improve Component Selection for both Ingestion and RAG
    • Added new configuraiton that will be passed between frontend and backend
    • Cleaned codebase, merged interfaces and managers to single files
    • Added clean endpoints for better code readability
    • Reworked on interfaces
  • Added better console and logging for ingestion
  • More Configuration
    • Enable/Disable Caching and Autocomplete Suggestions
    • Improved verba_config.json
  • Ability to enable/disable caching + autosuggestions
  • Add Google Gemini as new Embedder and Generator
  • Added .CSV support (all file types available in Unstructured IO)
  • More test data
  • Add Ollama as Generator and Embedding Component
  • Add Support for Cohere R+
  • Improved WindowRetriever Context Generation
  • Show RAW Context in Frontend + Save in LocalStorage
  • Save Settings and Configuration in Weaviate

Changed

  • Changed to AppRouter framework
  • Changed frontend project structure
  • Changed backend project structure
  • Removed Llama Generator Component

Fixed

  • Using Accelerator Library

[0.4.0] - 11.04.2024

Added

  • Improved Docker Documentation
  • Improved Docker Settings
  • New Environment Variables for OpenAI proxies: OpenAI_BASE_URL (LiteLLM support) (#56)
  • Increased version

Changed

  • Removed spaCy from project

Fixed

  • Python not working on version 3.12, 3.11, and 3.9
  • GitHub Links on README
  • Fix Docker Default Vectorizer (#50)
  • Fix requirements.txt spelling error
  • Minor Bug fixes

[0.3.1] - 15.11.2023

Added

  • PDFReader powered by PyPDF2
  • TokenChunker powered by tiktoken
  • Ruff Linting (set as pre-commit)
  • Markdown Formatting for chat messages (#48)

Fixed

  • Added missing dependencies
  • Fixed restart bug
  • Fixed MiniLM Cuda to_device bug (#41)
  • Fixed Config Issues (#51)
  • Fixed Weaviate Embedded Headers for Cohere

[0.3.0] - 12.09.2023

Added

  • Refactor modular architecture
  • Add ability to import data through the frontend, CLI, and script
  • Add Readers (SimpleReader, PathReader, GithubReader, PDFReader)
  • Add Chunkers (WordChunker, SentenceChunker)
  • Add Embedders (ADAEmbedder,SentenceTransformer, Cohere)
  • Add Generators (GPT3, GPT4, LLama, Cohere)
  • Status Page
  • Reset functionality
  • Streaming Token Generation
  • Lazy Document Loading
  • Add Copy and Cached Tag
  • Improved Semantic Cache
  • Added LLama 2 and Cohere support
  • Added new OpenAI models
  • Improved Documentation
  • Added technical docs and contribution guidelines

Fixed

  • Error handling for data ingestion (handling chunk size)
  • Schmea handling on startup

Changed

  • Removed Simple- and AdvancedEngine logic

[0.2.3] - 05.09.2023

Added

  • OpenAI API documentation example dataset

[0.2.2] - 31.08.2023

Release!

  • First version of Verba released! (many to come :)

Added

  • Verba favicon

Fixed

  • Add static files to package
  • Weaviate Embedded not shutting down

[0.1.0] - 29.08.2023

Added

  • Prepare Verba for first release