v0.15.10 - Bug Fix

• Chatbot - Fix error handling bug used to auto-detect max content length of LLM. Updated user input UI rendering to better handle indention.
• News Bot Script - Added logic to verify news summary from LLM to help prevent hallucinations.

Full Changelog: v0.15.9...v0.15.10

v0.15.9 - Vision Model Support

What's Changed

0.15.9 - DocMan Auth

• DocMan - Add basic authentication and secure connection options to Weaviate.

0.15.8 - Enhance Image Processing

• Chatbot - Add support for HEIC file type and resize all images to max dimensions of 1024. Handle image pasting into input field. Remove previous images from context thread.
• Chatbot - Clean up logging: non-critical logs are moved to DEBUG level.

0.15.7 - Vision Model Support

• Chatbot - Allows user to drag and drop images into the context window for multi-modal vision LLMs.

0.15.6 - Progressive Loading

• DocMan - Updated to use progressive loading to help with larger document and chunk lists. Performance and bug fixes.
• Chatbot - Updated /rag commands to allow turning auto-RAG on and off, setting the collection and result number.

0.15.5 - Async and SocketIO

• DocMan - Switch to async and socket communication to more responsive UI. Bug fixes.

0.15.2 - Weaviate Client Updates

• Chatbot and DocMan: Provide control for WEAVIATE_HOST and WEAVIATE_GRPC_HOST (and PORTs) settings separately via environmental variables.
• DocMan: Bug fixes

0.15.1 - Document Manager Updates

• DocMan: Fix some bugs and add features to process more document types (file or URL).
• Update version to v0.15.1 by @jasonacox in #11

Full Changelog: v0.15.0...v0.15.9

v0.15.0 - Document Manager

What's Changed

• Document Management - v0.15.0 by @jasonacox in #10
• Chatbot: Using Document class for RAG functions.
• DocMan: New web based UI for managing documents in the Weaviate vector database. Allows user to upload and embed content from URLs and uploaded files. Provides optional chunking and management of embedded documents.
• Screenshots:
  
  

Full Changelog: v0.14.13...v0.15.0

v0.14.13

0.14.13 - TPS Calculation

• Chatbot: Fix a bug that was counting null tokens.

0.14.12 - Toxic Filter

• Chatbot: Add toxic filter option (uses environmental variable TOXIC_THRESHOLD) to analyze and filter out bad prompts. Uses LLM to evaluate and score prompt. Set variable between 0 and 1 or 99 to disable (default).
• Chatbot: Add EXTRA_BODY variable (JSON string) to customize chat completion calls.

0.14.11 - OpenAI Support

• Chatbot: Add logic to detect OpenAI URL and disable non-OpenAI stop_token_ids.

Full Changelog: v0.14.10...v0.14.13

v0.14.10 - Minor Updates

Updates

• Chatbot v0.14.10: Fix issue where DOM was being corrupted by popup. New logic creates separate div for conversation debug.
• Chatbot v0.14.9: Add Debug Session link to footer to display conversation thread.
• Chatbot v0.14.8: Update RAG to remove duplicate documents.
• Chatbot v0.14.7: Update TemplateResponse arguments to current format as reported in #7.

Full Changelog: v0.14.6...v0.14.10

v0.14.6 - Ollama & News Updates

Chatbot Updates

• Expand /news/ RAG command to include reference URL links in news article headlines.
• Add response statistics (number of tokens and tokens per second) to footer.
• Serve up local copy of socket.io.js library to help with air-gap installations.

Ollama Support

• Add logic to chatbot to support OpenAI API servers that do not support the /v1/models API. This allows the Chatbot to work with Ollama provided the user specifies the LLM_MODEL, example docker run script:

docker run \
    -d \
    -p 5000:5000 \
    -e PORT=5000 \
    -e OPENAI_API_KEY="Asimov-3-Laws" \
    -e OPENAI_API_BASE="http://localhost:11434/v1" \
    -e LLM_MODEL="llama3" \
    -e USE_SYSTEM="false" \
    -e MAXTOKENS=4096 \
    -e TZ="America/Los_Angeles" \
    -v $PWD/.tinyllm:/app/.tinyllm \
    --name chatbot \
    --restart unless-stopped \
    jasonacox/chatbot

Full Changelog: v0.14.4...v0.14.6

v0.14.4 - Llama-3 Support

v0.14.4 - Llama-3 Support

• Add chatbot workaround for Meta Llama-3 support via stop token addition.
• Add logic to better handle model maximum context length errors with automated downsizing.
• Error handling and auto-retry for model changes on LLM.

v0.14.3 - Resize Control

• Add intuitive UI control at top of user input area to allow user to resize text input box.

v0.14.2 - Chatbot Stock RAG

• Add error checking and help for /stock {company} command.
• Allow user input textarea to be resized vertically.

v0.14.1 - Chatbot Baseprompt

• Fixed bug with baseprompt updates to respond to saved Settings or new sessions.
• Updated baseprompt to include date and guidance for complex and open-ended questions.
• Add TZ local timezone environmental variable to ensure correct date in baseprompt.

Full Changelog: v0.14.0...v0.14.1

v0.14.0 - Chatbot Controls

• Added ability to change LLM Temperature and MaxTokens in settings.
• Added optional prompt settings read-only options to allow viewing but prevent changes (PROMPT_RO=true).

Full Changelog: v0.13.0...v0.14.0

v0.13.0 - Use Weaviate for RAG

What's Changed

• Moved from Qdrant to Weaviate - This externalizes the sentence transformation work and lets the chatbot run as a smaller service. Activate by setting WEAVIATE_HOST to the address of the DB
• Added "References" text to output from /rag queries.
• Added ONESHOT environmental variable that if True will remove conversation threading allowing each query to be answered as a standalone sessions.
• Added RAG_ONLY environmental variable that if True will assume all queries should be directed to the default RAG database as set by WEAVIATE_LIBRARY.
• See #5

docker run \
    -d \
    -p 5000:5000 \
    -e PORT=5000 \
    -e OPENAI_API_BASE="http://localhost:8000/v1" \
    -e ONESHOT="true" \
    -e RAG_ONLY="false" \
    -e WEAVIATE_HOST="localhost" \
    -e WEAVIATE_LIBRARY="tinyllm" \
    -v $PWD/.tinyllm:/app/.tinyllm \
    --name chatbot \
    --restart unless-stopped \
    jasonacox/chatbot

Full Changelog: v0.12.5...v0.13.0