v1.1.0 - Conversational Mode

New Features

• Conversational Mode: A new mode setting has been added to config.ini.
  • pure_rag: The original behavior. Every message triggers a new search.
  • conversational: Only the first message in a chat triggers a RAG search. Follow-up messages will be about the initial context.

Other Changes

• Updated README.md with new instructions.
• Updated config.ini.template with the new mode setting.

v1.0.0 - Initial Public Release

• Universal ZIM Ingestion: Process any ZIM file into a searchable knowledge base.
• Robust Build Process: The indexer (build.py) is designed to be interruptible and can be resumed at any time, saving progress automatically. It supports both CPU and GPU for embedding creation.
• Flexible LLM Support: The server acts as a smart proxy, compatible with:
  • Offline Local Models: openai-compatible servers like LM Studio, GPT4All, and Jan.
  • Offline Ollama Models: Native support for local Ollama instances.
  • Online API Providers: Full support for OpenAI, Anthropic (Claude), and Google (Gemini).
• Verifiable & Cited Answers: The system prompt is engineered to force the LLM to cite the source article for every piece of information it provides.