SemanticSearchRAG 0.3.0
Description
SemanticSearchRAG- semantic search: text gets tokenized, and you can search text without knowing the exact phrase or order of worda.
- the default multi-language tokenizer lets you search documents without using the document language.i.e. you can search german and english files just with german, and you will also get matching meanings in the english files.
- full text search is available,but not needed an can be deactivated
- there are 3 factory actions: compare,summarize and check (for logic errors). you copy a file or more in the corresponding folder and get the result file.
- you can add your own prompts.md to the prompt folder as well. e.g. if you want to translate text to a language that has three different articles for spoon,fork,knife, you can tell the ai to translate the input file to german.
- if you want it to rewrite your input to riddles and rhymes,add a new rhyme.md with your prompt. you get the idea.
- i implemented 2 document categories: open and restricted. open is supposed to use cloud , restricted supposed to use local ai. but you can change that.
de facto, it provides RAG-functionality, but it will be more RAG in the next version, i.e. llm after the search.
What’s new
fixes in server configuration
fixes
fixes
