This repository contains the AI based knowledgebase. Based on research papers stored in a vector database, articles can be generated using predefined templates. These templates define the sections, titles and paragraphs which should be filled using an LLM.
- chroma vector database -> selfhosted
- firestore noSQL database -> google cloud
- web API -> python fast api
- frontend -> react
- llm framework -> haystack
- datasource sientific papers -> openalex
- LLM -> mostly ollama run localy (but interchangable)
- ollama -> local hosting llm
api back end
uvicorn main:app --reloadfrontend
bun run devTemplates can be created using the appropriate endpoint, they follow the structure outlined in 'pipelines/sample_template.json'.