Build an Agentic RAG AI App with Phidata, Qdrant, OllamaEmbedder & DeepSeek-R1 (No OpenAI API!)

In this project, an Agentic RAG AI Application has been built in Python with Phidata, Qdrant (vector db), OllamaEmbedder and Deepseek-r1 model (Open Source LLM). Qdrant is an open source Vector database of Phidata framework and we set it up locally through docker container. For embedding, we used OllamaEmbedder where "openhermes" model has been used after getting pulled through Ollama locally. There were no dependencies on OpenAI API key in this AI agentic RAG system. Traditional RAG has lot of limitations and it is suitable for only fixed set of questions and answers

for a context and it can't answer a question if the exact/matching information is not found in the retrived data, where as, Agentic RAG is quite smart, autonomous and it has some decision making capability. It will first go through the knowledge document and understand it, then if it finds the relevant informtion there, it will combine that with the LLM training knowledge and answer the query in a very structured and informative way with details. If it can't find the answer that way, it will take help from external tools like google search (if we provide that) and give the answer. In the App, user needs to upload the document (pdf) and also type the query in text box and then they just need to click on the "Answer" button.

GitHub Link: https://github.com/dharsandip/phidata_agentic_rag_qdrant_deepseek LinkedIn: https://www.linkedin.com/in/sandip-dhar-40145546/

#agenticrag, #phidata, #qdrant, #deepseek, #vectordatabase, #qdrantvectordatabase, #ollamaembedder, #groq, #agenticragai, #aiagents, #pythonproject, #streamlit