DSpy Ollama Custom Examples with Custom Ollama based Retriever for ChromaDB

Anchknowledgements and premise of writng this code

The methods and ways mentionned in most of the DSPy examples that uses ChromadbRM ,ColBERTv2, MilvusRM, QdrantRM, WeaviateRM etc fails to consider these points mentionned below:

1. Everybody is Not a seasoned progrmmaer who can easily spin-up local databases, web servers on a whim!
2. LLM != GPT-4 or GPT-3.5 or Grok or Claude or any other bigshot expensive money burning exernal api keys that makes a big hole in a Dev's Pocket and Morale
3. I Whole heartedly respect and love all those investors and Scientists who made it possible to develop and use all these Greate LLms, But Being a Biologist to turned himself from coding Dna to Coding LLMs, I believe

Life (and a pair of Kanjoose Bengali Engineers 😉 ) always finds a way! (To get away with doing things easy and for free 😝)

4. Sorry for diverting into the cavernous side alleys of insomniac thoughts.
5. And another issue is that everyone is not a guy who uses pthon with OOPs, and this can be a big OOps!! for a lot of newcommers =nc

What did I do

Well well, being a subset of nc myself (with a history of making complex programms in Blender's Geometry node) and having blessed with Guides, friends and Youtubers Like Arnav Singhvi & Herumb Shandilya from DSPy Connor Shorten from Weaviate and Tamoghna Das, My College senior who brought me into this world of LLms in November 2022. And The last but Not the least Perplexity for enabling millions like me who are a subset of nc to code without fear of syntaxes.

1. Studied the Ollama Library in python that is given by ollama and make an embedder:retiever function pair
2. made a local server of Ollama and Chromadb run from a given directory from urls http://localhost:11434/ and http://localhost:8000/ respectively.
3. Used NLTK's Sent_tokenize for chunking the text into sentences that will be fed to the ChromaDB
4. Made a Ollama based LLM run in DSPy
5. Created the default GenerateAnswer and GenerateSearchQuery Signatures for Rag (Never used the GenerateSearchQuery for being a noob myself 😝)
6. Created a markdown file that has the data I am doing RAG on (in a later example I will show you how to do rag on web search data)