README.md

metadata

title: Rag
emoji: 🐢
colorFrom: gray
colorTo: indigo
sdk: gradio
sdk_version: 5.12.0
app_file: app.py
pinned: false
short_description: Just another rag but with Images 🖼️

Just another RAG, dont bother much!!

File Descriptions

z_document_reader.py

Images are only useful to limited resource computer if it has a caption. So this file helps parse the wikipedia html strips it off the tags.

z_embedding.py

Generates vector store.

z_generate.py

Use LLM and prompting to find the relevant texts and images stored in the vector stores.

Adding more data sources

Currently limited to wikipedia pages downloaded as HTML.

1. Place the html in the folder _data
2. Run command python z_embedding.py
3. Output will be two FAISS vectors stores in the folder cache_vector...

Local Debug

Highly recommend VS Code, makes life easy.

1. Create Virtual environment with name sb-rag using the below command python -m venv sb-rag

2. Activate the environemnt (create new terminal VS Code to automatically do so)

3. Edit .vscode/launch.json. Fill in the environment variable HF_SERVERLESS_API.

4. Start VS Code debugger.

References

1. UI Blocks Concepts: https://huggingface.co/learn/nlp-course/en/chapter9/7
2. UI Row-Column Arrangement: https://www.gradio.app/guides/controlling-layout
3. Show caption in image gallery: https://github.com/gradio-app/gradio/issues/3364
4. HF Implementation of basic: https://huggingface.co/learn/cookbook/en/advanced_rag
5. https://python.langchain.com/docs/integrations/vectorstores/faiss/

Ideas

1. Shows frames of design patterns: https://www.falkordb.com/blog/advanced-rag/
2. HF Implementation of basic: https://huggingface.co/learn/cookbook/en/advanced_rag
3. HF RAG Evaluation: https://huggingface.co/learn/cookbook/en/rag_evaluation
4. HF Implementation by someone: https://medium.aiplanet.com/advanced-rag-implementation-on-custom-data-using-hybrid-search-embed-caching-and-mistral-ai-ce78fdae4ef6
5. HF Agentic Rag: https://huggingface.co/learn/cookbook/en/agent_rag
6. Future read, tooning https://huggingface.co/blog/lucifertrj/finetune-embeddings
7. Opinion on instruct embeddings: https://huggingface.co/blog/Tonic/instruct-embeddings-and-advanced-rag
8. Another Implementation: https://huggingface.co/learn/cookbook/en/rag_zephyr_langchain
9. Another Opinion on ray: https://www.anyscale.com/blog/retrieval-augmented-generation-with-huggingface-transformers-and-ray
10. Ray Follow up: https://github.com/run-llama/ai-engineer-workshop/blob/main/presentation.pdf?__s=2il5g6hpfc4mtmydioir
11. llama Index rag implementation: https://docs.llamaindex.ai/en/latest/optimizing/production_rag/
12. Just some termino book: https://www.projectpro.io/article/advanced-rag-techniques/1063
13. Another Evaluation Guide: https://pub.towardsai.net/evaluating-rag-metrics-across-different-retrieval-methods-770aa01380c8
14. Oracle Garbage: https://blogs.oracle.com/ai-and-datascience/post/ai-health-mixtral-oracle-23ai-rag-langchain-streamlit