RobinQu

more complex rag pipeline may invole agent frameworks #18 [Adaptive RAG](https://arxiv.org/abs/2403.14403)  https://github.com/langchain-ai/langgraph/blob/70c1c996a4c9fe8df518bcd849b3c6453dd0d58b/examples/rag/langgraph_adaptive_rag.ipynb

# Rerank, Colbert, ... ## Related work ### Methods * Keyword based - BM25 * Cross-encoders BCE reranker * Late-interaction - ColBERT, BGE-M3 * [2004.12832v2.pdf](https://github.com/RobinQu/instinct.cpp/files/15373991/2004.12832v2.pdf) * https://zhuanlan.zhihu.com/p/683483778 ### Opensourced Projects...

# Timeline * #21 * [x] experiemntes with BERT + BGE M3 model for better understanding of forward pass, ~3 days~ , 1 day. * [x] bge-m3 inference code. A...

OpenAI officials parameter for RAG: https://platform.openai.com/docs/assistants/tools/file-search/how-it-works > By default, the file_search tool uses the following settings: > Chunk size: 800 tokens > Chunk overlap: 400 tokens > Embedding model: text-embedding-3-large...

wont't compile in gcc 13 due to problem of pdfium recipe: https://github.com/conan-io/conan-center-index/issues/23529

intial notebook is located at https://github.com/RobinQu/instinct.cpp/blob/sprint/v0.1.1/modules/instinct-examples/doc-agent/test/rag_evaluation/rag_evaluation.ipynb

Inconsistency problem of `RecursiveCharacterSplitter` is resolved. Evaulation should be fair now.

Close this issue. The evaulation script is stable since last version.

> I found solution in other issue ([#1573](https://github.com/abetlen/llama-cpp-python/issues/1573)), this install command worked for me: CMAKE_ARGS="-DGGML_CUDA=on -DLLAVA_BUILD=off" pip install llama-cpp-python --upgrade --force-reinstall --no-cache-dir Solved the pip install , but why?

I agree with you that callbacks are important. Event-driven design may be more diffucult to implmenet in cpp. But I've already introduced dependencies like `rpp` so it would be straightforward...