[KongchangAI]
ProductMuffakir RAG Runner

Muffakir

An open-source RAG experiment runner that supports automated combinatorial experiments across dimensions including chunking strategy, embedding model, retrieval method, query transformation, reranking, top-k, and generation model. It provides quantitative comparisons of retrieval metrics, generation metrics, latency, and cost, with results exportable as Python code.

Timeline (last 90 days)

Sep 20

Muffakir is an open-source RAG experiment runner created by a developer.

Unverified50%
Sep 20

Muffakir supports experimental comparison of RAG components across dimensions including chunking, embedding models, retrieval, query transformation, reranking, top-k, and generation models.

Unverified50%
Sep 20

Muffakir automatically runs various RAG component combinations as reproducible trials.

Unverified50%
Sep 20

Muffakir uses four dimensions—retrieval metrics, generation metrics, latency, and cost—to compare the results of each trial.

Unverified50%
Sep 20

Muffakir allows users to export selected configurations as Python code for use in production environments.

Unverified50%
Sep 20

Muffakir currently runs locally and is still in early development.

Expired50%
Sep 20

The new package version of Muffakir has not yet been published to PyPI, and at this stage it can only be installed from the source code in the GitHub repository.

Expired50%
Sep 20

The GitHub repository address for Muffakir is github.com/Mohamed28112003/Muffakir.

Unverified50%
Sep 20

The developer's primary purpose in creating Muffakir was to solicit feedback from people who are actually building RAG systems.

Expired50%

All Facts (6)

Source Articles