ezechiel-sawadogo-nlp/merimee-rag

How ezechiel-sawadogo-nlp/merimee-rag works

merimee-rag is a local French-monuments question-answering RAG system, with an accompanying evaluation workflow. Users ask questions in the Streamlit demo; the configured retriever finds passages, the pipeline supplies cited context to an Ollama model, and the UI presents the answer, sources, and map. Corpus ingestion and indexing support that workflow. The evaluation branch builds a test set and measures retrieval and generated answers. Edges reflect the excerpts, SOURCE INDEX references, and README; unsampled implementation details are not inferred.

Main entry points: pipeline.py RAG pipeline · app.py Streamlit demo · eval_generation.py Generation evaluation · store.py Index and system

Components 19

Corpus and indexing

CLI commands src/merimee_rag/cli.py
Mérimée data data/monuments.json.gz
Notice loading src/merimee_rag/corpus.py
Index and system src/merimee_rag/store.py

Question answering

Streamlit demo app.py
Configuration src/merimee_rag/config.py
Retrievers src/merimee_rag/retrievers.py
RAG pipeline src/merimee_rag/pipeline.py
LLM cache src/merimee_rag/llm.py

Evaluation

Test-set builder eval/build_testset.py
Evaluation questions eval/testset.jsonl
Retrieval evaluation eval/eval_retrieval.py
Generation evaluation eval/eval_generation.py
Answer judge eval/judge.py
Evaluation metrics eval/metrics.py
Evaluation report eval/report.py

Other components

Questioner
Solr index
Ollama model

Connections 25

Questioner → Streamlit demo: asks README.md
Streamlit demo → RAG pipeline: submits question app.py
Streamlit demo → Configuration: reads settings app.py
CLI commands → Configuration: uses settings src/merimee_rag/cli.py
CLI commands → Notice loading: loads corpus src/merimee_rag/cli.py
CLI commands → Index and system: builds index src/merimee_rag/cli.py
Mérimée data → Notice loading: loads notices README.md
Notice loading → Index and system: provides notices src/merimee_rag/store.py
Index and system → Retrievers: builds indexes src/merimee_rag/store.py
Index and system → RAG pipeline: provides chunks app.py
RAG pipeline → Retrievers: searches passages src/merimee_rag/pipeline.py
RAG pipeline → Ollama model: generates answer src/merimee_rag/pipeline.py
Streamlit demo → LLM cache: configures cache app.py
Retrievers → Solr index: queries index README.md
Test-set builder → Notice loading: loads notices eval/build_testset.py
Test-set builder → Ollama model: generates questions eval/build_testset.py
Test-set builder → Evaluation questions: writes questions README.md
Evaluation questions → Retrieval evaluation: evaluates retrieval eval/eval_retrieval.py
Evaluation questions → Generation evaluation: evaluates answers eval/eval_generation.py
Retrieval evaluation → Retrievers: queries systems eval/eval_retrieval.py
Generation evaluation → RAG pipeline: runs configurations eval/eval_generation.py
Generation evaluation → Answer judge: scores answers eval/eval_generation.py
Generation evaluation → Evaluation metrics: computes metrics eval/eval_generation.py
Retrieval evaluation → Evaluation metrics: computes metrics eval/eval_retrieval.py
Evaluation report → Configuration: reads paths eval/report.py

More Python diagrams: oca/product-configurator · oshekharo/jio-gemini-activation-scanner · paulonteri/django-serverless-cron · pratham-bhatnagar/gesture-volume-control · riccardo8888/agent-link

Drawn by GitDiagram on September 29, 2026 from the repository's file tree, README and source files. Read as Markdown ·