Thomas BrochardDevelopment · Data · AI
Let’s talk
Back to selected work

Retrieval systems

Case study / 09

Cultural Events RAG

Carry source evidence through the retrieval path.

Connect natural-language questions to cultural events through metadata-aware retrieval, grounded answers and source references.

Technologies & concepts

  • Python
  • LangChain
  • FAISS
  • Mistral embeddings
  • Gemini / Mistral
  • Streamlit

Technical demonstration / selected source

Private repository

01 / The context

The problem behind the design.

Event catalogues contain long descriptions, repeated sessions and inconsistent metadata. Natural-language discovery needs to connect questions to relevant events while preserving dates, locations and the original source links.

02 / Engineering decisions

Choices made explicit.

Preserve event metadata

The ingestion and chunking workflow keeps event identity, dates, location and source references beside text. Retrieval prefers candidates overlapping months mentioned in French or English questions.

Provide bounded fallbacks

A lexical path handles unavailable embeddings or rate limits. When generation is limited, an extractive event list exposes retrieved metadata.

Separate evaluation signals

Retrieval recall and reference-based answer judgment are assessed independently. Per-stage instrumentation distinguishes ingestion, cleanup, chunking, embedding, search and generation.

03 / System view

Follow the flow.

Cultural-events retrieval design: a normalized catalogue becomes metadata-aware chunks for vector retrieval and grounded answers with source references.

Illustrative architecture from the project presentation. It describes the selected design; it does not imply a deployed service or a measured operational result.

View the full-size diagram
  1. Event catalogue

  2. Normalize and filter

  3. Metadata-aware chunks

  4. Vector retrieval

  5. Answer + sources

04 / Selected artifacts

What the repository contains.

  • scripts/fetch_events.py
  • scripts/preprocess_events.py
  • scripts/build_vector_store.py
  • scripts/rag.py
  • scripts/evaluate_rag.py
  • scripts/benchmark_pipeline.py
  • docs/retrieval-design.md
  • docs/evaluation.md

05 / Scope & limitations

How to read this project.

Demonstration and source showcase without a hosted assistant, ready-made catalogue, provider credentials, serialized FAISS indexes or historical evaluation results. The retained settings use one example city. Date preference may fill remaining slots outside the requested period and is not a strict filter. Conversation memory and multi-city discovery are outside scope. No accuracy, latency, availability or test-pass claims are made.

Explore the source with access.

The GitHub repository is private. Its link works for authorized collaborators; other visitors can contact me to discuss the project.

Next case study / 10

RAG Platform Architecture

Retrieval systems

Explore the collection

The next conversation

Let’s talk about your project.

↗
A need, an idea, a technical question.
LinkedIn ↗GitHub ↗