OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
RAG & Knowledge Systems

Bird Search

Bird Search is a Python Marimo teaching notebook that demonstrates and evaluates RAG retrieval methods on about 2,000 North American bird articles paired with photos. It shows Pinecone calls for keyword, dense semantic, visual cross-modal, and combined retrieval, with results graded against a small hand-built evaluation set.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Bird Search — OpenViST project profile in RAG & Knowledge Systems.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it as a teaching or exploration environment when you want to see how several RAG retrieval approaches behave on the same text-and-image corpus and compare them against a small evaluation set.

Who it is for

  • Learners exploring how to select retrieval techniques for RAG.
  • Developers who want to compare keyword, semantic, image-based, and combined retrieval on a concrete corpus.
  • Educators presenting retrieval methods through notebook examples and an evaluation scorecard.

Documented uses

  • Compare exact-term keyword search with meaning-based dense semantic search.
  • Explore image-based retrieval when the relevant visual detail is absent from article text.
  • Study combinations such as filtering with visual reranking and coarse-to-fine retrieval.
  • Inspect retrieval results and scores across a small hand-built set of test questions.
Before you choose

Requirements & limits

The repository is public and MIT licensed. Running the notebook locally requires uv and service API keys; Claude is only needed for the agentic section and RAG answers.

  • The examples use a specialized corpus of bird articles and photos rather than a reader's own data.
  • The project separates image/full-text indexing from chunked text-dense indexing across two Pinecone indexes.
  • The first image-index build embeds about 2,000 photos and may take 15–30 minutes; embeddings are cached locally.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up and open the Bird Search notebook

Prepare the documented local environment and launch the Marimo notebook.

Before you start

  • uv, the Python runner used by the repository.
  • Pinecone and Google Gemini API keys for the core retrieval techniques.
  • An Anthropic Claude API key for the agentic section and RAG answers.

Steps

  1. Install uv if it is not already installed.

  2. Copy .env.example to .env and add the API keys for the services you plan to use.

  3. Run the command below to install dependencies, build missing Pinecone indexes, and open the notebook.

    Documented command
    make run

Expected outcome

The notebook opens with its retrieval examples and evaluation set. The first run builds the indexes; image embedding for about 2,000 photos may take 15–30 minutes and is cached locally.

Guide limitations

  • Claude is only required for the agentic section and RAG answers; keyword, dense semantic, and visual cross-modal retrieval use Pinecone and Gemini.
  • The repository documents that the first image-index build can take 15–30 minutes.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
pinecone-io/better-rag-retrieval-marimo · Official homepage title · retained in its source language
How we review sourcesSuggest a correction