Pinecone Examples

This repository provides a collection of sample applications and Jupyter Notebooks designed for hands-on experience with Pinecone vector databases and common AI patterns, tools, and algorithms. It is divided into two main categories:

1. Production-ready examples: Located in ./docs, these are regularly reviewed and supported by the Pinecone engineering team.
2. Learning and exploration examples: Located in ./learn, these are maintained by the Pinecone Developer Advocacy team and are optimized for exploring AI techniques and application patterns.

The Ask Lex plugin is a demonstration ChatGPT plugin that illustrates how to apply Retrieval Augmented Generation (RAG) to ChatGPT using video content.

For a complete walkthrough of the build process, refer to the ChatGPT plugins tutorial.

This directory contains notebooks and scripts demonstrating how to build a GIF Search application using semantic search. The demo covers data preparation, indexing, and querying workflows.

Key components include:

• gif-search.ipynb: A notebook demonstrating the end-to-end pipeline: data preparation, indexing vectors into Pinecone, and querying the index.
• app.py: A Streamlit script that powers the interactive user interface for the GIF Search App.

Use the 00-data-build.ipynb notebook to replicate the data preparation process for the YouTube Search App. This notebook demonstrates how to use the Kaggle YTTTS dataset along with Beautiful Soup and other libraries to scrape and structure the data used in the application.

https://colab.research.google.com/github/pinecone-io/examples/blob/master/search/semantic-search/yt-search/00-data-build.ipynb

The Generative Pseudo-Labeling (GPL) implementation is provided as a series of Jupyter notebooks that follow the workflow described in the Generative Pseudo-Labeling (GPL) article. The workflow consists of five distinct stages: downloading the dataset, synthetic query generation, negative mining, pseudo-labeling scoring, and bi-encoder fine-tuning.

Each notebook can be run locally or via Google Colab.

You can deploy the Prometheus and Grafana monitoring stack using Docker Compose.

1. Start the services:

docker-compose up

2. Verify the running services and check the assigned ports for Grafana and Prometheus:

docker-compose ps

3. Access Grafana in your browser and navigate to the Dashboards section. The preloaded dashboard is located in a folder titled Pinecone.

The Shields Checker script validates Colab and nbviewer shields at the beginning of notebooks in the examples repository. It ensures links are valid and can optionally update them or raise errors if shields are missing.

To run the script, navigate to the root directory of the examples repository and execute the command below. By default, this command runs the check across all directories, does not update links, and does not raise an error if shields are missing.

python scripts/shields-checker.py run --path . --shield-error False --update False

It is recommended to use a dedicated Python environment. You can create one using Conda with Python 3.12:

conda create -n ollama-langgraph python=3.12
conda activate ollama-langgraph

This directory contains Jupyter notebooks related to the BERTopic article. Specifically, Script 06 provides a complete walkthrough demonstrating how to implement BERTopic using custom components, including:

• Custom transformers
• UMAP (Uniform Manifold Approximation and Projection)
• HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise)
• Count Vectorizer

Install the necessary dependencies by first installing poetry via pip, then running poetry install within the project directory.

pip install -qU poetry
poetry install