Example applications and notebooks

Choose an item by what you want to build or learn. Sample applications show an end-to-end user experience. Notebooks focus on one workflow or feature at a time.

Follow the README in each sample application directory for prerequisites, installation, configuration, and run instructions. For notebooks, open the notebook in Jupyter or Colab and follow its setup instructions.

Sample applications

Doc Chatbot: RAG-based document question answering

Build a retrieval-augmented generation (RAG) application that accepts PDF or TXT documents and generates a grounded answer from the retrieved content.

Covers: Document upload, text chunking, embeddings, metadata, vector retrieval, and answer generation.

Open the Doc Chatbot sample  

Semantic Code Search

Search a local Python codebase with natural-language queries. The sample parses code into functions, classes, and other snippets, then returns matching source through a FastAPI backend and React frontend.

Covers: Code parsing, code embeddings, file and line-range metadata, semantic retrieval, and a web search interface.

Open the Semantic Code Search sample  

Fashion Product Search

Search fashion products with a text description or an uploaded image. The sample stores image and text embeddings in separate Oracle VecDB vector tables and returns filtered product results.

Covers: Text and image embeddings, multiple vector tables, shared product metadata, filtering, and multimodal search.

Open the Fashion Product Search sample  

Semantic Product Search and Visual Discovery

Combine text and image similarity search over a product catalog. The sample generates CLIP embeddings and combines top-k vector search with category and price filters.

Covers: CLIP embeddings, text and image search, top-k retrieval, category filters, price filters, and product discovery.

Open the Semantic Product Search and Visual Discovery sample  

Ask the Parks: semantic and geospatial search

Search National Park Service data by meaning, with optional metadata and location filters. A natural-language query returns ranked parks, and a map location adds a geospatial filter.

Covers: Semantic search, metadata filters, geospatial filters, GeoJSON locations, ranked results, and a map-based interface.

Open the Ask the Parks sample  

Browse the complete sample application catalog.

Notebooks

Use the notebooks for focused, runnable examples. They are grouped by the developer task they demonstrate, with a short description of what each notebook covers.

Start here

Notebook	Covers
Quickstart	The first Python vector-search workflow.
Connect	Configure and reuse a Python client.

Embeddings and models

Notebook	Covers
Embedding & Ingest Patterns	Compare integrated embeddings and BYO vectors.
OCI Embedding Pipeline	Generate application-supplied embeddings with OCI Generative AI.
Finance Accounts BYO	Generate and query vectors for finance account data.
In-database Reranking	Load and use a database-loaded reranking model.
OCI-hosted Reranking	Rerank candidates with an external provider.

Ingest and search data

Notebook	Covers
Integrated Embedding bulk load	Bulk load records with integrated embeddings.
BYO vectors with assigned IDs	Bulk load BYO vectors with application-assigned IDs.
BYO vectors with generated IDs	Bulk load BYO vectors with database-generated IDs.
Search Diagnostics	Diagnose metadata filters and search responses.
Parks Demo	Combine semantic, location, and metadata filters.
Ask the Parks	Search park data with vector, metadata, and geospatial filters.
Gemini Vector RAG	Build a grounded RAG workflow with Gemini.

Indexes and maintenance

Notebook	Covers
HNSW Index Tuning	Tune HNSW search with efsearch.
Index & Model Workbench	Explore index lifecycle operations and model inspection.
Maintenance Indexing Playbook	Review table maintenance, deletion, reindexing, and cleanup.

Browse the complete notebook catalog.