Semantic search with vector databases

Learn to create vector database indexes via code with Pinecone's api, use Python, and implement a two-layer api and service architecture for natural language vector searches.

Set up a Python semantic search project by creating a virtual environment, structuring API and services folders, and installing OpenAI and Pinecone libraries.

Learn to create and manage Pinecone APIs, set up environment variables with Python dotenv, and build authenticated Python functions for indexing and integration with OpenAI in semantic search workflows.

Learn to create a Pinecone index in Python by authenticating, then executing a create index call with a 1536-dimension, cosine metric, and a serverless configuration on AWS.

Learn to build a clean, professional api workflow with fastapi, where a user enters a bank name and the api creates an index via a service that communicates with pinecone.

Learn to list Pinecone indexes for semantic search with vector databases in Python by authenticating, calling the list index method, and exposing results via an api route for testing.

Display index details by creating and listing indices through two routes, authorize and authenticate users, and fetch index metadata via a two-layer API and service workflow.

Learn to manipulate a vector index in a vector database by turning natural language into 1536-dimensional embeddings and using upsert to insert or update vectors.

Learn to build an independent embed service that converts natural language text into OpenAI embeddings, handle token limits, authenticate, choose the right embed model, and extract 1536-dimension vectors.

Learn to convert pdf text to vectors for a vector database by extracting text, generating embeddings with OpenAI ada 002, and indexing via app search.

Split large documents into smaller pieces to fit Ada 002's 8191 token limit for embeddings in a vector database.

Explore advanced split techniques with overlap for semantic chunking in vector databases, using and comparing from-scratch approaches versus libraries like Lang Chain, with practical guidance on chunk size and overlap.

Implement a split text function to break text into overlapping chunks for embedding, specifying chunk size and overlap, returning a list; expose it as a miscellaneous API route.

Learn a purist, library-free embedding workflow for vector databases, converting PDFs to text, splitting into chunks, embedding with a custom service, and upserting vectors for app search.

Learn how to upsert data into a vector database using pinecone's app search, including generating vector embeddings, assigning vector ids, and optionally adding metadata to enable future filtering.

Create an app search workflow that extracts text from a PDF, breaks it into chunks, embeds them, and upserts into a Pinecone index with namespace and metadata.

Learn how to prepare and insert embeddings into a vector database, generate unique ids with uuid, and validate app search insertions using Pinecone, including error handling.

Explore adding metadata to Pinecone vectors, embedding text chunks, and performing similarity search with OpenAI embeddings, while handling client-side JSON metadata and preserving the original text.

Assemble and embed text chunks into vectors, attach metadata and ids, then upload the vector set to the service via the API.

Validate and test metadata in Pinecone by registering documents, debugging keys, and confirming saved entries, preparing for AI-powered document search.

Learn how to perform queries in a vector database by embedding questions with the Ada model, searching in pinecone, and using metadata chunks to retrieve vectorized results.

Build and test a simple vector database query using fast api, including authentication, embedding, namespace and vector inputs, and preview results with similarity scores and metadata.

Explore formatting responses to vector queries by converting results into a JSON response, using metadata for semantic search, and leveraging an AI assistant to generate answers.

Create an ai assistant that answers user questions by querying a vector database with embeddings, and pass metadata to an openai llm agent that returns a reply through context.