Using the CLI

The package includes a Command-Line Interface (CLI) designed primarily for debugging, quick testing, and simple integrations. It allows you to explore functionality, manage the vector database, and even integrate insightvault into a CI/CD pipeline.

Note: The CLI is optimized for simplicity, not performance. It is slower by design compared to using the Python package directly, as it initializes components on every command.

To get started, display the available commands and their descriptions by running:

insightvault --help

Command: manage

The manage command allows you to interact with the vector database used by the SearchApp and ChatApp. You can add, view, or clear documents stored in the database.

Adding Documents

You can add documents in two ways:

1. From a text file:

insightvault manage add-file <path-to-file>

This splits the input file into chunks, generates embeddings for each chunk, and ingests them into the database.

2. From raw text:

insightvault manage add-text "<text>"

This creates and ingests a document directly from the provided string.

Listing Documents

To view the documents stored in the database:

insightvault manage list

Clearing the Database

To delete all documents and reset the database:

insightvault manage delete-all

Command: summarize

The summarize command uses the SummarizerApp to generate concise summaries of text. This command does not use the vector database.

Summarizing Raw Text

Provide a string to summarize directly:

insightvault summarize "This is a very long string which we must summarize."

Summarizing a Text File

Provide the path to a text file for summarization:

insightvault summarize --file "./data/my-file.txt"

Command: search

The search command performs semantic searches on the document chunks in the vector database.

Example Search

Run a query to retrieve the most relevant documents:

insightvault search "Why is the sky blue?"

The results include the names of the best-matching documents, with the maximum number of results configurable in your config.yaml.

Command: chat

The chat command provides an interactive session for chatting with the documents stored in the vector database. It leverages Retrieval-Augmented Generation (RAG) to generate context-aware responses.

Starting a Chat

Ask a question and receive a response based on your documents:

insightvault chat "What is Rayleigh scattering?"

Important: The CLI chat does not preserve history between commands. For persistent conversations and enhanced features, consider building custom apps as described in Building Apps.

Notes

1. Performance: The CLI initializes components on each invocation, making it slower than direct programmatic use. For production-grade performance, use the Python package.

2. Shared Database: The manage, search, and chat commands operate on the same shared database, which is accessible across multiple CLI sessions.

3. Debugging and Prototyping: The CLI is ideal for debugging and quickly testing new ideas but is not recommended for latency-critical or high-volume use cases.