RAGo

A fast and extensible Retrieval-Augmented Generation (RAG) CLI, written in Go.

Overview

RAGo is a flexible RAG framework and CLI tool, designed for developers who want full control over:

• Models: Which LLMs are used to generate responses
• Embedders: How documents are embedded into vector space
• VectorStores: Where embeddings are stored and retrieved from
• Loaders: How information is fetched, parsed, and chunked for storage

All of these types are factory-based and configurable via a kubeconfig-like config file at ~/.rago/config.

Factories allow easy plugin registration and extension without the boilerplate. Current plugins primarily align with the interfaces from langchaingo.

Goals

RAGo is a pluggable RAG engine, configured by YAML, powered by Go.

• Fast, configurable RAG pipelines
• Easy plugin system for real-world applications with custom logic
• Support for any backend without lock-in
• Powered by Go for performance and portability

Current Plugins

Models

• ollama

Embedders

• ollama

VectorStores

• pgvector

Loaders

• markdown

Configuration

All configuration for all plugins is managed through a single YAML file located at ~/.rago/config.

A sample config looks like this:

models:
- name: mistral
  kind: ollama
  options:
    model: mistral
    serverURL: http://localhost:11434

vectorStores:
- name: pgvector
  kind: pgvector
  options:
    connectionURL: "postgres://testuser:testpass@localhost:5432/testdb"
    preDeleteCollection: true
    embedder:
      name: nomic-embed-text
      kind: ollama
      options:
        model: nomic-embed-text
        serverURL: http://localhost:11434

loaders:
- name: my-wiki-loader
  kind: markdown
  options:
    directory: ~/my-wiki

# Contexts are the glue that map all the pieces together, and allow
# you to easily mix/match models, stores, and loaders in quick, new
# configurations
contexts:
- name: wiki-rag
  model: mistral
  vectorStore: pgvector
  loader: my-wiki-loader

# The current-context is what you are currently set up to use
# on your next rago command.
current-context: wiki-rag

Usage

Basic commands:

# Add documents to your vectorstore (again, assuming your config choices!)
rago vectorstore add-documents

# More commands comming soon...

## Delete all documents from your vectorstore
#rago vectorstore delete
#
## Start chatting with your LLM, with retrieval augmentation
#rago chat

For Developers - Extensibility

Adding new models, loaders, embedders, or vectorstores is easy:

• Create a new plugin in /plugins
• Implement the proper interface based on the factory you want to Register to
• Register your plugin with the Factory via init() using the factory system, and add your plugin to the top-level register.go

No boilerplate, just what makes your RAG yours.

Acknowledgements

• github.com/tmc/langchaingo: for the foundational interfaces