GoodMem Documentation

GoodMem is the memory layer between your agents and their data. It turns documents, knowledge, and conversation history into persistent, searchable memory; manages embedders, rerankers, and LLMs as first-class resources you wire together at query time; and governs every request with a full authorization model.

Key Features

• Semantic Retrieval: Dense, sparse, and hybrid search with reranking and AI-powered summarization
• Multi-modal Memory: Text, image, audio, and video content
• Open Model Stack: First-class provider types for OpenAI, Alibaba Cloud Model Studio (DashScope), TEI, vLLM, llama.cpp, Cohere, Jina, and Voyage — hosted APIs or models you serve yourself
• Governed Access: Principals, roles, grants, and scoped API keys decide who can retrieve what, with auditable retrieval logging
• Flexible Chunking: Configurable text chunking strategies for optimal retrieval
• Real-time Streaming: Stream search results and AI-generated summaries
• Cloud Tuner (private beta): On GoodMem Cloud, the Retrieval Optimizer searches embedder and reranker combinations against your own corpus and queries, alongside embedder and reranker fine-tuning

OCR is available via the GoodMem OCR add-on service/image and is not included in the base install.

Quick Start

Before you use an SDK or the CLI, install and start a GoodMem server. The recommended installer is curl -s https://get.goodmem.ai | bash, which installs both the server and CLI, creates a local CLI profile, and leaves the default local setup ready to use. The CLI can self-upgrade with goodmem upgrade.

 

Core Concepts

Spaces

Spaces are the fundamental storage units in GoodMem. All memories belong to a space, and each space can be configured with specific embedders and chunking strategies.

Memories

Memories are your stored content with associated metadata. They are automatically chunked and embedded for efficient retrieval.

Embedders

Embedders convert your content into vector representations. GoodMem supports multiple embedding providers and models.

Retrieval

Semantic search across your memories with AI-powered summarization and context-aware results.

Principals and Access Control

Every request is authenticated as a principal — a human user or a service identity — and authorized through roles, grants, and API-key ceilings. Start with the Security Model.

Architecture

GoodMem provides:

• gRPC APIs generated from Protocol Buffers for type-safe, efficient communication
• REST APIs automatically generated from the gRPC definitions
• Client SDKs generated from OpenAPI specifications in multiple languages
• CLI tool for command-line operations and administration

Learn how the runtime publishes ports and files in Server runtime footprint.

Getting Started

1. Choose your interface: Start with our Client SDKs for programmatic access or the CLI for command-line usage
2. Set up authentication: Create API keys to authenticate your requests
3. Create a space: Set up your first memory space with embedder configuration
4. Store memories: Add your content as memories in the space
5. Retrieve and search: Use semantic search to find relevant information

Ready to dive in? Check out our SDK documentation or CLI reference to get started!