Overview

Introduction

DreamLake Server is the central API that manages the entire DreamLake ecosystem. It handles namespaces, spaces, episodes, assets, folder hierarchy, and semantic search.

Core Concepts

Concept	Description
Namespace	Top-level owner scope (user or team). Maps to a URL slug like `alice`.
Space	A project or workspace within a namespace. Contains episodes and assets.
Episode	A time-bounded recording unit — e.g. a robot run, an experiment session.
Node	A folder in the hierarchical file tree (materialized paths).
Asset	A file (video, audio, label track, text track) stored in BSS and indexed here.

Architecture

CLI (dreamlake-py)
    │
    ├── Upload file ──→ BSS (S3 storage)
    │                     │
    └── Register asset ──→ DreamLake Server (MongoDB index)
                            │
                            └──→ DreamLake Frontend (browse + search)

DreamLake Server is the metadata layer. All binary data lives in BSS / S3. The server provides:

Asset registration — link uploaded files to spaces and episodes
Node hierarchy — folder tree with browse, rename, move, delete
Episode management — create, update, organize recording sessions
Search — vector-based semantic search via Qdrant
Logs & parameters — structured logging and experiment config storage

Tech Stack

Component	Technology
Framework	Fastify
Database	MongoDB (via Prisma)
Auth	JWT (vuer-auth integration)
Search	Qdrant (vector embeddings)
Storage	Delegates to BSS (S3)

Quick Start

# Start dependencies
cd docker && docker compose up -d
 
# Start the server
cd dreamlake-server && pnpm dev
# → http://localhost:10334