Current Project Phase: Pre-Alpha
See Getting Started for development instructions.
This project is fairly new (though, I've been conceptualizing it in my head for quite some time), and it seems like a perfect opportunity to start developing it in the open.
Bosca is an Open Source (Apache 2.0) AI powered content management system.
Key Features (not an exhaustive list):
- Content Management
- Collections
- Semantic Search
- Accounts, Authentication, Authorization, etc.
- Content Workflows
- Content Processing
- Videos
- Images
- Text
- Content Graph
- Personalization
- Profiles
- Recommendations
- Client SDKs (GraphQL)
Roadmap:
- Stabilize Service Interfaces
- Build Administration Interfaces
- Stabilize Kubernetes Environment
- ...
This system leverages a monorepo containing all the components for Bosca. However, the build pipeline will deploy Bosca in a microservice style architecture. There are some exceptions to this rule, such as components that aren't part of the Bosca Core. For example, administrative interfaces that will be part of a commercial offering (or, at least that is the current thinking, open to other thoughts about how to make Bosca sustainable).
- /docs - Documentation for Bosca
- /backend - Contains backend API components
- /protobuf - Contains protobuf definitions for the API and Clients
- /database - Contains various database scripts and definitions
- /services - Contains a development environment for the services Bosca leverages and integrates
- /.run - Contains scripts for running Bosca Commands in IntelliJ (using the GoLand plugin)
Technologies:
These are the general directions for the technology stack of Bosca. There will be some variations in a production environment, such as possibly using Vertex AI over Ollama for processing different LLM related functions. But to allow for local only processing, something like Ollama will be integrated into the system.
To enable these types of variations, there will be modular microservices and components that can be picked from. For instance, you may be running in Google Cloud and want to use GCP cloud buckets for storage of assets. The default implementation (for portability) will be MinIO, but with minimal effort, you can leverage a different storage provider. In the interest of being fully functional as quickly as possible, we will forgo certain implementations until a later times. But we will try and make decisions that support the modularity needed to make Bosca flexible, such that it can fit into any organization's architecture well.
- Go
- GRPC
- Temporal
- Ory
- NextJS
- PostgreSQL
- Kubernetes
- Qdrant
- LangChain
- TypeScript
- Ollama
- MinIO
- GraphQL
- Ray
For more information, you can look in the /docs folder
Architecture:
