Sign inSign up

umangsingh123/ragyn-base

By umangsingh123

•Updated 4 months ago

Ragyn base image — shared Python deps and config for the graph and vector RAG platform

Image
Developer tools
0

893

umangsingh123/ragyn-base repository overview

⁠ragyn-base

Shared base execution layer and dependency framework for the Ragyn⁠ graph RAG platform.

⁠What is this?

ragyn-base is the parent image for all Ragyn container services. It compiles and packages core OS libraries, Python 3.12 virtualenvs, database driver binaries (Neo4j, PostgreSQL), and the complete set of enterprise connector libraries and document parsers.

Note: This image serves as a builder parent and is not meant to be run directly as a container. ragyn-search and ragyn-ingest are built on top of this image.


⁠Bundled Enterprise Features & Libraries

To support full-scale enterprise ingestion and parsing, this base layer pre-compiles and wraps the following SDKs and libraries:

⁠1. Document Parsers & OCR
  • PDF Parsing: PyMuPDF (fitz) for fast, local layout and text extraction.
  • Word Processing: python-docx for parsing Microsoft Word document structures.
  • Markdown Parsing: markdown-it-py for structural parsing of markdown content.
  • Image Processing & OCR: Pillow (PIL) for image processing and layout parsing. Supports optional OCR layer processing for scanned PDFs and images (PNG, JPG, JPEG).
⁠2. Cloud Storage Connectors (Sources)
  • AWS S3 & S3-Compatible: s3fs and boto3 SDKs.
  • Azure Blob Storage: adlfs and Azure Storage SDKs.
  • Google Cloud Storage (GCS): gcsfs and Google Cloud Storage SDKs.
  • Microsoft SharePoint: Office365-REST-Python-Client for connecting to secure SharePoint Online Document Libraries.
⁠3. Enterprise Database Drivers
  • PostgreSQL: psycopg[binary,pool] and asyncpg for high-performance async database connections.
  • Neo4j Graph Database: neo4j official driver with connection pooling.

⁠Features & Hardening Standards

  • Multi-Architecture: Built for both linux/amd64 and linux/arm64 (AWS Graviton) CPU architectures.
  • Embedded Security: Compiled using buildx with embedded SLSA Provenance and SBOM (Software Bill of Materials) attestations.
  • Non-Root Execution: Pre-configured with a secure non-root appuser (UID 1000) shell-disabled account.
  • Intellectual Property Protection: Python source files are pre-compiled into .pyc bytecode, and the raw .py source code files are deleted to prevent reverse engineering.
  • Optimized Size: Built using slim Debian OS bases with minimal APT packages to ensure high security scores on cloud scanners (Prisma, Wiz, Trivy).

⁠Tags

TagDescription
latestMost recent stable base layer
1.0.0-alphaBase layer for the 1.0.0-alpha release

ImagePurpose
ragyn-search⁠Query and retrieval service
ragyn-ingest⁠Document ingestion pipeline

⁠License

Proprietary — © Ragyn. All rights reserved.

Tag summary

Content type

Image

Digest

sha256:c6dcc300b…

Size

380.8 MB

Last updated

4 months ago

docker pull umangsingh123/ragyn-base