langchain-oracledb-helper

What problem does it solve?

It solves the friction of wiring LangChain to Oracle Database for vector retrieval and chat history persistence, especially when metadata handling and chat table semantics don’t match what langchain-oracledb provides out of the box.

Core Features & Use Cases

• Multi-collection Oracle vector store scaffold: Creates a reusable, multi-collection wrapper so each logical collection kind (e.g., DOCUMENTS, CONVERSATIONS) maps to an Oracle table with a consistent naming convention.
• Embedder setup with guaranteed dimensionality: Supports multiple embedder options and enforces an expected embedding dimension to prevent silent retrieval failures from dim mismatches.
• Oracle chat history backend support: Generates an OracleChatHistory layer (plus the required idempotent DDL migration) for projects that need RunnableWithMessageHistory integration.
• Metadata-as-string monkeypatch: Provides a required import-time fix that makes similarity_search metadata parsing consistent for OracleVS-backed retrieval.

Quick Start

Use the langchain-oracledb-helper skill to scaffold a new vector store and optional chat history layer in your project, then run its provided bootstrap to verify embedding dimension correctness.