build-paths-advanced

What problem does it solve?

It scaffolds an advanced, production-feeling agent project where Oracle AI Database is the only durable state store, so you can demo and iterate on agentic workflows without sidecar databases or hidden state.

Core Features & Use Cases

• Oracle DB as the only state store: Enforces a forbidden-imports constraint and routes all durable memory and retrieval through Oracle components (OracleVS, in-DB ONNX embeddings, and OAMP where applicable).
• Skill-library composition: Orchestrates three reusable building blocks—Oracle DB docker setup, LangChain Oracle vector wiring, and an MCP server that exposes SQL/schema tools—then layers in only the application logic.
• Three advanced demo projects: Scaffolds (1) a hybrid NL2SQL + document RAG analyst with OAMP-backed durable per-user memory, (2) a self-improving research agent that logs tool runs and retrieves prior executions via OracleVS plus OAMP conversation context, or (3) a conversational schema designer with DDL confirmation gating and JSON Duality view generation (with read_write MCP).
• Verification + executed notebook payoff: Includes a verify.py checklist (DB/memory/MCP smoke tests and state-store invariants) and requires a clean, executed notebook for shareable demos.

Quick Start

Run the scaffold with a non-empty empty target directory, choose one of the three advanced project ideas (or provide a pitch mapped to them), confirm OCI GenAI usage, and generate the project with an executable notebook and verification script.