Merge pull request #160 from oracle-devrel/feature/agent_harness_update

notebooks: add agent_harness with enterprise_data_agent_heavyweight

Author: WSPluta

notebooks/agent_harness/README.md

@@ -0,0 +1,27 @@
+# Agent Harness
+
+End-to-end notebook that constructs a production-grade agent _harness_ on Oracle AI Database 26ai. An agent harness is everything around the model — memory, retrieval, tool dispatch, identity, budgets — built here from Oracle primitives (OAMP, in-DB ONNX embeddings, HNSW + Oracle Text, MLE, DBFS, Duality Views, DDS) running on a single local container.
+
+## Contents
+
+| Content                                                                              | Description                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              | Open                                                                                                                                                                                                                                   |
+| ------------------------------------------------------------------------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| [`enterprise_data_agent_heavyweight.ipynb`](enterprise_data_agent_heavyweight.ipynb) | Full build on a local Oracle AI Database 26ai container. Walks through DSN + dedicated `AGENT` user, in-database ONNX embeddings + HNSW indexes, cross-encoder reranking, hybrid vector + Oracle Text retrieval fused via RRF, vector-indexed `toolbox` and `skillbox` registries, a DBFS scratchpad with a `UTL_TO_TEXT` / `UTL_TO_CHUNKS` promotion path to OAMP, sandboxed JavaScript via Oracle MLE, the agent loop, JSON Relational Duality Views, `DBMS_SCHEDULER`-driven re-scans, and identity-aware authorization via Deep Data Security (DDS). | [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/oracle-devrel/oracle-ai-developer-hub/blob/main/notebooks/agent_harness/enterprise_data_agent_heavyweight.ipynb) |
+
+## Layout
+
+```
+agent_harness/
+├── README.md
+├── enterprise_data_agent_heavyweight.ipynb
+└── images/        # diagrams referenced inline by the notebook
+    ├── cover-oracle-native-arch.png
+    ├── cover-oracle-reference-arch.png
+    ├── cover-duality-view.png
+    ├── cover-skillbox-flow.png
+    ├── cover-toolbox-flow.png
+    ├── oamp_memory_discpline.png
+    ├── oamp_breakdown.png
+    ├── dual_memory_substrate.png
+    └── ingestion-pipeline.png
+```