[RichmondAlake]

WMS ID

Pending — not supplied yet. This PR is opened as a draft so the WMS ID, WMS self-QA and hosted workshop delivery can be confirmed before it is made ready for review.

Summary

Add Agent Memory: Introduction to context and memory engineering in dev-rel-con-mem-engineering, following the same 10-file structure as dev-rel-agent-memory and dev-rel-rag-to-agents.

• Introduce the Everyday Goods inventory handover: 80 mugs proposed, 120 approved under PO-1048, and no supplier order submitted. Explain the manual workflow, current friction and why the assistant needs both context and durable memory.
• Guide learners through a bare LLM, a transient tool-using loop and a memory aware agent using Oracle Agent Memory 26.8, Oracle AI Database, LangChain and Ollama serving llama3.2. Cover working, episodic, semantic and procedural memory, extraction, refinement, offloading, compaction, retrieval, context budgets and action boundaries.
• Provide LiveLabs connection guidance, a 17-TODO learning checklist, five supporting images, an 18-question quiz and the standard sandbox manifest/landing page.
• Incorporate the latest notebook corrections: per-ingestion extraction checkpoints, fresh capture IDs and a precise explanation of the context-card retrieval budget.
• Keep the reference-folder pattern: no extra README, AGENTS.md, notebook copies, helper scripts or synchronization files.

Companion notebooks and delivery dependency

The notebooks and runtime assets are maintained separately in NoSQLKnowHow/LiveLabs-Image-Dev PR #2, at commit 30c5de976baa4fda6e6985fa164d7f411631db38 when this PR was prepared. This PR adds only the LiveLabs workshop pages and assets, not notebook or image provisioning code.

The companion complete notebook ran 80/80 code cells with Llama 3.2 on 9 October 2026 (run 3738399430), with actual outputs saved in place. The memory-aware agent recalled 120 approved under PO-1048, with no order submitted, matching the previous Qwen run's core answer. Overall quality was not equivalent: the no-memory agent invented details, a handled tool call failed argument validation, extraction omitted some facts, and the full handover altered an evidence ID. The connection page explicitly asks learners to inspect these differences. Its optional approved-write path was not exercised. This is local notebook evidence, not a hosted LiveLabs reservation test or a model-reliability benchmark.

Before launch, the workshop runtime must supply notebooks/agent_memory_zero_to_hero, both notebook editions, their fixture/diagrams, dependencies, a dedicated Oracle account, the imported embedding model and the configured Ollama service. Confirm the companion changes are included in the selected lab image.

Validation

• Repository LiveLabs Markdown validator passed for all three pages.
• The folder's exact 10-file set matches both reference workshops.
• Manifest JSON and all local page/image references resolve; the manifest includes the help email.
• All five PNGs were visually inspected and meet the repository's 1,280-pixel maximum in either dimension.
• All 18 quiz questions have exactly one marked answer; the learner checklist contains TODOs 1–17.
• All paths are lowercase; no file exceeds 100 MB; credential-pattern checks found no secrets in the text files.
• git diff --cached --check passed. Only this workshop folder is included.

Checklist

• WMS ID supplied and WMS Self-QA completed.
• Workshop structure follows the existing reference workshops.
• Filenames and folders are lowercase and descriptive.
• Markdown images have alternate text.
• Manifest contains the help email address.
• No install packages, compiled binaries or files over 100 MB; the only binary assets are the five workshop PNGs.
• Hosted reservation/landing-page end-to-end QA and WMS URLs confirmed before publication.
• Contributor/OCA eligibility verified through the repository's normal contribution process; no agreement acceptance is asserted here.

Review and publishing notes

Please follow the repository's review SLAs. Update the WMS URLs after approval as required by the PR template.

The workshop deliberately distinguishes application scope labels from authorization, extraction from verified truth, and live-row cleanup from verified erasure. The assistant does not place supplier orders or change stock.

[tnechita]

Hello @RichmondAlake

Since the Oracle AI World LiveLabs Code Freeze is currently active, only emergency PRs will be approved.

To request an emergency PR, please send an email to livelabs-help-db_us@oracle.com with the following details:

• Session ID* (Must be an AI World-related HOL)
• WMS ID
• Justification for the PR

Note: All non-AI World related PRs will be approved only after October 28th.

Thank you!

[hope-fisher]

Per Richmond, this draft is to be closed.