Practical AI Labs — Index

Practical AI Labs

Lab Build Main evidence
1 — Prompt and verification Repeatable research prompt Claim/citation audit
2 — Local RAG Small document assistant Retrieval and grounding scores
3 — Read-only tool agent Bounded tool workflow Authorization and trace record
4 — Production harness capstone Versioned system with gates Eval, observability, recovery, threat model

Existing implementation notebook set: Local RAG-CAS project.

Lab safety

Use synthetic/non-sensitive data, isolated environments, scoped credentials, no unapproved external effects, and explicit cleanup. A successful output without supporting evidence does not pass.