aidp-models-catalog

aidp-models-catalog — available models & parameters

Two distinct surfaces:

• Platform model catalog — what you can call from ai_generate() / agent flows (LA Models API).
• MLOps/MLflow registry — registered models + versions (the aidp-mlops group; see also aidp-mlops).

CLI (preferred) — registry browse: aidp mlops <command> --instance-id <DATALAKE_OCID> --auth api_key --profile DEFAULT --region <r>

• aidp mlops list-registered-models | get-registered-model | list-model-versions
• (The official CLI folds the model registry into the mlops group — there is no separate models group.)

Fallback (no CLI) / platform catalog: the LA Models REST API via oci raw-request (identical endpoint + auth) — this is the verified path for resolving an ai_generate() model name.

Verify-first (no-fabrication): the platform Models API is Limited Availability on 20240831; the ?modelType= list call is LIVE-VERIFIED 200 below. The MLOps registry side is Preview (not yet probed). Confirm a name with a live read before relying on it; record in references/rest-endpoint-map.md.

When to use

• "What models are available?", "what's the model name for ai_generate?", "show model parameters", "list registered models / model versions".

Platform catalog — endpoints (DataLake-scoped; LA 20240831)

Base: https://aidp.<region>.oci.oraclecloud.com/20240831/dataLakes/<DATALAKE_OCID>/…

• GET /models?modelType=<GENERATIVE_AI|BASE|EMBEDDING|LLM> — list installed models (LIVE-VERIFIED 200)
• GET /models/{id} — model detail
• GET /models/{id}/modelParameters — parameters

Workflow

1. Platform catalog: GET /models?modelType=GENERATIVE_AI (auth ladder; on 401/403 → session refresh) → present available model names/ids. Repeat with BASE / EMBEDDING / LLM to enumerate other catalogs. GET /models/{id} / …/modelParameters for detail.
2. Registry browse: aidp mlops list-registered-models → list-model-versions (CLI preferred).
3. Feed the confirmed model name to aidp-ai-sql (ai_generate('<model>', …)) or aidp-agent-flows.

oci raw-request --http-method GET \
  --target-uri "https://aidp.us-ashburn-1.oci.oraclecloud.com/20240831/dataLakes/<DATALAKE_OCID>/models?modelType=GENERATIVE_AI" \
  --profile DEFAULT

Notes

• Use this to resolve a real model name before ai_generate rather than guessing one.
• Read-only; no destructive ops.
• A bare GET /models may 400 on a required param — pass modelType (verified above) to enumerate.
• ai_generate availability is independent of this catalog. GET /models?modelType=GENERATIVE_AI can return items: [] on a fresh instance while ai_generate('openai.gpt-5.4', …) still executes fine — the model is resolved at the Spark engine level, not via this REST catalog. An empty list here does not mean ai_generate is unavailable. Confirm AI-in-SQL with the smoke test in aidp-ai-sql (a trivial SELECT ai_generate('<model>', 'hello') cell), not with this endpoint.

References

• references/aidp-cli-map.md · references/oci-raw-request.md · references/no-mcp-rest-map.md · references/rest-endpoint-map.md · pairs with aidp-ai-sql, aidp-agent-flows, aidp-mlops