aidp-fusion-config

# `aidp-fusion-config` — fill `aidp.config.yaml` from names, not OCIDs ## Bundled CLI Resolve `<plugin-root>` as the directory two levels above this `SKILL.md`. On macOS/Linux, invoke `"<plugin-root>/bin/aidp-fusion-autopilot"`; on Windows, invoke `"<plugin-root>\\bin\\aidp-fusion-autopilot.cmd"`. Command examples below use `aidp-fusion-autopilot` as shorthand only—do not assume the plugin's `bin/` directory is on `PATH`. The worst part of fusion-autopilot setup is `aidp.config.yaml`: it wants three opaque identifiers — `aiDataPlatformId` (a long OCID), `workspaceKey`, and `clusterKey` (UUIDs). Customers should not hunt those out of console URLs and REST responses by hand. This skill collects **human-friendly names** and resolves the keys live. This is a **control-plane** skill — no MCP, no notebook session required. ## When to use - "Set up aidp.config.yaml", "configure the fusion autopilot connection", "what do I put for workspaceKey/clusterKey", "I don't have the OCIDs", "fill in the AIDP coords by name". - The user ran `init`, sees `aidp.config.yaml` full of `*-PLACEHOLDER` values, and asks what goes there. - A precondition check (e.g. the `aidp-fusion-seed` ladder) reports the config is missing or has placeholders. ## When NOT to use - Authoring `bundle.yaml`'s `fusion:` connectivity / credentials — customer-supplied policy the CLI can't discover (use `init` + hand-edit). This skill only fills the **coordinates** in `aidp.config.yaml`. - Resolving tenant data variation (column aliases, semantic variants) — that's `bootstrap` / `medallion-author`. ## Engine — the self-contained `aidp-fusion-autopilot init-config` command The skill shells out to one CLI command; it never re-implements OCI signing or REST calls. The command reuses the plugin's own `AidpRestClient` discovery primitives (`find_workspace_by_name` / `find_cluster_by_name`, OCI-signed) to turn names into keys, validates against the `AidpConfig` schema, then writes the env block. | Op | What you give | What it resolves / writes | |---|---|---| | Resolve workspace | `--workspace "<display name>"` | → `workspaceKey` | | Resolve cluster | `--cluster "<display name>"` | → `clusterKey` + live `state` (warns if not ACTIVE) | | Anchor | `--aidp-id <DATALAKE_OCID>` | the one root id (copied once from the console URL) | | Target env | top-level `--env <name>` (default `dev`) | writes `environments.<name>`, preserves siblings | | Preview | `--dry-run` | prints resolved keys + YAML, writes nothing | ```bash # 1. Dry-run first — resolve names → keys and show the YAML before writing anything aidp-fusion-autopilot init-config \ --aidp-id ocid1.datalake.oc1.iad.<...> \ --workspace "My Workspace" \ --cluster "My Cluster" \ --region us-ashburn-1 \ --dry-run # 2. Write it for real (drop --dry-run). Target a non-dev env with the top-level --env flag: aidp-fusion-autopilot --env prod init-config \ --aidp-id ocid1.datalake.oc1.iad.<...> \ --workspace "My Workspace" --cluster "My Cluster" # 3. Overwrite an env that already exists: aidp-fusion-autopilot init-config --aidp-id <...> --workspace "..." --cluster "..." --force ``` All calls sign with the local OCI profile (`--oci-profile`, default `DEFAULT`). For a session-token profile, have the user run `oci session authenticate` first. ## The one id the user must provide Two of the three coordinates are name-resolved; the **root anchor** is not: | Input | How the user gets it | Required? | |---|---|---| | **AIDP / DataLake OCID** | Copy once from the AIDP console URL — the `.../aiDataPlatforms/<OCID>/...` segment. | ✅ yes | | **Workspace name** | The display name in the AIDP console. | ✅ yes | | **Cluster name** | The Spark cluster's display name. | ✅ yes | | Region / OCI profile | Defaults `us-ashburn-1` / `DEFAULT`. | default | | Catalog | Reminder only — lives in `bundle.yaml` under `aidp.catalog`. | optional | ## Patterns - **Dry-run, then write.** Always run `--dry-run` first so the user confirms the resolved keys and YAML. - **Name miss → list + pick.** If a workspace/cluster name isn't found, the command prints the **available** names. Relay that list and ask the user to choose the exact one, then re-run. - **Inactive cluster.** If the cluster resolves but isn't `ACTIVE`, the config still writes — surface the warning and tell the user to start it before `run` (see `aidp-cluster-ops` / `bootstrap`). - **Multi-env.** Use `--env <name>` to add `prod` alongside `dev`; sibling environments are preserved. `--force` is only needed to overwrite an env that already exists. - **Hand off.** After writing, point the user at the printed next steps: `aidp-fusion-autopilot validate` → `aidp-fusion-autopilot bootstrap`. Remind them the **catalog** goes in `bundle.yaml` (`aidp.catalog`). ## Notes - **Comments aren't preserved.** Writing re-emits the YAML, so hand-written comments in an existing `aidp.config.yaml` are dropped (resolved values are kept). Mention this if the file is heavily commented. - **Skill family.** Sibling of `aidp-fusion-seed`; both wrap the self-contained CLI rather than importing the orchestrator. The seed skill's precondition ladder can invoke this skill when the config is the missing rung.