aidp-connectors-bootstrap

# `aidp-connectors-bootstrap` — first-time setup of the helper package in AIDP ## When to use - The user just installed the plugin and asks "how do I set this up?", "what's the first step?", "install the helpers". - The user runs a connector skill for the first time and gets `ModuleNotFoundError: No module named 'oracle_ai_data_platform_connectors'`. - The user explicitly asks to upload the helper package to AIDP. ## Outcome of running this skill - `/Workspace/Shared/oracle_ai_data_platform_connectors/scripts/oracle_ai_data_platform_connectors/` exists in the user's AIDP workspace, populated from the plugin's local `scripts/` directory. - The user has run `examples/00_bootstrap_helpers.ipynb` once and it printed `BOOTSTRAP OK`. - From that point on, every connector skill works without further setup. ## What the assistant should do ### Step 1 — locate the plugin's `scripts/` directory on disk The plugin lives wherever the user installed it. Common locations: - `~/.claude/plugins/<marketplace>/oracle-ai-data-platform-workbench-spark-connectors/scripts/oracle_ai_data_platform_connectors/` - `~/.codex/plugins/<marketplace>/oracle-ai-data-platform-workbench-spark-connectors/scripts/oracle_ai_data_platform_connectors/` - A local clone or repo checkout when developing. Search the installed plugin cache and the local repo checkout for `scripts/oracle_ai_data_platform_connectors/`. Confirm the source and destination paths with the user before uploading. ### Step 2 — create the destination directory in AIDP Use the AIDP MCP tools: ``` mcp__aidp__create_directory( workspace_id="<user's workspace id>", path="/Workspace/Shared/oracle_ai_data_platform_connectors" ) mcp__aidp__create_directory( workspace_id=..., path="/Workspace/Shared/oracle_ai_data_platform_connectors/scripts" ) ``` (If the workspace_id isn't already known from the conversation, ask the user.) ### Step 3 — upload the package files For each `.py` file under the local `scripts/oracle_ai_data_platform_connectors/`, upload to the matching path under `/Workspace/Shared/oracle_ai_data_platform_connectors/scripts/`. Use `mcp__aidp__upload_file` (or the equivalent in this MCP server). The package layout to preserve: ``` oracle_ai_data_platform_connectors/ ├── __init__.py ├── auth/{__init__,wallet,dbtoken,oci_config,user_principal,secrets}.py ├── jdbc/{__init__,oracle,hive}.py ├── rest/{__init__,fusion,epm,essbase}.py └── streaming/{__init__,kafka}.py ``` ### Step 4 — push the bootstrap notebook to AIDP and run it Upload `examples/00_bootstrap_helpers.ipynb` to `Shared/connectors-tests/00_bootstrap_helpers.ipynb` via `mcp__aidp__nb_save_file`. Then `mcp__aidp__nb_create_session` against the user's chosen cluster (typically `tpcds`), and `mcp__aidp__nb_execute_code` for each cell. The final cell prints `BOOTSTRAP OK` if everything works. ### Step 5 — confirm Tell the user: - Where the helpers landed. - That every connector skill in this plugin will now work. - The next step is to pick a connector (e.g., `aidp-atp`) and supply that connector's env vars / Vault secrets. ## Alternative: ask the user to install via PyPI (v1.0+) Once the package is published to PyPI, this skill should pivot to telling the user to run `%pip install oracle-ai-data-platform-connectors` in any AIDP cell instead of uploading. Until v1.0 ships, the Workspace-upload path above is the only way. ## What NOT to do - Do not upload anything to `/Workspace/Shared/` without confirming the path with the user (in case they have an existing convention). - Do not write secrets, .env contents, or PEM keys anywhere in `/Workspace/Shared/`. The helper package is code-only. - Do not skip the sanity-import notebook — it's how you (and the user) confirm the upload worked, not just that files exist. ## References - Bootstrap notebook: [`examples/00_bootstrap_helpers.ipynb`](../../examples/00_bootstrap_helpers.ipynb) - Plugin README install section: [`README.md`](../../README.md)