Skip to content

datarobot-oss/datarobot-agent-skills / skills/datarobot-agent-assist/agent-assist-deploy-workload

datarobot-agent-assist-deploy-workload

Flagged

Use when an agent assist user wants to deploy their agent, put it on DataRobot, get a link to it, or share it with others, the agent can run as a single container, and the project has no `infra/Pulumi.yaml`. Not for agents built from the agent template, which deploy with Pulumi through datarobot-agent-assist. Users usually just say "deploy" or "deploy to DataRobot" and won't mention containers, workloads or the CLI.

FileKindSize
SKILL.mdskill md10.5 KB
references/app-contract.mdmarkdown3.2 KB
references/deploy-fallback.mdmarkdown7.2 KB
references/deploy-path.mdmarkdown4.4 KB
references/tool-secrets.mdmarkdown1.3 KB

references/deploy-path.md

View on GitHub
# Choose the deploy path

Agent assist can deploy an agent two ways:

- **Workload API.** For a project with no `infra/Pulumi.yaml`. The agent, its tools, its API and its UI run in one container, and `dr workload up` builds and deploys it. Follow [../SKILL.md](../SKILL.md).
- **Custom application with Pulumi.** The agent template's own deploy. The agent, MCP server and web app are deployed separately. Follow [pre-deployment-checklist.md](../../agent-assist-build/references/pre-deployment-checklist.md).

The Workload API is for single Python projects that don't use a Pulumi template. Read `agent_spec.yaml` when present, otherwise the legacy `agent_spec.md`, along with the code in `<target_dir>`, and make the call yourself. Don't ask the user to choose between architectures, and don't ask them the questions below. Run the checks yourself.

## First, check for a Pulumi template

If `<target_dir>/infra/Pulumi.yaml` exists, use Pulumi and stop here. The project is the agent template or another Pulumi project, even when it would fit in one container. If it's already deployed with Pulumi, keep it there, or the user ends up with two copies. Switch only if the user asks, and tell them the old deployment keeps running until they remove it.

## Then, check the Workload API is on for this account

The Workload API is switched on per account or organization. Check it before the other checks. The check needs the CLI installed and signed in. If `dr` is missing or `dr auth check` fails, do steps 1 to 3 of section 2 in [../SKILL.md](../SKILL.md) first.

```bash
eval "$(dr auth export)"   # never print the token
curl -sS -H "Authorization: Bearer $DATAROBOT_API_TOKEN" "$DATAROBOT_ENDPOINT/workloads/?limit=1"
```

A 403 that says `Workload API is disabled in feature flags` means it's off. That's an account setting, not something the fallback can fix. Use Pulumi and tell the user: "The Workload API isn't turned on for your DataRobot account, so I'll use the standard deployment path. If you want the faster path, ask your DataRobot admin to enable the Workload API for your organization." A 200 means it's on.

## Use Pulumi if any of these are true

- The code doesn't fit in one container, or a Workload deploy already failed for that reason: a dependency outside `<target_dir>` (a `file://` path, a symlink, or `Distribution not found at: file:///...` in the build log), or an app that needs more than one process or port. Don't try the Workload API again for this project.
- A tool needs a GPU, a lot of memory, or moves a lot of data (tens of GB). That tool should be its own deployment so it can scale on its own.
- Some tools call complicated third-party systems that the user wants to deploy, debug and version separately from the agent.
- The tools have to be published as a standalone MCP server that other agents or teams use on their own schedule. One agent serving chat, A2A, MCP and a web UI from the same process still fits in a single container.
- The user needs features that today only come with a custom model deployment: guard models, moderation, custom metrics, registry approvals.
- The agent has more than about 10 tools, or several interfaces owned by different people.
- The spec uses `llm_deployment_id`. The Workload API path hasn't been tried with it.

## Otherwise use the Workload API

This is the default. Most agents fit. It's easier to debug, deploys in a couple of minutes, and can still be scaled up with more CPU, memory or replicas. If the code later turns out not to fit in one container ([app-contract.md](app-contract.md)), come back here and use Pulumi.

## Unclear or overridden

If a condition is really unclear, ask one plain question about what the agent needs, for example: "Will other teams need to use these tools on their own, without your agent?"

If the user asks for a path ("deploy with the Workload API", "simple agent assist", "use Pulumi", "the standard way") or disagrees with the call, go with the user.

## Tell the user and hand off

Tell them in one plain sentence, for example: "Your agent is simple enough to run as one app, so I'll deploy it with the Workload API," or "Your agent needs the bigger setup, so I'll use the standard deployment path."

- Workload API: continue with section 2 of [../SKILL.md](../SKILL.md).
- Pulumi: read and follow [pre-deployment-checklist.md](../../agent-assist-build/references/pre-deployment-checklist.md) end to end. If the project wasn't built from the template, that checklist offers option 2 first.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 322kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall