Can you update our canonical improvement script runner now that our backend is modal , not digital ocean based ?

Can you update our canonical improvement script runner now that our backend is modal , not digital ocean based ?

I’ll inspect the runner and deployment/docs references first, then patch the smallest surface that still reflects the Modal backend accurately. Since your repo instructions require it, I’ll also commit, push, and run deploy if there’s an existing deploy target once the change is verified.

The first pass shows a lot of stale DigitalOcean language in docs, but the likely work is the canonical dataset prompt/script path rather than generated report artifacts. I’m narrowing on the runner files and package scripts now so the patch doesn’t churn old evaluation output.

I found the concrete issue in scripts/start-canonical-dataset-improvement-jobs.mjs: it still sends a DigitalOcean-style runnerSize slug and requires deploymentStatus === "ready" before starting improvements. For Modal-backed datasets, I’m switching that runner metadata to a Modal resource profile and making readiness tolerate datasets that no longer report a separate deployment status.

There’s a second path that can launch a single canonical improvement run through npm run canonical:dataset -- improve. I’m updating its exported resource contract too so batch and single-dataset runners don’t drift.

The code changes are in place. I also updated the architecture/lifecycle docs that directly described DigitalOcean runners, because the runner change would otherwise leave contradictory guidance in the repo. Next I’m running the focused canonical tests and checks.

The focused test passes. The batch dry run also emitted Modal resources correctly, but exited nonzero because this session only has econ present and the script treats missing canonical datasets as failures. I found stale runnerSize examples in docs/HARNESS.md, so I’m correcting those before the broader checks.

The direct runnerSize references are gone from the active runner surfaces. I’m also cleaning the remaining harness wording that names DigitalOcean as the live infrastructure dependency, replacing it with Modal/backend infrastructure where appropriate.

Focused checks are green: canonical tests, docs check, and architecture check. Since this touches runner contracts and docs, I’m starting the repo’s full agent:check gate now; after that I’ll handle git commit/push per your repository instructions.

agent:check passed, including the deploy readiness check. I don’t see an actual deploy workflow beyond deploy:check in this repo, so the deploy-related requirement is satisfied by the existing deployment readiness target. I’m checking the worktree now, then I’ll commit and push.

The commit is created locally. I’m pushing it to the current codex/initial-substrate branch now.

Updated the canonical improvement runner for Modal:

• Replaced DigitalOcean runnerSize slugs with Modal resource fields: backend, resourceProfile, cpu, and memoryGb.
• Made the batch improvement runner tolerate Modal datasets that no longer report a separate deploymentStatus.
• Updated single-dataset canonical resource config, docs, and a regression test.

Validation passed:

• npm run test:canonical
• npm run docs:check
• npm run architecture:check
• npm run agent:check including deploy:check

Committed and pushed: 60fc2f3 Update canonical runner for Modal backend

::git-stage{cwd="/Users/example/Documents/Projects/alpha-datasets-canonical"} ::git-commit{cwd="/Users/example/Documents/Projects/alpha-datasets-canonical"} ::git-push{cwd="/Users/example/Documents/Projects/alpha-datasets-canonical" branch="codex/initial-substrate"}

only econ should be present. Get rid of references to the old ones

I’ll remove the non-econ canonical dataset entries from the runner/docs surfaces so dry runs and guidance match the current one-dataset state, then re-run the focused checks and push another commit.

There are three kinds of old references: scheduled runner arrays, public docs pages/briefings, and catalog/onboarding docs. I’m going to keep econ and remove the other canonical pages rather than leave them orphaned.

The expansion runner and one harness test still use sociology as the canonical example, and the docs still have legacy public dataset pages. I’m converting those examples to econ and removing the old public dataset documents.

The old dataset references are now removed from active code and docs. I’m running the repository search again before tests so we catch any lingering references outside archived generated run prompts/reports.

A full repo search still shows the old DigitalOcean ops bundle and AGENT docs. I’m removing that stale deployment target, replacing it with a Modal ops note, and updating deploy:check so the required deployment-readiness check points at the current backend target.

The stale references are down to the negative regression test that asserts runnerSize is gone. I’m running the focused dry runs and checks now, including the CLI harness because resource profiles changed in apps/cli/src/agent.ts.

Focused checks passed, including dry-run output now showing only econ. I’m running the full gate once more because this touched docs, CLI behavior, and deployment readiness.

The full gate passed and the remaining search hit is only the regression test proving runnerSize is absent. I’m committing and pushing the cleanup now.

Done. I removed the old non-econ canonical dataset references from runners, docs, generated briefings, tests, and public dataset pages. I also removed the stale DigitalOcean ops target and replaced it with ops/modal.

Verified:

• npm run canonical:improve:dry-run now returns only econ
• npm run canonical:expand:dry-run now returns only econ
• npm run test:canonical
• npm run test:cli
• npm run docs:check
• npm run deploy:check
• npm run agent:check

Committed and pushed: 504b3c0 Keep canonical jobs scoped to econ

::git-stage{cwd="/Users/example/Documents/Projects/alpha-datasets-canonical"} ::git-commit{cwd="/Users/example/Documents/Projects/alpha-datasets-canonical"} ::git-push{cwd="/Users/example/Documents/Projects/alpha-datasets-canonical" branch="codex/initial-substrate"}