Draft improvements from assistant conversations: run moves and nested config#2830
Draft improvements from assistant conversations: run moves and nested config#2830mintlify[bot] wants to merge 2 commits into
Conversation
|
Preview deployment for your docs. Learn more about Mintlify Previews.
|
📚 Mintlify Preview Links📝 Changed (2 total)📄 Pages (2)
🤖 Generated automatically when Mintlify deployment succeeds |
🔗 Link Checker Results✅ All links are valid! No broken links were detected. Checked against: https://wb-21fd5541-mintlify-8476678c.mintlify.site |
📋 Validation of technical assertionsEvery claim below was checked against the truth on the ground using Glean
Open items — recommended next steps
Deviation from the sourceThe proposed additions contain two universal claims that the implementation contradicts: deletion is not the Public API's only run-management operation, and 🤖 Generated by the CoreWeave Docs Team's Validator ( |
|
Validator and style guide corrections applied. |
Summary
Closes two recurring documentation gaps surfaced by recent assistant conversation buckets: users repeatedly ask whether runs can be moved programmatically, and how to put a nested dictionary into
wandb.config.Changes
models/runs/manage-runs.mdx: state explicitly that moving runs is UI-only with no SDK or Public API equivalent, that copying runs is not supported, and point toRun.delete()as the one run-management call available from the Public API.models/track/config.mdx: add a "Nested dictionaries in config" subsection explaining thatconfig-defaults.yaml(desc/value) does not support nested dicts inline, showing the Pythonwandb.init(config=...)pattern, and noting that W&B flattens nested keys with dots in the UI and Public API filters.Context
Recent assistant traffic shows repeated questions like "are there API endpoints to move runs", "are there programmatic ways to move runs", and YAML/config questions where users tried to encode nested dicts in
config-defaults.yaml. The existing pages described the supported paths but did not call out the unsupported ones, so the assistant had to infer the negative answer each time.