Compare commits
1 Commits
main
...
docs/readm
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
baa828675c |
25
README.md
25
README.md
@ -50,6 +50,31 @@ workflow rebuilds this mirror + uploads the new wheel to PyPI.
|
||||
See [`docs/workspace-runtime-package.md`](https://github.com/Molecule-AI/molecule-core/blob/main/docs/workspace-runtime-package.md)
|
||||
for the full publishing flow.
|
||||
|
||||
## PyPI is canonical, this git tree may lag
|
||||
|
||||
**The PyPI wheel is the canonical artifact** for runtime consumers
|
||||
(workspace template images, the controlplane's runtime smoke harness,
|
||||
local `pip install` users). What's in PyPI and what's checked into this
|
||||
repo's git tree **may differ** — sometimes by hours, sometimes by days.
|
||||
|
||||
Why: the [`publish-runtime`](https://github.com/Molecule-AI/molecule-core/blob/main/.github/workflows/publish-runtime.yml)
|
||||
workflow always publishes a fresh wheel built from `molecule-core`'s
|
||||
`staging` HEAD, but the git mirror push that updates this repo is a
|
||||
separate step and can lag (or, on transient failures, be skipped while
|
||||
the wheel still ships). The mirror exists for **human readability** and
|
||||
to give `mirror-guard` a concrete branch to enforce its
|
||||
"no direct PRs" policy against — not as the wire-truth artifact.
|
||||
|
||||
**To check what's actually in a wheel:**
|
||||
|
||||
```bash
|
||||
pip download molecule-ai-workspace-runtime==X.Y.Z --no-deps
|
||||
unzip -l molecule_ai_workspace_runtime-X.Y.Z-*.whl
|
||||
```
|
||||
|
||||
If you're chasing "is module X in the published runtime yet?", trust
|
||||
the wheel listing, not this repo's `git log`.
|
||||
|
||||
## Why this split
|
||||
|
||||
The runtime needs to ship as a PyPI artifact (so the 8 workspace template
|
||||
|
||||
Loading…
Reference in New Issue
Block a user