mirror of
https://github.com/hanzoai/frames.git
synced 2026-09-24 00:08:03 +00:00
zooqueen
62bbc7766c
Every SKILL.md already declared the new name — name: frames, frames-cli, frames-core, and so on — but the directories were never moved, so the tree carried frames-* in content and hyperframes-* on disk. Everything that resolves a skill by path missed: the core-skill list, the plugin marketplace pin, and the docs the CLI reads at runtime. Eight tests failed on main because the files they open do not exist. The directories now match what the files say they are, and the guide the claude-design contract reads takes the name that contract already used.
63 lines
3.5 KiB
Markdown
63 lines
3.5 KiB
Markdown
# Cloud Run rendering on Google Cloud
|
|
|
|
Use `frames cloudrun` only when the user explicitly wants self-managed Google Cloud infrastructure. It deploys Cloud Run, Workflows, and Cloud Storage. For a managed default use `frames cloud`; for AWS use `frames lambda`.
|
|
|
|
## Prerequisites
|
|
|
|
- `gcloud` is authenticated and the target project has billing enabled.
|
|
- Terraform 1.5 or newer is on `PATH`.
|
|
- Docker or permission to use Cloud Build is available.
|
|
|
|
## Lifecycle
|
|
|
|
```bash
|
|
npx @hanzo/frames cloudrun deploy --project <gcp-project> --region us-central1
|
|
npx @hanzo/frames cloudrun sites create ./project
|
|
npx @hanzo/frames cloudrun render ./project --width 1920 --height 1080 --wait
|
|
npx @hanzo/frames cloudrun progress <execution-name>
|
|
npx @hanzo/frames cloudrun destroy --project <gcp-project>
|
|
```
|
|
|
|
`deploy` enables the required Google APIs, builds or accepts a container image, applies the bundled Terraform module, and stores the resulting coordinates in `~/.frames/cloudrun-state.json`. Use deploy flags such as `--image`, `--repo`, `--cpu`, `--memory`, `--max-instances`, and `--timeout` only when the infrastructure needs those overrides.
|
|
|
|
`sites create` uploads a content-addressed project archive for reuse by `render-batch`; pass its `--site-id` to that command to skip the batch upload. Single `cloudrun render` currently resolves the project from its directory and does not consume `--site-id`. Both render commands require `--width` and `--height`; supported output formats are `mp4`, `mov`, `png-sequence`, and `webm`. Use `--output-resolution 4k` to supersample an authored composition without changing its layout dimensions.
|
|
|
|
Common render flags are `--fps 24|30|60`, `--quality draft|standard|high`, `--codec h264|h265` for MP4, `--chunk-size`, `--max-parallel-chunks`, `--target-chunk-frames`, `--render-id`, `--output-key`, `--wait`, and `--wait-interval-ms`. Use `--json` for machine-readable output.
|
|
|
|
For a variable-driven single render:
|
|
|
|
```bash
|
|
npx @hanzo/frames cloudrun render ./template \
|
|
--width 1920 --height 1080 \
|
|
--variables-file ./alice.json \
|
|
--strict-variables \
|
|
--wait
|
|
```
|
|
|
|
Use exactly one of `--variables` and `--variables-file`. Read [`variables-and-media.md`](../../frames-core/references/variables-and-media.md#variables) for the composition-side contract.
|
|
|
|
## JSONL batches
|
|
|
|
```bash
|
|
npx @hanzo/frames cloudrun render-batch ./template \
|
|
--batch ./users.jsonl \
|
|
--width 1920 --height 1080 \
|
|
--max-concurrent 10 \
|
|
--site-id <site-id> \
|
|
--json
|
|
```
|
|
|
|
Each nonblank line must contain an `outputKey`; `variables` is optional:
|
|
|
|
```json
|
|
{ "outputKey": "renders/alice.mp4", "variables": { "name": "Alice" } }
|
|
```
|
|
|
|
- `--max-concurrent` defaults to `50` and limits in-flight executions. `--max-parallel-chunks` separately limits chunks inside one render.
|
|
- `--dry-run` parses the file and prints `would-start` rows without starting executions.
|
|
- The template uploads once unless `--site-id` is supplied.
|
|
- Per-entry start errors remain visible and make the command exit nonzero.
|
|
- Do not rely on `--strict-variables` for `cloudrun render-batch`: the current command accepts the flag but does not validate batch rows. Validate the JSONL variable objects against the composition schema before dispatch. The strict gate does work for single `cloudrun render`.
|
|
|
|
`render` without `--wait` returns an execution name. Use `cloudrun progress <execution-name>` until it succeeds, then verify the reported GCS output. `destroy` removes the Terraform-managed stack and its scratch bucket; preserve any deliverables that must outlive the stack.
|