reflex-dev/reflex

reflex

Performance History

Latest Results

Review fixes: restore provider on failed switch, arg precedence, rollback hint Addresses three review comments: - P1 (cli.py): a provider switch on an already-deployed app tears its old resources down before hostname reservation, export, and deployment creation — any of which can fail, stranding the app on the new provider. Wrap that pipeline in `_restore_provider_on_failure`: on failure, best-effort re-pin the previous provider so its last deployment stays a valid `reflex cloud apps rollback` target, and tell the user how to recover. - P2 (apps.py `_resolve_app_id`): honor an explicit `--app-name` before the cloud.yml/pyproject `appid` fallback, so `--app-name` is no longer silently ignored when a config appid is present. Precedence is now --app-id > --app-name > config. - P2 (apps.py rollback): include the deployment id and `--watch` in the post-rollback tracking hint (the bare `reflex cloud apps status` exits with a missing-argument error). Adds regression tests for all three. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K1YRndqwXapfwj5Rtx9Mzz

claude/gcp-support-reflex-cli-j78sg3

2 hours ago

Default provider-switch teardown confirmation to "n" The "Continue?" prompt shown before switching a deployed app to a different provider (which tears down the current deployment) defaulted to "y", so pressing Enter alone proceeded with the destructive teardown. Default it to "n" to match the rollback confirmation, and add a regression test. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K1YRndqwXapfwj5Rtx9Mzz

claude/gcp-support-reflex-cli-j78sg3

17 hours ago

Fix pyright errors in test_hosting.py (pre-commit) The pre-commit `pyright` hook (which checks `reflex tests`) flagged the new tests: - The shared `_ok` helper is typed for dict payloads; add a small `_ok_body` helper for the list/str JSON responses (provider list, history list, deployment id). - rollback_deployment / update_deployment_description return `str | None`, so narrow with `assert result is not None` before `.startswith(...)`. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K1YRndqwXapfwj5Rtx9Mzz

claude/gcp-support-reflex-cli-j78sg3

17 hours ago

OPS-1673: audit & correct Integrations/Connections docs Verified the integration/connection docs against the flexgen codebase and fixed inaccuracies: - GitHub: repo visibility is plan-dependent (Free=public, Enterprise=private, Pro=choice), not "private unless you choose public". - Git providers: Azure DevOps is supported; added it to the list. - Database: added a "Connecting a Private or Remote Database" section (IPv4/pooler reachability, Enterprise static-egress allowlisting) and noted the app connects via REFLEX_DB_URL / REFLEX_ASYNC_DB_URL. - Snowflake: added the missing frontmatter (tags + description) so its gallery card renders correctly. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

aditya/ops-1673-audit-update-integrations-connections-docs

18 hours ago

Address review: safer rollback default, drop user-facing "fly" alias - reflex cloud apps rollback: default the confirmation to "n" so pressing Enter cancels rather than replacing the running deployment (matches apps delete). - Remove the "fly" alias from PROVIDER_ALIASES so `--provider fly` is not accepted — users shouldn't couple deploy scripts to how Reflex Cloud is hosted. Only provider-agnostic names (reflex-cloud/reflex/cloud/gcp/...) are exposed; "fly" stays an internal wire value. De-emphasize "fly" in docstrings and comments in favor of "Reflex Cloud". - Tests: rollback now defaults to cancel; normalize_provider("fly") is None. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K1YRndqwXapfwj5Rtx9Mzz

claude/gcp-support-reflex-cli-j78sg3

18 hours ago

Add GCP managed provider, rollbacks, and deployment descriptions to hosting CLI Port the cloud GCP provider feature set to the reflex-hosting-cli client and document it. The backend endpoints already exist; this wires them into the CLI with a focus on UX. reflex deploy: - Prompt for "Reflex Cloud vs. Google Cloud (GCP)" when the org has a GCP account connected and GCP deploys are allowed (Enterprise). Add --provider to skip the prompt and a `provider` config-file field. The provider is pinned before the hostname is reserved so a GCP app bakes its *.run.app backend URL into the exported frontend, and region/VM-type are dropped for GCP (they come from the connected account). - Add --description to record an optional changelog note on the deployment. reflex cloud apps: - rollback DEPLOYMENT_ID: redeploy a previous deployment's image without rebuilding from source. - describe DEPLOYMENT_ID --description: set/clear a deployment's note. - history now surfaces each deployment's description and rollback eligibility. reflex cloud providers (new group): - status / list: inspect the cloud provider accounts connected to the org. Docs: new "Cloud Providers" page, app-management history/description/rollback sections, a self-service-vs-managed note on deploy-to-gcp, CLI-reference wiring, and towncrier news fragments. Tests: unit coverage for the new hosting client functions, the deploy provider-selection flow, the apps rollback/describe commands, and the providers group. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K1YRndqwXapfwj5Rtx9Mzz

claude/gcp-support-reflex-cli-j78sg3

19 hours ago

fix: track completed post_compile hooks by identity, not hash Real plugins are dataclasses (eq=True sets __hash__ = None), so storing them in a set raised TypeError: unhashable type on every app that uses the default TailwindV4Plugin. Key the completed-hook record by id() instead, keeping the plugin as the value so its id cannot be recycled; identity is also the right semantics, since equal-but-distinct plugin instances must each run their hook. Also add the reflex-base news fragment required by the per-package changelog check for the Plugin.post_compile docstring change.

FarhanAliRaza:farhan/eng-9917-post-compile-once-per-app

21 hours ago

fix(vars): carry custom Field attrs by reference, not deepcopy (#6809) * fix(vars): carry custom Field attrs by reference, not deepcopy The source_field carry-over introduced in #6726 deep-copied each custom attribute onto the rebuilt Field. That broke two things for downstream tag consumers (e.g. Reflex Enterprise's _auth field tag): - a stateful callable marker (auth check object) ran as a deep-copied clone instead of the original instance - a marker holding non-copyable state (lock, client) crashed state class creation with TypeError: cannot pickle '_thread.lock' object The source Field is a throwaway namespace value replaced during class creation, so sharing by reference is safe, and it matches how _copy_fn carries a function's __dict__ (#6725) — the sister fix in the same release. Flips the test that asserted deepcopy isolation and adds identity + non-copyable regression tests. * chore: add changelog entry for #6809 * test: fold by-reference regression tests into one identity check Any copy scheme breaks object identity, so a single is-assert on a lock-holding marker covers reference carry, callable identity, and the non-copyable crash in one test.

main

22 hours ago

Latest Branches

Add managed GCP provider, deployment rollback, and descriptions#6810

2 hours ago

56f83e9

claude/gcp-support-reflex-cli-j78sg3

-4%

OPS-1673: Audit & update Integrations & Connections docs#6811

18 hours ago

19f5a93

aditya/ops-1673-audit-update-integrations-connections-docs

ENG-9917 fix: run plugin post_compile hooks at most once per app instance#6733

22 hours ago

f315ecf

FarhanAliRaza:farhan/eng-9917-post-compile-once-per-app

Home Terms Privacy Docs