[AGENT] Add contributor collections, local uploads and media review - #343
Conversation
PR verification reportAll configured preview and verification checks passed.
Changed visual baselines (2)
Updated from Baseline Checks run 35907709649, attempt 1 for |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
[AGENT] Claude ReviewClaude review for commit This is not a current review or approval. |
|
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f79c68748c
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
[AGENT] The GitHub Claude review stopped before inference because this PR exceeds its 500,000-byte textual-diff limit. That job remains unavailable; its safety limit and skip controls are unchanged. A local read-only Claude fallback reviewed public base 74cf6b8 through f79c687 in three file-boundary partitions. The returned model was claude-opus-5. The recorded reads cover all 21,956 text-diff lines. It did not run tests or inspect screenshot binaries. Source validation confirmed one OAuth consent-recovery issue for the new contribution tools, included in the current correction round. Its concern about daily cleanup re-arming was not confirmed as a blocker: that schedule intentionally reconciles late writes, and processed rows move to the following day in the ordered queue. The review does not establish hosted storage behavior or cleanup throughput. |
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
|
[AGENT] @codex review Please review the correction delta from f79c687 to 3ac183b and verify the resolved technical findings. Each correction round passed independent scoped review. The final candidate passed 892 backend tests, 487 web tests, typechecks, API/MCP package checks, documentation build and local Convex runtime validation. The two exact-copy approval threads intentionally remain open pending BASIC's decision. Hosted feature enablement still has separate S3, authenticated journey and installed-client proof gates. |
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
|
[AGENT] @greptileai Please review the correction delta from f79c687 to 3ac183b and verify the resolved technical findings. Both correction rounds passed independent scoped review. The exact-copy approval threads intentionally remain open pending BASIC's decision. |
|
[AGENT] @greptileai Please review the correction from 3ac183b to bf2f21d. The local read-only Claude delta review completed through 3ac183b and found a replay signing failure being reported as terminal while its reservation remained live. db2f6ff fixes that classification: only confirmed settlement produces terminal refusal; no-op or uncertain settlement returns same-key retry guidance. Its independent scoped review passed, including real adapter/backend and registered MCP callback cases for lost acknowledgements. This head also integrates main's merged event-lineup PR342. Independent integration review confirmed both parent MCP test sets and generated contracts are retained. The remaining follow-up only removes EOF whitespace. CI is now running on bf2f21d. The two exact-copy approval threads remain open pending BASIC's decision. GitHub Claude's full-PR review remains unavailable because of its diff-size cap, and the Codex re-review request hit the account's code-review quota. Local review evidence does not turn those unavailable reviews into approvals. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8658531ba3
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
[AGENT] The twelve technical review threads have individual dispositions and are resolved. The corrections are in follow-up PR #345, which is undergoing its own CI and review. This link tracks the postmerge fixes; it does not change the already deployed PR343 release. |


[AGENT] Contributors can organize collections, upload local images through a temporary multipart endpoint, and inspect and review stored media through MCP or the website. Shared versioned commands and durable receipts recover lost responses. Separate reviewer, publisher and capacity grants preserve independent review and restrict trusted publication to eligible empty slots.
This combines the three approved phases in one PR: assigned review, native current/candidate previews, withdrawal, paginated contributor history, bounded intake and retention, cleanup, and migration controls. Admission and retries retain current authorization. Private current images use reviewer-authorized access, OAuth challenges identify the required grant, and archived or empty batches remain subject to retained-capacity limits.
Local evidence includes the mixed 30-item/20-image collection and 1,100-item capacity proofs using injected storage and synthetic accounts, desktop/mobile comparison and publication-recovery cases, and concurrency, revocation, migration and cleanup regressions. A large-history regression verifies byte-bounded envelope counting, honest partial counts and same-key recovery. Signing recovery tests cover no-op settlement and lost acknowledgements through the real adapter, registered MCP callback and local backend. Each correction passed independent scoped review.
Feature-specific hosted S3 transfer/cleanup, authenticated multi-user journeys, installed Codex/Claude image rendering and exact product-copy approval remain release gates. Local upload intake defaults to disabled, and elevated capacity policy is restricted to the local proof identity. The operator checkpoint covers migration/accounting reconciliation, approved-target prerequisites and rollback.