Add app-server current-time impl (varlatency 3/n)

[rka-oai]

What

Server should request:

{
  "id": 42,
  "method": "currentTime/read",
  "params": {
    "threadId": "11111111-1111-1111-1111-aaaaafdc2c11"
  }
}

Client should respond with something like:

{
  "id": 42,
  "result": {
    "currentTimeAt": 1781717655
  }
}

Why

Sessions configured with clock_source = "app_server_client" need a thread-specific external time source before inference. The system clock remains the default production provider.

Validation

• cargo test -p codex-app-server-protocol
• cargo test -p codex-app-server --test all current_time_read_round_trip_adds_reminder_to_model_input
• cargo test -p codex-app-server first_attestation_capable_connection_for_thread_only_uses_thread_subscribers
• cargo test -p codex-analytics
• just fix -p codex-app-server-protocol
• just fix -p codex-app-server

Stacked on #28824.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d473fdc80a

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".

[chatgpt-codex-connector]

Wait for subscriber attachment before reading app-server time

For threads created outside the synchronous thread/start path, such as subagents, client subscription is attached later by the thread_created_rx background branch, but the first model request can call this provider immediately. With clock_source = "app_server_client" and a reminder due on the first request, this lookup can see zero current-time-capable subscribers and fail the turn even though an initialized client would be attached moments later; make creation/attachment ordered for these threads or have the provider wait/retry for the initial subscriber.

Useful? React with 👍 / 👎.

[rka-oai]

Fixed cleanly for the supported single-client case in 5617874.

[chatgpt-codex-connector]

Keep current-time capability out of v1 initialize

Adding request_current_time to v1::InitializeCapabilities expands the v1 initialize contract and regenerates schema/json/v1/InitializeParams.json with a new current-time field. This current-time negotiation is new app-server API surface, so it should be modeled in v2/common protocol instead of extending v1.

AGENTS.md reference: AGENTS.md:L257-L260

Useful? React with 👍 / 👎.

[rka-oai]

iiuc, this shouldn't be an issue

[chatgpt-codex-connector]

Expose current-time opt-in through the client facade

For callers that use app-server-client and handle AppServerEvent::ServerRequest, this hard-coded false means they cannot advertise the new current-time capability even if they implement currentTime/read. With clock_source = "app_server_client", app-server will therefore find zero capable subscribers and fail the turn before inference; add a start/connect argument and forward it here (and in the remote connect args) so facade clients can opt in.

Useful? React with 👍 / 👎.

[rka-oai]

shouldn't be an issue. this should only support clients that connect directly to the app-server using JSON-RPC