Include platform_cost in cloud agent credit totals

[jefflloyd]

Description

The conversation details panel often rendered no Credits used field at all, even when the Oz web app showed a non-zero total for the same run. Two compounding bugs were responsible:

1. ConversationDetailsData::from_task used the ? operator inside an and_then closure, so if either inference_cost or compute_cost was null on RequestUsage, the entire credits value became None and the row was dropped from render. Oz web handled the same data with ?? 0 and still rendered a number.
2. The client's RequestUsage struct was missing the server's platform_cost field entirely, so even after the Oz-web fix/oz-credits-include-platform-cost change, the local panel would still under-count by whatever was sitting in the platform bucket.

What changed

• Added platform_cost: Option<f64> (with #[serde(default)] for forward compat with older server responses) to the client's RequestUsage in app/src/ai/ambient_agents/task.rs.
• Updated AmbientAgentTask::credits_used() to sum inference + compute + platform, each defaulting to 0.0 when null. This is the single helper every cloud-agent surface (details panel, tombstone, conversation list entries) now goes through.
• Collapsed the CreditsInfo enum in conversation_details_panel.rs down to credits: Option<f32> and deleted the inference/compute split renderer (render_credits_with_split / render_cost_sub_row and the two info-tooltip mouse states). Per product direction we now want a single Credits used: X.X line everywhere; no split UI.
• Both code paths land on that single total:
  • Local conversations: conversation.credits_spent() (already sums platform credits at the controller).
  • Cloud agent runs: task.credits_used() (now also sums platform).

Linked Issue

• The linked issue is labeled ready-to-spec or ready-to-implement.
• Where appropriate, screenshots or a short video of the implementation are included below (especially for user-visible or UI changes).

Testing

• Updated conversation_details_panel_tests::test_from_conversation_populates_local_conversation_fields to assert data.credits.is_some() after dropping the CreditsInfo enum.

• Updated conversation_ended_tombstone_view_tests to populate platform_cost on the test RequestUsage and to expect inference + compute + platform in the formatted credits string (catches future regressions if a bucket gets dropped from the sum again).

• cargo clippy -p warp --all-targets --tests --no-deps -- -D warnings is clean.

• cargo nextest run -p warp -E 'test(conversation_details_panel) or test(conversation_ended_tombstone_view)' — 15 passed.

• I have manually tested my changes locally with ./script/run

Screenshots / Videos

N/A — purely additive to an existing row in the details panel; the row now appears (with the correct total) for runs that previously rendered no credits at all.

Agent Mode

• Warp Agent Mode - This PR was created via Warp's AI Agent Mode

CHANGELOG-BUG-FIX: Cloud agent run details now include platform cost in the "Credits used" total (and no longer drop the field entirely when one cost bucket is null).

Co-Authored-By: Oz oz-agent@warp.dev

[oz-for-oss]

@jefflloyd

I'm starting a first review of this pull request.

You can view the conversation on Warp.

I completed the review and no human review was requested for this pull request.

Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).

Powered by Oz

[oz-for-oss]

Overview

This PR adds platform_cost to ambient task usage, centralizes cloud-run credit summing through AmbientAgentTask::credits_used(), and simplifies the details panel to display a single Credits used total.

Concerns

• Task-backed conversation entries can still hide the credits row when the raw task is not loaded, even though the entry already carries a normalized total.
• This is a user-visible details-panel behavior change, but the PR has no screenshot/screen recording and the manual testing checkbox is unchecked; manual testing evidence or a justification is required.

Verdict

Found: 0 critical, 2 important, 0 suggestions

Request changes

Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).

Powered by Oz

[jefflloyd]

Instead of splitting everything out by cost bucket, we're just going to sum up to one overall number. The API always exists for more detailed breakdowns.