🤖 fix: attribute session usage sources and dedupe abort cleanup

[ibetitsmike]

Summary

Add source-level session usage attribution and harden stream abort cleanup so interrupted streams cannot double-count token usage or continue consuming work after a soft interrupt.

Background

We needed clearer attribution for high input-token sessions (main chat vs. system overhead) and identified interrupt-path races that could overcount usage during abort handling.

Implementation

• Added SessionUsageSource categories (main, system1, plan, subagent) and threaded attribution through:
  • usage persistence (SessionUsageService)
  • stream/session schemas
  • workspace rollup delta events
  • AI request entry points (AIService, System1ToolWrapper)
  • renderer state hydration/merge logic (WorkspaceStore)
  • Costs tab source breakdown UI
• Hardened StreamManager abort teardown:
  • deduplicate overlapping cleanup paths with a shared abort-cleanup promise
  • signal abort immediately before awaiting partial-write flushing, preventing extra stream/tool work after a soft interrupt
  • added regression coverage for cleanup dedupe and blocked-flush interrupt ordering
• Updated WorkspaceStore repricing to keep bySource costs aligned with repriced byModel totals using blended session component rates, preventing stale source-cost displays after pricing/mapping changes.

Validation

• make static-check
• bun test src/node/services/streamManager.test.ts src/node/services/sessionUsageService.test.ts src/browser/stores/WorkspaceStore.test.ts src/common/orpc/schemas/chatStats.test.ts

Risks

• Low-to-moderate: touches stream interrupt teardown and usage accounting paths. Regressions would most likely appear as incorrect source-level cost attribution or interrupt edge-case behavior; both paths now have focused regression tests.

---

Generated with mux • Model: openai:gpt-5.3-codex • Thinking: xhigh • Cost: $13.74

[ibetitsmike]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 17ddca6fef

ℹ️ About Codex in GitHub

Codex has been enabled to automatically 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 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[ibetitsmike]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8f55612f93

ℹ️ About Codex in GitHub

Codex has been enabled to automatically 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 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[ibetitsmike]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. 🎉

ℹ️ About Codex in GitHub

Codex has been enabled to automatically 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 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".