fix(ci): unpin upstream LiteLLM prices that are failing every open PR

[ohong]

Why this is urgent

CI is red on every open PR — #143, #145, #146 and #149 — for one shared reason that none of them caused:

AssertionError: expected 0.28400000000000003 to be close to 0.355
  ❯ __tests__/ccusage-pricing.integration.test.ts:94:34

packages/cli/__tests__/ccusage-pricing.integration.test.ts pins hardcoded dollar amounts for the GPT-5.6 family. Those rates are owned by LiteLLM upstream and are fetched live by the bundled ccusage binary. LiteLLM moved gpt-5.6-terra from $0.355 to $0.284, and the whole repo's Test (cli) job went red on the next run. 202 of 203 CLI tests pass; this one assertion is the blocker.

#149 is the clearest proof it isn't the PRs: it changes .gitignore and deletes committed artifacts, touches no CLI code at all, and still fails.

What changed

The test now asserts the invariant we actually own instead of the number upstream owns:

• every member of the GPT-5.6 family resolves to a non-zero LiteLLM price (catches an unpriced or renamed model, which is the real regression this test exists to catch);
• the day total equals the sum of the per-model breakdown (catches aggregation bugs);
• the 440,000-token fixture total stays pinned, since token counts are ours, not upstream's.

What it no longer does is fail when a vendor reprices a model overnight.

docs/CHANGELOG.md had the old $1.917 figure written into the ccusage entry as a durable claim; corrected to describe the invariant.

Also in here

docs/ROADMAP.md picks up two findings from reviewing the open PRs:

• Straude already ingests Gemini, Qwen, Kimi, Copilot and the rest of ccusage's sources. The collector runs unscoped ccusage daily --json, CcusageAgent is string, and /api/usage/submit has no agent allowlist. feat: add Gemini CLI usage tracking via gemistat #77 and feat: add Gemini, Qwen Code, and Mistral Vibe providers #22 both hand-wrote parsers for capability that already ships. What's missing is prettifyModel cases, model colours, and copy telling users.
• calculate_user_streak is SECURITY DEFINER with EXECUTE granted to anon, so anyone can read any user's streak by id, private profiles included. Long-standing, from 20260430172022, not from an open PR.

Validation

bun test __tests__/ccusage-pricing.integration.test.ts in packages/cli — 1 pass, 13 assertions, against the real bundled binary and live pricing.

Worth merging before the others so their CI can go green.

🤖 Generated with Claude Code

Summary by CodeRabbit

• Bug Fixes

  • Improved GPT-5.6 pricing validation to ensure all supported models return valid pricing and totals remain consistent with model-level costs.

• Documentation

  • Updated changelog guidance to reflect resilient pricing validation without relying on fixed dollar amounts.
  • Added roadmap items for multi-agent CLI support and strengthening access controls around user streak calculations.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
straude	Ready	Preview	Aug 9, 2026 8:09am

[coderabbitai]

📝 Walkthrough

Walkthrough

The GPT-5.6 pricing test now checks model coverage, positive resolved pricing, and cost consistency without fixed dollar values. The changelog records this behavior. The roadmap adds multi-agent CLI support and a security item about streak RPC grants.

Changes

GPT-5.6 pricing validation

Layer / File(s)	Summary
Pricing assertions and changelog packages/cli/__tests__/ccusage-pricing.integration.test.ts, docs/CHANGELOG.md	The test validates all GPT-5.6 variants, positive LiteLLM pricing, and consistency between model, daily, and summary totals. The changelog documents the non-pinned dollar values.

Multi-agent CLI roadmap

Layer / File(s)	Summary
Multi-agent CLI support docs/ROADMAP.md	The roadmap adds planned model naming, color support, supported-agent documentation, and Mistral Vibe support.

Streak RPC security roadmap

Layer / File(s)	Summary
Streak RPC grant review docs/ROADMAP.md	The roadmap documents arbitrary-user streak access through anon and authenticated grants and proposes call-site verification followed by grant revocation.

Estimated code review effort: 2 (Simple) | ~10 minutes

Suggested reviewers: alexesprit

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly describes the main change: removing hardcoded upstream LiteLLM price assertions from the CI pricing test.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch oh-review-followups

Warning

There were issues while running some tools. Please review the errors and either fix the tool's configuration or disable the tool if it's a critical failure.

🔧 ESLint

If the error stems from missing dependencies, add them to the package.json file. For unrecoverable errors (e.g., due to private dependencies), disable the tool in the CodeRabbit configuration.

ESLint install failed. For unrecoverable errors, disable the tool in CodeRabbit configuration.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

🧹 Nitpick comments (1)

docs/ROADMAP.md (1)

40-40: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Define the remaining documentation deliverable.

packages/cli/README.md, Lines 15-20, already lists Kimi, Qwen, Copilot CLI, and Gemini CLI. State whether this item adds landing-page copy, a dedicated supported-agents section, or an explicit support guarantee. Link to the existing README section. This gives future work a clear, non-duplicative acceptance target.

As per coding guidelines, docs/ROADMAP.md must include feature requests, ideas, and improvements with sufficient context for future pickup.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@docs/ROADMAP.md` at line 40, Clarify the roadmap item about supported-agent
messaging by specifying the exact documentation deliverable, such as
landing-page copy or a linked supported-agents section. Reference the existing
supported-agents section in packages/cli/README.md and define a non-duplicative
acceptance target rather than leaving the scope as general README and
landing-page copy.

Source: Coding guidelines

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@docs/ROADMAP.md`:
- Line 40: Clarify the roadmap item about supported-agent messaging by
specifying the exact documentation deliverable, such as landing-page copy or a
linked supported-agents section. Reference the existing supported-agents section
in packages/cli/README.md and define a non-duplicative acceptance target rather
than leaving the scope as general README and landing-page copy.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: b9a7675a-4924-4d96-a7b7-718a1d92cc44

📥 Commits

Reviewing files that changed from the base of the PR and between f007465 and 17593a3.

📒 Files selected for processing (3)

• docs/CHANGELOG.md
• docs/ROADMAP.md
• packages/cli/__tests__/ccusage-pricing.integration.test.ts