Walkthrough

This PR documents quota_fallback as deprecated/ignored at runtime and clarifies that Gemini quota fallback between Antigravity and Gemini CLI is always enabled. It adds a description to the JSON schema and updates configuration and multi-account docs to reflect the always-on, bidirectional fallback. In code, header routing logic is centralized via a new exported resolveHeaderRoutingDecision(urlString, family, config) that returns cliFirst, preferredHeaderStyle, explicitQuota, and allowQuotaFallback; resolveQuotaFallbackHeaderStyle is simplified to accept only family, headerStyle, and alternateStyle. Tests and __testExports were updated to expose the new routing decision hook.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

No actionable comments were generated in the recent review. 🎉

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.

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

Greptile Overview

Greptile Summary

This PR centralizes Gemini routing logic and removes the quota_fallback config flag, making bidirectional fallback between Antigravity and Gemini CLI quotas automatic for all Gemini requests.

Key changes:

• Introduced resolveHeaderRoutingDecision() to centralize routing decisions
• Removed cliFirst condition from antigravity→gemini-cli fallback path, enabling bidirectional fallback
• Deprecated quota_fallback flag (accepted but ignored)
• Updated tests to reflect new bidirectional behavior

Critical issue:
The implementation changes documented behavior for explicit quota suffixes. Previously, models with explicit :antigravity or :gemini-cli suffixes would never fallback to the alternate quota (they would only switch accounts). The old schema docs stated: "Only applies when model is requested without explicit quota suffix. Explicit suffixes like :antigravity or :gemini-cli always use that specific quota and switch accounts if exhausted."

The new code sets allowQuotaFallback: family === "gemini" unconditionally, ignoring the explicitQuota flag. This means explicit quota models will now fallback, breaking the documented contract and potentially affecting users who rely on strict quota isolation.

Confidence Score: 2/5

• This PR contains a breaking behavior change that affects explicit quota model routing
• The code quality is good and the refactoring improves maintainability, but there's a logic error where explicitQuota is no longer checked when setting allowQuotaFallback. This breaks the documented behavior that explicit quota suffixes should not fallback to alternate quotas. The test at line 157 confirms this new behavior but contradicts the old documentation.
• Pay close attention to src/plugin.ts:2834 where the routing decision logic needs to respect explicitQuota

Important Files Changed

Sequence Diagram

sequenceDiagram
    participant Client
    participant Plugin
    participant RoutingDecision
    participant AccountManager
    participant Antigravity
    participant GeminiCLI

    Client->>Plugin: Request Gemini model (e.g., gemini-3-flash)
    Plugin->>RoutingDecision: resolveHeaderRoutingDecision(url, family, config)
    
    alt cli_first = false (default)
        RoutingDecision-->>Plugin: preferredHeaderStyle="antigravity", allowQuotaFallback=true
        Plugin->>AccountManager: getCurrentOrNextForFamily("antigravity")
        Plugin->>Antigravity: Request with antigravity headers
        
        alt Antigravity exhausted on all accounts
            Plugin->>AccountManager: Check for gemini-cli availability
            Plugin->>GeminiCLI: Fallback to gemini-cli quota
            GeminiCLI-->>Plugin: Response
        else Antigravity available
            Antigravity-->>Plugin: Response
        end
    else cli_first = true
        RoutingDecision-->>Plugin: preferredHeaderStyle="gemini-cli", allowQuotaFallback=true
        Plugin->>AccountManager: getCurrentOrNextForFamily("gemini-cli")
        Plugin->>GeminiCLI: Request with gemini-cli headers
        
        alt Gemini-CLI exhausted
            Plugin->>AccountManager: Check for antigravity availability
            Plugin->>Antigravity: Fallback to antigravity quota
            Antigravity-->>Plugin: Response
        else Gemini-CLI available
            GeminiCLI-->>Plugin: Response
        end
    end
    
    Plugin-->>Client: Response