feat: Add support for the Antigravity client

[bruce-shi]

feat: Add support for the Antigravity client, including session parsing from .pb files and token usage extraction via language server RPC:
POST /exa.language_server_pb.LanguageServerService/GetCascadeTrajectory

---

Summary by cubic

Add Antigravity (Windsurf/Codeium) client support to parse local .pb conversations and fetch token usage via the language server RPC. Synced with upstream and fixed client index/count to avoid ID conflicts.

• New Features

  • Added Antigravity client scanning ~/.gemini/antigravity/conversations for *.pb.
  • Discovers the language server and POSTs to /exa.language_server_pb.LanguageServerService/GetCascadeTrajectory to extract input/output/cache-read/reasoning tokens.
  • Deduplicates by responseId, resolves model aliases/placeholders, and infers provider for pricing.
  • Integrated into parse_all_messages_with_pricing and parse_local_clients; TUI shows "Antigravity" with hotkey a.
  • Added Windsurf model aliases in pricing to normalize variants.

• Migration

  • Ensure the Windsurf/Antigravity language server is running during parsing; otherwise no usage is returned.
  • No config changes required.

^{Written for commit 71880af. Summary will update on new commits.}

[vercel]

@bruce-shi is attempting to deploy a commit to the Inevitable Team on Vercel.

A member of the Team first needs to authorize it.

[cubic-dev-ai]

3 issues found across 8 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="crates/tokscale-core/src/sessions/antigravity.rs">

<violation number="1" location="crates/tokscale-core/src/sessions/antigravity.rs:310">
P2: Unchecked `u64`→`i64` cast can wrap large token counts negative and then clamp them to 0, causing silent token undercounting.</violation>

<violation number="2" location="crates/tokscale-core/src/sessions/antigravity.rs:325">
P2: Unknown/unmapped Antigravity models are incorrectly attributed to Google because provider fallback is hardcoded to `google` when inference fails.</violation>

<violation number="3" location="crates/tokscale-core/src/sessions/antigravity.rs:495">
P2: RPC JSON body is built via string interpolation, so `cascade_id` is not escaped and can break request serialization for valid filename stems.</violation>
</file>

---

Since this is your first cubic review, here's how it works:

• cubic automatically reviews your code and comments on bugs and improvements
• Teach cubic by replying to its comments. cubic learns from your replies and gets better over time
• Add one-off context when rerunning by tagging @cubic-dev-ai with guidance or docs links (including llms.txt)
• Ask questions if you need clarification on any suggestion

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

P2: Unknown/unmapped Antigravity models are incorrectly attributed to Google because provider fallback is hardcoded to google when inference fails.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At crates/tokscale-core/src/sessions/antigravity.rs, line 325:

<comment>Unknown/unmapped Antigravity models are incorrectly attributed to Google because provider fallback is hardcoded to `google` when inference fails.</comment>

<file context>
@@ -0,0 +1,745 @@
+
+/// Get provider from model name (e.g. "gemini-2.5-pro" → "google").
+fn get_provider(model: &str) -> &'static str {
+    provider_identity::inferred_provider_from_model(model).unwrap_or("google")
+}
+
</file context>

Suggested change

	provider_identity::inferred_provider_from_model(model).unwrap_or("google")
	provider_identity::inferred_provider_from_model(model).unwrap_or("unknown")

[cubic-dev-ai]

P2: RPC JSON body is built via string interpolation, so cascade_id is not escaped and can break request serialization for valid filename stems.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At crates/tokscale-core/src/sessions/antigravity.rs, line 495:

<comment>RPC JSON body is built via string interpolation, so `cascade_id` is not escaped and can break request serialization for valid filename stems.</comment>

<file context>
@@ -0,0 +1,745 @@
+            None => continue,
+        };
+
+        let body = format!(r#"{{"cascadeId":"{}"}}"#, cascade_id);
+        let resp = match rpc_call(&server, "GetCascadeTrajectory", &body) {
+            Some(r) => r,
</file context>

Suggested change

	let body = format!(r#"{{"cascadeId":"{}"}}"#, cascade_id);
	let body = serde_json::json!({ "cascadeId": cascade_id }).to_string();

[cubic-dev-ai]

P2: Unchecked u64→i64 cast can wrap large token counts negative and then clamp them to 0, causing silent token undercounting.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At crates/tokscale-core/src/sessions/antigravity.rs, line 310:

<comment>Unchecked `u64`→`i64` cast can wrap large token counts negative and then clamp them to 0, causing silent token undercounting.</comment>

<file context>
@@ -0,0 +1,745 @@
+    value
+        .and_then(|v| {
+            v.as_i64()
+                .or_else(|| v.as_u64().map(|n| n as i64))
+                .or_else(|| v.as_str().and_then(|s| s.parse::<i64>().ok()))
+        })
</file context>