feat: generalize ensureThinkingBudget to all providers + preserve server-side tool invocations on antigravity (#6979)

* fix(6914,6912): enable server-side tool invocations on antigravity + remove clinepass gate from ensureThinkingBudget

#6914: Antigravity executor was not passing include_server_side_tool_invocations:
true in toolConfig, causing server-side tool calls to be silently dropped.

#6912: ensureThinkingBudget was gated to clinepass providers only, leaving
non-clinepass reasoning models (nvidia, deepseek, etc.) vulnerable to empty
content when the thinking budget consumed all of max_tokens. Gate removed so
the budget floor applies universally.

* fix(6914,6912): address code review + CI file-size

antigravity.ts: preserve includeServerSideToolInvocations through
sanitizeAntigravityGeminiRequest by reading it from the raw toolConfig
before rebuilding (gemini-code-assist high).

default.ts: use whichever key (max_tokens or max_completion_tokens)
was already on the body, avoiding re-introducing max_tokens alongside
max_completion_tokens for recent OpenAI models (gemini-code-assist medium).

file-size-baseline.json: rebaseline executor-antigravity.test.ts
942->977 (+35, server-side tool invocation test) and default.ts
877->879 (+2, tokenKey logic).

* fix(ci): update default.ts file-size baseline 879->881 (thinking-budget generalization)

* refactor(executors): compact generalized thinking-budget block (file-size cap)

default.ts is frozen at 877 LOC with zero headroom; the generalized
ensureThinkingBudget() + max_completion_tokens-key handling added ~4 net
lines. Tighten the accompanying comments (no behavior change) so the file
stays within the existing 877 cap instead of raising it.

Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>

* chore(quality): drop obsolete antigravity-test rebaseline, annotate codex-test bump

Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>

* test(antigravity): move #6914 server-side-tools cases to own file (test-size cap)

Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>

---------

Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
Co-authored-by: rafaumeu <53516504+rafaumeu@users.noreply.github.com>

config/quality/file-size-baseline.json

@@ -143,6 +143,7 @@
   "_rebaseline_2026_06_20_1449_1444_test_route": "Re-baseline providers test route.ts 842->887: combined growth of sibling fixes #1449 (bound OAuth connection-test probe with a timeout) + #1444 (label a deactivated account distinctly from a revoked token), both at the same connection-test chokepoint. Cohesive route handler; not extractable without hiding the test flow.",
   "_rebaseline_2026_06_20_1409_1294_models": "Re-baseline src/lib/db/models.ts 1184->1221: combined growth of sibling fixes #1409 (cascade-delete orphaned model aliases when a provider is removed) + #1294 (persist max_input_tokens/max_output_tokens on custom models), both adding CRUD at the existing models domain module. Cohesive db module; not extractable.",
   "_rebaseline_2026_06_20_4389_thinking_toolchoice": "Re-baseline base.ts 1387->1399 (#4389): tool_choice-forced thinking guard at the existing Claude wire-image injection chokepoint (effThinking gate avoids the Anthropic 400 when tool_choice forces a tool). Cohesive guard; structural shrink tracked in #3501.",
+  "_rebaseline_2026_07_18_6979_codex_test": "PR #6979 own growth: executor-codex.test.ts 1340->1347 (+7 = generalized ensureThinkingBudget assertion added to the existing codex thinking-budget cases). antigravity-test bump 942->977 REVERTED here: #7408's test split dropped that file to 888, so this PR's +35 fits under the original 942 frozen cap.",
   "cap": 800,
   "frozen": {
     "_rebaseline_2026_07_02_5816_qoder": "PR #5816 (@AgentKiller45, qoder PAT via qodercli): qoderCli.ts 666->989, new-above-cap frozen (owner-approved baseline freeze). The growth is the legitimate PAT job-token exchange + quota parsing CLI transport (the pure-JS Cosy path 500'd on every PAT request); extracting the spawn/parse helpers now would just add indirection to a contributor PR mid-merge. Test frozen also raised for this PR's coverage growth: providers-page-utils.test.ts 1052->1092. Additionally clears an inherited base-red from the already-merged #5933 (codex json_schema->text.format): translator-openai-responses-req.test.ts 1097->1172 (+75 regression tests, no offending branch left). All remain frozen (cannot grow further); release captain's rebaseline-at-release supersedes.",
@@ -316,7 +317,7 @@
     "tests/unit/db-settings-crud.test.ts": 941,
     "tests/unit/deepseek-web.test.ts": 1092,
     "tests/unit/executor-antigravity.test.ts": 942,
-    "tests/unit/executor-codex.test.ts": 1340,
+    "tests/unit/executor-codex.test.ts": 1347,
     "tests/unit/executor-default-base.test.ts": 1523,
     "tests/unit/grok-web.test.ts": 2437,
     "tests/unit/image-generation-handler.test.ts": 2019,

open-sse/executors/antigravity.ts

@@ -475,7 +475,18 @@ function sanitizeAntigravityGeminiRequest(
   const geminiTools = buildGeminiTools(request.tools);
   if (geminiTools) {
     clean.tools = geminiTools;
-    clean.toolConfig = { functionCallingConfig: { mode: "VALIDATED" } };
+    // #6914: Preserve includeServerSideToolInvocations from the raw request's
+    // toolConfig when present (set by transformRequest when tools exist). The
+    // sanitize whitelist would otherwise rebuild toolConfig without it.
+    const rawToolConfig = asRecord(request.toolConfig);
+    const rawFnConfig = asRecord(rawToolConfig?.functionCallingConfig);
+    const includeServerSide = rawFnConfig?.includeServerSideToolInvocations === true;
+    clean.toolConfig = {
+      functionCallingConfig: {
+        mode: "VALIDATED",
+        ...(includeServerSide ? { includeServerSideToolInvocations: true } : {}),
+      },
+    };
   } else if (asRecord(request.toolConfig)) {
     clean.toolConfig = request.toolConfig;
   }
@@ -708,7 +719,7 @@ export class AntigravityExecutor extends BaseExecutor {
       safetySettings: getAntigravitySafetySettings(normalizedRequest?.safetySettings),
       toolConfig:
         Array.isArray(normalizedRequest?.tools) && normalizedRequest.tools.length > 0
-          ? { functionCallingConfig: { mode: "VALIDATED" } }
+          ? { functionCallingConfig: { mode: "VALIDATED", includeServerSideToolInvocations: true } }
           : normalizedRequest?.toolConfig,
     };
 

open-sse/executors/default.ts

@@ -562,9 +562,7 @@ export class DefaultExecutor extends BaseExecutor {
       withDefaults &&
       typeof withDefaults === "object" &&
       !Array.isArray(withDefaults) &&
-      (this.provider === "cerebras" ||
-        this.provider === "mistral" ||
-        this.provider === "nvidia") &&
+      (this.provider === "cerebras" || this.provider === "mistral" || this.provider === "nvidia") &&
       Object.prototype.hasOwnProperty.call(withDefaults, "client_metadata")
     ) {
       const withoutClientMetadata = { ...(withDefaults as Record<string, unknown>) };
@@ -732,10 +730,8 @@ export class DefaultExecutor extends BaseExecutor {
       }
     }
 
-    // ClinePass reasoning models burn all of max_tokens on the thinking phase
-    // when the budget is too small, leaving content empty (finish_reason:
-    // "length"). Bump max_tokens to a safe floor when reasoning is enabled and
-    // the budget is undersized. CLINEPASS-GATED — no-op for every other provider.
+    // Reasoning models burn all of max_tokens on the thinking phase when the budget is too
+    // small, leaving content empty (finish_reason: "length"); applies to all providers (#6912).
     if (typeof withDefaults === "object" && withDefaults !== null) {
       this.ensureThinkingBudget(withDefaults as Record<string, unknown>, model);
     }
@@ -760,12 +756,10 @@ export class DefaultExecutor extends BaseExecutor {
     return withDefaults;
   }
 
-  // ClinePass / OpenRouter-style thinking models leave content empty when the
-  // reasoning budget consumes all of max_tokens. Bump max_tokens to a safe
-  // minimum only when reasoning is enabled and the budget is undersized.
-  // CLINEPASS-GATED: returns early for every other provider.
+  // Reasoning models (ClinePass, OpenRouter, etc.) leave content empty when the reasoning
+  // budget consumes all of max_tokens; bump max_tokens to a safe minimum when undersized.
   ensureThinkingBudget(body: Record<string, unknown>, model: string): Record<string, unknown> {
-    if (!body || this.provider !== "clinepass") return body;
+    if (!body) return body;
 
     const outboundModel = typeof body.model === "string" ? body.model : model;
     const entry = getRegistryEntry(this.provider);
@@ -789,10 +783,15 @@ export class DefaultExecutor extends BaseExecutor {
     const target = Math.min(MIN_TOKENS, maxOutput);
     const current = body.max_tokens ?? body.max_completion_tokens;
 
+    // #6912: keep whichever token key transformRequest already set (o1/o3/o4/gpt-5 use
+    // max_completion_tokens) instead of re-introducing max_tokens alongside it.
+    const tokenKey =
+      body.max_completion_tokens !== undefined ? "max_completion_tokens" : "max_tokens";
+
     if (typeof current !== "number" || current <= 0) {
-      body.max_tokens = target;
+      body[tokenKey] = target;
     } else if (current < MIN_TOKENS && current < maxOutput) {
-      body.max_tokens = MIN_TOKENS;
+      body[tokenKey] = MIN_TOKENS;
     }
     return body;
   }

tests/unit/antigravity-server-side-tools-6914.test.ts

@@ -0,0 +1,43 @@
+// #6914: server-side tool invocations must be requested from Antigravity whenever the
+// caller sends tools, and must NOT be forced when the request carries none. Lives in
+// its own file (not executor-antigravity.test.ts) because that suite is frozen at the
+// test-file-size cap.
+import { test } from "node:test";
+import assert from "node:assert/strict";
+
+import { AntigravityExecutor } from "../../open-sse/executors/antigravity.ts";
+
+test("AntigravityExecutor.transformRequest includes includeServerSideToolInvocations when tools are present", async () => {
+  const executor = new AntigravityExecutor();
+  const body = {
+    request: {
+      contents: [{ role: "user", parts: [{ text: "Hello" }] }],
+      tools: [{ functionDeclarations: [{ name: "search" }] }],
+    },
+  };
+
+  const result = await executor.transformRequest("antigravity/gemini-3.1-pro", body, true, {
+    projectId: "project-1",
+  });
+
+  if (result instanceof Response) throw new Error("Unexpected Response from transformRequest");
+  assert.deepEqual(result.request.toolConfig, {
+    functionCallingConfig: { mode: "VALIDATED", includeServerSideToolInvocations: true },
+  });
+});
+
+test("AntigravityExecutor.transformRequest does not include includeServerSideToolInvocations when no tools", async () => {
+  const executor = new AntigravityExecutor();
+  const body = {
+    request: {
+      contents: [{ role: "user", parts: [{ text: "Hello" }] }],
+    },
+  };
+
+  const result = await executor.transformRequest("antigravity/gemini-3.1-pro", body, true, {
+    projectId: "project-1",
+  });
+
+  if (result instanceof Response) throw new Error("Unexpected Response from transformRequest");
+  assert.equal(result.request.toolConfig, undefined);
+});

tests/unit/clinepass-thinking-budget.test.ts

@@ -3,8 +3,9 @@ import assert from "node:assert/strict";
 
 import { DefaultExecutor } from "../../open-sse/executors/default.ts";
 
-// DefaultExecutor.ensureThinkingBudget — clinepass-gated max_tokens floor for
+// DefaultExecutor.ensureThinkingBudget — max_tokens floor for
 // reasoning models (prevents empty content when the budget is undersized).
+// Previously gated to clinepass only; now applies to all providers (#6912).
 
 test("bumps undersized max_tokens to 4096 for a clinepass reasoning model", () => {
   const executor = new DefaultExecutor("clinepass");
@@ -64,14 +65,17 @@ test("no-op for a non-reasoning clinepass model", () => {
   assert.equal(body.max_tokens, 100);
 });
 
-test("no-op for a non-clinepass provider (gate)", () => {
-  const executor = new DefaultExecutor("openrouter");
+test("bumps undersized max_tokens for a non-clinepass reasoning provider (gate removed, #6912)", () => {
+  // Issue #6912: ensureThinkingBudget was gated to clinepass only.
+  // Now it applies to all providers. Use nvidia (non-clinepass) which has
+  // deepseek-ai/deepseek-v4-pro with supportsReasoning in the registry.
+  const executor = new DefaultExecutor("nvidia");
   const body = {
-    model: "cline-pass/deepseek-v4-pro",
+    model: "deepseek-ai/deepseek-v4-pro",
     reasoning_effort: "high",
     max_tokens: 100,
   } as Record<string, unknown>;
 
-  executor.ensureThinkingBudget(body, "cline-pass/deepseek-v4-pro");
-  assert.equal(body.max_tokens, 100);
+  executor.ensureThinkingBudget(body, "deepseek-ai/deepseek-v4-pro");
+  assert.equal(body.max_tokens, 4096);
 });

tests/unit/executor-antigravity.test.ts

@@ -149,7 +149,7 @@ test("AntigravityExecutor.transformRequest normalizes model, project and content
   assert.equal(generationConfig.topK, 40);
   assert.equal(generationConfig.topP, 1.0);
   assert.deepEqual(result.request.toolConfig, {
-    functionCallingConfig: { mode: "VALIDATED" },
+    functionCallingConfig: { mode: "VALIDATED", includeServerSideToolInvocations: true },
   });
   assert.deepEqual(result.request.contents[0].parts, [{ text: "keep me" }]);
   assert.equal(result.request.contents[1].role, "user");