fix(gemini-cli): raise output floor for thinking and add validated toolConfig (#2486)

Gemini CLI requests with small max_tokens spend the whole output budget on
thoughts after reasoning_effort maps to thinkingConfig, returning blank
content or finish=length. Raise maxOutputTokens floors per thinking level/
budget (clamped to caps.maxOutput). Also emit toolConfig
functionCallingConfig.mode=VALIDATED for Gemini CLI tool requests to avoid
MALFORMED_FUNCTION_CALL.

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
DOMANHDUC
2026-07-10 11:27:41 +07:00
committed by decolua
parent 0d4d4bc261
commit 7610f28f42
4 changed files with 107 additions and 12 deletions

View File

@@ -148,18 +148,57 @@ function toKimiReasoningEffort(cfg) {
return null;
}
const GEMINI_LEVEL_OUTPUT_FLOOR = {
minimal: 4096,
low: 8192,
medium: 16384,
high: 65535,
};
function geminiBudgetOutputFloor(budget) {
if (budget === -1) return 32768;
if (!Number.isFinite(budget)) return 32768;
if (budget <= 1024) return 8192;
if (budget <= 8192) return 16384;
if (budget <= 24576) return 32768;
return 65535;
}
function geminiLevelOutputFloor(level) {
return GEMINI_LEVEL_OUTPUT_FLOOR[level] || GEMINI_LEVEL_OUTPUT_FLOOR.high;
}
// Gemini nests thinkingConfig under generationConfig. gemini-cli / antigravity wrap
// the whole request in a { request: { generationConfig } } envelope — target the
// envelope's generationConfig when present, else the top-level one.
function getGeminiGenerationConfig(body) {
if (body.request && typeof body.request === "object") {
if (!body.request.generationConfig || typeof body.request.generationConfig !== "object") {
body.request.generationConfig = {};
}
return body.request.generationConfig;
}
if (!body.generationConfig || typeof body.generationConfig !== "object") {
body.generationConfig = {};
}
return body.generationConfig;
}
function setGeminiThinking(body, tc) {
const gc = body.request?.generationConfig
? body.request.generationConfig
: (body.generationConfig && typeof body.generationConfig === "object"
? body.generationConfig
: (body.generationConfig = {}));
const gc = getGeminiGenerationConfig(body);
gc.thinkingConfig = tc;
}
function ensureGeminiOutputFloor(body, floor, caps) {
const cap = Number.isFinite(caps?.maxOutput) ? caps.maxOutput : floor;
const target = Math.min(floor, cap);
const gc = getGeminiGenerationConfig(body);
const current = Number(gc.maxOutputTokens);
if (!Number.isFinite(current) || current < target) {
gc.maxOutputTokens = target;
}
}
// Strip every known thinking field from a body (used before re-applying / when unsupported).
function stripAll(body) {
delete body.thinking;
@@ -203,12 +242,14 @@ function applyFormat(fmt, body, cfg, caps) {
case "gemini-level": {
const level = none ? "minimal" : toGeminiThinkingLevel(eff);
setGeminiThinking(body, { thinkingLevel: level, includeThoughts: level !== "minimal" });
ensureGeminiOutputFloor(body, geminiLevelOutputFloor(level), caps);
break;
}
case "gemini-budget": {
if (none && canDisable) { setGeminiThinking(body, { thinkingBudget: 0, includeThoughts: false }); break; }
const budget = toBudget(eff, caps.thinkingRange);
setGeminiThinking(body, { thinkingBudget: budget ?? -1, includeThoughts: true });
ensureGeminiOutputFloor(body, geminiBudgetOutputFloor(budget ?? -1), caps);
break;
}
case "zai": {

View File

@@ -281,18 +281,17 @@ function wrapInCloudCodeEnvelope(model, geminiCLI, credentials = null, isAntigra
// Antigravity specific fields
if (isAntigravity) {
envelope.requestType = "agent";
// Add toolConfig for Antigravity
if (geminiCLI.tools?.length > 0) {
envelope.request.toolConfig = {
functionCallingConfig: { mode: "VALIDATED" }
};
}
} else {
// Keep safetySettings for Gemini CLI
envelope.request.safetySettings = geminiCLI.safetySettings;
}
if (geminiCLI.tools?.length > 0) {
envelope.request.toolConfig = {
functionCallingConfig: { mode: "VALIDATED" }
};
}
return envelope;
}