From ee9b49838033de3223fd07270b0f261e79cf8f7f Mon Sep 17 00:00:00 2001 From: Luke Foster <116062686+LukeFost@users.noreply.github.com> Date: Sat, 20 Dec 2025 22:10:47 -0600 Subject: [PATCH] Add Gemini 3 preview models to google-gemini-cli provider - Add gemini-3-pro-preview and gemini-3-flash-preview to Cloud Code Assist - Handle thinkingLevel config for Gemini 3 (vs thinkingBudget for Gemini 2.x) - Gemini 3 Pro: LOW/HIGH levels only - Gemini 3 Flash: all four levels (MINIMAL/LOW/MEDIUM/HIGH) --- packages/ai/scripts/generate-models.ts | 24 ++++++++++++ packages/ai/src/models.generated.ts | 36 ++++++++++++++++- .../ai/src/providers/google-gemini-cli.ts | 8 +++- packages/ai/src/stream.ts | 39 ++++++++++++++++++- packages/coding-agent/CHANGELOG.md | 4 ++ 5 files changed, 107 insertions(+), 4 deletions(-) diff --git a/packages/ai/scripts/generate-models.ts b/packages/ai/scripts/generate-models.ts index 8d17fdcf..f72db4bd 100644 --- a/packages/ai/scripts/generate-models.ts +++ b/packages/ai/scripts/generate-models.ts @@ -522,6 +522,30 @@ async function generateModels() { contextWindow: 1048576, maxTokens: 8192, }, + { + id: "gemini-3-pro-preview", + name: "Gemini 3 Pro Preview (Cloud Code Assist)", + api: "google-gemini-cli", + provider: "google-gemini-cli", + baseUrl: CLOUD_CODE_ASSIST_ENDPOINT, + reasoning: true, + input: ["text", "image"], + cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 }, + contextWindow: 1048576, + maxTokens: 65535, + }, + { + id: "gemini-3-flash-preview", + name: "Gemini 3 Flash Preview (Cloud Code Assist)", + api: "google-gemini-cli", + provider: "google-gemini-cli", + baseUrl: CLOUD_CODE_ASSIST_ENDPOINT, + reasoning: true, + input: ["text", "image"], + cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 }, + contextWindow: 1048576, + maxTokens: 65535, + }, ]; allModels.push(...cloudCodeAssistModels); diff --git a/packages/ai/src/models.generated.ts b/packages/ai/src/models.generated.ts index 00f631dc..c51c2b31 100644 --- a/packages/ai/src/models.generated.ts +++ b/packages/ai/src/models.generated.ts @@ -3225,7 +3225,7 @@ export const MODELS = { cost: { input: 0.24, output: 0.38, - cacheRead: 0.11, + cacheRead: 0.02, cacheWrite: 0, }, contextWindow: 163840, @@ -6854,6 +6854,40 @@ export const MODELS = { contextWindow: 1048576, maxTokens: 8192, } satisfies Model<"google-gemini-cli">, + "gemini-3-pro-preview": { + id: "gemini-3-pro-preview", + name: "Gemini 3 Pro Preview (Cloud Code Assist)", + api: "google-gemini-cli", + provider: "google-gemini-cli", + baseUrl: "https://cloudcode-pa.googleapis.com", + reasoning: true, + input: ["text", "image"], + cost: { + input: 0, + output: 0, + cacheRead: 0, + cacheWrite: 0, + }, + contextWindow: 1048576, + maxTokens: 65535, + } satisfies Model<"google-gemini-cli">, + "gemini-3-flash-preview": { + id: "gemini-3-flash-preview", + name: "Gemini 3 Flash Preview (Cloud Code Assist)", + api: "google-gemini-cli", + provider: "google-gemini-cli", + baseUrl: "https://cloudcode-pa.googleapis.com", + reasoning: true, + input: ["text", "image"], + cost: { + input: 0, + output: 0, + cacheRead: 0, + cacheWrite: 0, + }, + contextWindow: 1048576, + maxTokens: 65535, + } satisfies Model<"google-gemini-cli">, }, "google-antigravity": { "gemini-3-pro-high": { diff --git a/packages/ai/src/providers/google-gemini-cli.ts b/packages/ai/src/providers/google-gemini-cli.ts index 59ee332e..b9c43109 100644 --- a/packages/ai/src/providers/google-gemini-cli.ts +++ b/packages/ai/src/providers/google-gemini-cli.ts @@ -4,7 +4,7 @@ * Uses the Cloud Code Assist API endpoint to access Gemini and Claude models. */ -import type { Content, ThinkingConfig } from "@google/genai"; +import type { Content, ThinkingConfig, ThinkingLevel } from "@google/genai"; import { calculateCost } from "../models.js"; import type { Api, @@ -26,6 +26,7 @@ export interface GoogleGeminiCliOptions extends StreamOptions { thinking?: { enabled: boolean; budgetTokens?: number; + level?: ThinkingLevel; // For Gemini 3 models }; projectId?: string; } @@ -424,7 +425,10 @@ function buildRequest( generationConfig.thinkingConfig = { includeThoughts: true, }; - if (options.thinking.budgetTokens !== undefined) { + // Gemini 3 models use thinkingLevel, older models use thinkingBudget + if (options.thinking.level !== undefined) { + generationConfig.thinkingConfig.thinkingLevel = options.thinking.level; + } else if (options.thinking.budgetTokens !== undefined) { generationConfig.thinkingConfig.thinkingBudget = options.thinking.budgetTokens; } } diff --git a/packages/ai/src/stream.ts b/packages/ai/src/stream.ts index ba00c29f..f2c15050 100644 --- a/packages/ai/src/stream.ts +++ b/packages/ai/src/stream.ts @@ -237,12 +237,24 @@ function mapOptionsForApi( } case "google-gemini-cli": { - // Cloud Code Assist uses thinking budget tokens like Gemini 2.5 if (!options?.reasoning) { return { ...base, thinking: { enabled: false } } satisfies GoogleGeminiCliOptions; } const effort = clampReasoning(options.reasoning)!; + + // Gemini 3 models use thinkingLevel instead of thinkingBudget + if (model.id.includes("3-pro") || model.id.includes("3-flash")) { + return { + ...base, + thinking: { + enabled: true, + level: getGeminiCliThinkingLevel(effort, model.id), + }, + } satisfies GoogleGeminiCliOptions; + } + + // Gemini 2.x models use thinkingBudget const budgets: Record = { minimal: 1024, low: 2048, @@ -304,6 +316,31 @@ function getGemini3ThinkingLevel(effort: ClampedReasoningEffort, model: Model<"g } } +function getGeminiCliThinkingLevel(effort: ClampedReasoningEffort, modelId: string): ThinkingLevel { + if (modelId.includes("3-pro")) { + // Gemini 3 Pro only supports LOW/HIGH (for now) + switch (effort) { + case "minimal": + case "low": + return ThinkingLevel.LOW; + case "medium": + case "high": + return ThinkingLevel.HIGH; + } + } + // Gemini 3 Flash supports all four levels + switch (effort) { + case "minimal": + return ThinkingLevel.MINIMAL; + case "low": + return ThinkingLevel.LOW; + case "medium": + return ThinkingLevel.MEDIUM; + case "high": + return ThinkingLevel.HIGH; + } +} + function getGoogleBudget(model: Model<"google-generative-ai">, effort: ClampedReasoningEffort): number { // See https://ai.google.dev/gemini-api/docs/thinking#set-budget if (model.id.includes("2.5-pro")) { diff --git a/packages/coding-agent/CHANGELOG.md b/packages/coding-agent/CHANGELOG.md index 47b461ea..637e44cf 100644 --- a/packages/coding-agent/CHANGELOG.md +++ b/packages/coding-agent/CHANGELOG.md @@ -2,6 +2,10 @@ ## [Unreleased] +### Added + +- **Gemini 3 preview models**: Added `gemini-3-pro-preview` and `gemini-3-flash-preview` to the google-gemini-cli provider. + ## [0.25.2] - 2025-12-21 ### Fixed