Merge remote-tracking branch 'origin/main' into cs/codex-native-web-search-spec

# Conflicts:
#	src/agents/pi-embedded-runner/extra-params.ts
#	src/agents/pi-embedded-runner/openai-stream-wrappers.ts
This commit is contained in:
Christof Salis
2026-03-17 06:25:55 +01:00
926 changed files with 21587 additions and 13022 deletions

View File

@@ -1,10 +1,10 @@
import { fetchGeminiUsage } from "../../src/infra/provider-usage.fetch.js";
import { buildOauthProviderAuthResult } from "../../src/plugin-sdk/provider-auth-result.js";
import { buildOauthProviderAuthResult } from "openclaw/plugin-sdk/core";
import type {
OpenClawPluginApi,
ProviderAuthContext,
ProviderFetchUsageSnapshotContext,
} from "../../src/plugins/types.js";
} from "openclaw/plugin-sdk/core";
import { fetchGeminiUsage } from "openclaw/plugin-sdk/provider-usage";
import { loginGeminiCliOAuth } from "./oauth.js";
import { isModernGoogleModel, resolveGoogle31ForwardCompatModel } from "./provider-models.js";

View File

@@ -1,16 +1,16 @@
import { emptyPluginConfigSchema, type OpenClawPluginApi } from "openclaw/plugin-sdk/core";
import { createProviderApiKeyAuthMethod } from "openclaw/plugin-sdk/provider-auth";
import {
GOOGLE_GEMINI_DEFAULT_MODEL,
applyGoogleGeminiModelDefault,
} from "openclaw/plugin-sdk/provider-models";
import {
createPluginBackedWebSearchProvider,
getScopedCredentialValue,
setScopedCredentialValue,
} from "../../src/agents/tools/web-search-plugin-factory.js";
import {
GOOGLE_GEMINI_DEFAULT_MODEL,
applyGoogleGeminiModelDefault,
} from "../../src/commands/google-gemini-model-default.js";
import { emptyPluginConfigSchema } from "../../src/plugins/config-schema.js";
import { createProviderApiKeyAuthMethod } from "../../src/plugins/provider-api-key-auth.js";
import type { OpenClawPluginApi } from "../../src/plugins/types.js";
} from "openclaw/plugin-sdk/provider-web-search";
import { registerGoogleGeminiCliProvider } from "./gemini-cli-provider.js";
import { googleMediaUnderstandingProvider } from "./media-understanding-provider.js";
import { isModernGoogleModel, resolveGoogle31ForwardCompatModel } from "./provider-models.js";
const googlePlugin = {
@@ -51,6 +51,7 @@ const googlePlugin = {
isModernModelRef: ({ modelId }) => isModernGoogleModel(modelId),
});
registerGoogleGeminiCliProvider(api);
api.registerMediaUnderstandingProvider(googleMediaUnderstandingProvider);
api.registerWebSearchProvider(
createPluginBackedWebSearchProvider({
id: "gemini",

View File

@@ -0,0 +1,149 @@
import { normalizeGoogleModelId, parseGeminiAuth } from "openclaw/plugin-sdk/google";
import {
assertOkOrThrowHttpError,
describeImageWithModel,
describeImagesWithModel,
normalizeBaseUrl,
postJsonRequest,
type AudioTranscriptionRequest,
type AudioTranscriptionResult,
type MediaUnderstandingProvider,
type VideoDescriptionRequest,
type VideoDescriptionResult,
} from "openclaw/plugin-sdk/media-understanding";
export const DEFAULT_GOOGLE_AUDIO_BASE_URL = "https://generativelanguage.googleapis.com/v1beta";
export const DEFAULT_GOOGLE_VIDEO_BASE_URL = "https://generativelanguage.googleapis.com/v1beta";
const DEFAULT_GOOGLE_AUDIO_MODEL = "gemini-3-flash-preview";
const DEFAULT_GOOGLE_VIDEO_MODEL = "gemini-3-flash-preview";
const DEFAULT_GOOGLE_AUDIO_PROMPT = "Transcribe the audio.";
const DEFAULT_GOOGLE_VIDEO_PROMPT = "Describe the video.";
async function generateGeminiInlineDataText(params: {
buffer: Buffer;
mime?: string;
apiKey: string;
baseUrl?: string;
headers?: Record<string, string>;
model?: string;
prompt?: string;
timeoutMs: number;
fetchFn?: typeof fetch;
defaultBaseUrl: string;
defaultModel: string;
defaultPrompt: string;
defaultMime: string;
httpErrorLabel: string;
missingTextError: string;
}): Promise<{ text: string; model: string }> {
const fetchFn = params.fetchFn ?? fetch;
const baseUrl = normalizeBaseUrl(params.baseUrl, params.defaultBaseUrl);
const allowPrivate = Boolean(params.baseUrl?.trim());
const model = (() => {
const trimmed = params.model?.trim();
if (!trimmed) {
return params.defaultModel;
}
return normalizeGoogleModelId(trimmed);
})();
const url = `${baseUrl}/models/${model}:generateContent`;
const authHeaders = parseGeminiAuth(params.apiKey);
const headers = new Headers(params.headers);
for (const [key, value] of Object.entries(authHeaders.headers)) {
if (!headers.has(key)) {
headers.set(key, value);
}
}
const prompt = (() => {
const trimmed = params.prompt?.trim();
return trimmed || params.defaultPrompt;
})();
const body = {
contents: [
{
role: "user",
parts: [
{ text: prompt },
{
inline_data: {
mime_type: params.mime ?? params.defaultMime,
data: params.buffer.toString("base64"),
},
},
],
},
],
};
const { response: res, release } = await postJsonRequest({
url,
headers,
body,
timeoutMs: params.timeoutMs,
fetchFn,
allowPrivateNetwork: allowPrivate,
});
try {
await assertOkOrThrowHttpError(res, params.httpErrorLabel);
const payload = (await res.json()) as {
candidates?: Array<{
content?: { parts?: Array<{ text?: string }> };
}>;
};
const parts = payload.candidates?.[0]?.content?.parts ?? [];
const text = parts
.map((part) => part?.text?.trim())
.filter(Boolean)
.join("\n");
if (!text) {
throw new Error(params.missingTextError);
}
return { text, model };
} finally {
await release();
}
}
export async function transcribeGeminiAudio(
params: AudioTranscriptionRequest,
): Promise<AudioTranscriptionResult> {
const { text, model } = await generateGeminiInlineDataText({
...params,
defaultBaseUrl: DEFAULT_GOOGLE_AUDIO_BASE_URL,
defaultModel: DEFAULT_GOOGLE_AUDIO_MODEL,
defaultPrompt: DEFAULT_GOOGLE_AUDIO_PROMPT,
defaultMime: "audio/wav",
httpErrorLabel: "Audio transcription failed",
missingTextError: "Audio transcription response missing text",
});
return { text, model };
}
export async function describeGeminiVideo(
params: VideoDescriptionRequest,
): Promise<VideoDescriptionResult> {
const { text, model } = await generateGeminiInlineDataText({
...params,
defaultBaseUrl: DEFAULT_GOOGLE_VIDEO_BASE_URL,
defaultModel: DEFAULT_GOOGLE_VIDEO_MODEL,
defaultPrompt: DEFAULT_GOOGLE_VIDEO_PROMPT,
defaultMime: "video/mp4",
httpErrorLabel: "Video description failed",
missingTextError: "Video description response missing text",
});
return { text, model };
}
export const googleMediaUnderstandingProvider: MediaUnderstandingProvider = {
id: "google",
capabilities: ["image", "audio", "video"],
describeImage: describeImageWithModel,
describeImages: describeImagesWithModel,
transcribeAudio: transcribeGeminiAudio,
describeVideo: describeGeminiVideo,
};

View File

@@ -1,6 +1,6 @@
import { createHash, randomBytes } from "node:crypto";
import { createServer } from "node:http";
import { isWSL2Sync } from "../../src/infra/wsl.js";
import { isWSL2Sync } from "openclaw/plugin-sdk/infra-runtime";
import { resolveOAuthClientConfig } from "./oauth.credentials.js";
import { AUTH_URL, REDIRECT_URI, SCOPES } from "./oauth.shared.js";

View File

@@ -1,4 +1,4 @@
import { fetchWithSsrFGuard } from "../../src/infra/net/fetch-guard.js";
import { fetchWithSsrFGuard } from "openclaw/plugin-sdk/infra-runtime";
import { DEFAULT_FETCH_TIMEOUT_MS } from "./oauth.shared.js";
export async function fetchWithTimeout(

View File

@@ -1,39 +1,14 @@
import { normalizeModelCompat } from "../../src/agents/model-compat.js";
import type {
ProviderResolveDynamicModelContext,
ProviderRuntimeModel,
} from "../../src/plugins/types.js";
} from "openclaw/plugin-sdk/core";
import { cloneFirstTemplateModel } from "../../src/plugins/provider-model-helpers.js";
const GEMINI_3_1_PRO_PREFIX = "gemini-3.1-pro";
const GEMINI_3_1_FLASH_PREFIX = "gemini-3.1-flash";
const GEMINI_3_1_PRO_TEMPLATE_IDS = ["gemini-3-pro-preview"] as const;
const GEMINI_3_1_FLASH_TEMPLATE_IDS = ["gemini-3-flash-preview"] as const;
function cloneFirstTemplateModel(params: {
providerId: string;
modelId: string;
templateIds: readonly string[];
ctx: ProviderResolveDynamicModelContext;
}): ProviderRuntimeModel | undefined {
const trimmedModelId = params.modelId.trim();
for (const templateId of [...new Set(params.templateIds)].filter(Boolean)) {
const template = params.ctx.modelRegistry.find(
params.providerId,
templateId,
) as ProviderRuntimeModel | null;
if (!template) {
continue;
}
return normalizeModelCompat({
...template,
id: trimmedModelId,
name: trimmedModelId,
reasoning: true,
} as ProviderRuntimeModel);
}
return undefined;
}
export function resolveGoogle31ForwardCompatModel(params: {
providerId: string;
ctx: ProviderResolveDynamicModelContext;
@@ -55,6 +30,7 @@ export function resolveGoogle31ForwardCompatModel(params: {
modelId: trimmed,
templateIds,
ctx: params.ctx,
patch: { reasoning: true },
});
}