@openclaw/deepinfra-provider 2026.9.1-beta.1 → 2026.9.2

package/dist/api.js

@@ -1,7 +1,7 @@
-import { DEEPINFRA_DEFAULT_MODEL_REF } from "./provider-models.js";
 import { deepinfraEmbeddingProviderAdapter } from "./embedding-adapter.js";
 import { buildDeepInfraImageGenerationProvider } from "./image-generation-provider.js";
 import { deepinfraMediaUnderstandingProvider } from "./media-understanding-provider.js";
+import { DEEPINFRA_DEFAULT_MODEL_REF } from "./provider-models.js";
 import { applyDeepInfraConfig } from "./onboard.js";
 import { buildDeepInfraProvider, buildStaticDeepInfraProvider } from "./provider-catalog.js";
 import { buildDeepInfraSpeechProvider } from "./speech-provider.js";

package/dist/capability-catalog.js

@@ -0,0 +1,5 @@
+import { buildDeepInfraSpeechProvider } from "./speech-provider.js";
+//#region extensions/deepinfra/capability-catalog.ts
+var capability_catalog_default = { speechProviders: [buildDeepInfraSpeechProvider()] };
+//#endregion
+export { capability_catalog_default as default };

package/dist/doctor-contract-api.js

@@ -1,4 +1,4 @@
-import { t as modelCatalog } from "./openclaw.plugin-DomWf43f.js";
+import { t as modelCatalog } from "./openclaw.plugin-BRaVociw.js";
 import { asObjectRecord } from "openclaw/plugin-sdk/runtime-doctor-migrations";
 //#region extensions/deepinfra/doctor-contract-api.ts
 const PROVIDER_PATH = "models.providers.deepinfra";

package/dist/embedding-adapter.js

@@ -1,4 +1,4 @@
-import { DEEPINFRA_BASE_URL } from "./provider-models.js";
+import { DEEPINFRA_BASE_URL } from "./media-models.js";
 import { DEFAULT_DEEPINFRA_EMBEDDING_MODEL, createDeepInfraEmbeddingProvider } from "./embedding-provider.js";
 import { embeddingProviderOwnsDestination, sanitizeEmbeddingCacheHeaders } from "openclaw/plugin-sdk/memory-core-host-engine-embeddings";
 //#region extensions/deepinfra/embedding-adapter.ts
@@ -8,19 +8,6 @@ const EXCLUDED_EMBEDDING_HEADERS = [
 	"x-api-key",
 	"api-key"
 ];
-function textFromEmbeddingInput(input) {
-	return typeof input === "string" ? input : input.text;
-}
-function adaptMemoryEmbeddingProvider(provider) {
-	return {
-		id: provider.id,
-		model: provider.model,
-		...typeof provider.maxInputTokens === "number" ? { maxInputTokens: provider.maxInputTokens } : {},
-		embed: async (input, options) => await provider.embedQuery(textFromEmbeddingInput(input), { signal: options?.signal }),
-		embedBatch: async (inputs, options) => await provider.embedBatch(inputs.map(textFromEmbeddingInput), { signal: options?.signal }),
-		...provider.close ? { close: async () => await provider.close?.() } : {}
-	};
-}
 function buildDeepInfraEmbeddingAdapter(options) {
 	const defaultModel = options?.embedModels?.[0]?.id ?? DEFAULT_DEEPINFRA_EMBEDDING_MODEL;
 	return {
@@ -29,13 +16,9 @@ function buildDeepInfraEmbeddingAdapter(options) {
 		transport: "remote",
 		authProviderId: "deepinfra",
 		create: async (createOptions) => {
-			const { dimensions, ...memoryOptions } = createOptions;
 			const { provider, client } = await createDeepInfraEmbeddingProvider({
-				...memoryOptions,
+				...createOptions,
 				provider: "deepinfra",
-				fallback: "none",
-				outputDimensionality: dimensions,
-				taskType: createOptions.taskType,
 				defaultModel
 			});
 			const headers = sanitizeEmbeddingCacheHeaders(client.headers, EXCLUDED_EMBEDDING_HEADERS);
@@ -44,7 +27,7 @@ function buildDeepInfraEmbeddingAdapter(options) {
 				providerBaseUrl: DEEPINFRA_BASE_URL
 			});
 			return {
-				provider: provider ? adaptMemoryEmbeddingProvider(provider) : null,
+				provider,
 				runtime: {
 					id: "deepinfra",
 					cacheKeyData: {

package/dist/embedding-provider.js

@@ -1,5 +1,4 @@
-import { DEEPINFRA_BASE_URL } from "./provider-models.js";
-import { DEEPINFRA_EMBED_FALLBACK_MODELS, normalizeDeepInfraModelRef } from "./media-models.js";
+import { DEEPINFRA_BASE_URL, DEEPINFRA_EMBED_FALLBACK_MODELS, normalizeDeepInfraModelRef } from "./media-models.js";
 import { createRemoteEmbeddingProvider, resolveRemoteEmbeddingClient } from "openclaw/plugin-sdk/memory-core-host-engine-embeddings";
 //#region extensions/deepinfra/embedding-provider.ts
 const DEFAULT_DEEPINFRA_EMBEDDING_MODEL = DEEPINFRA_EMBED_FALLBACK_MODELS[0];
@@ -18,7 +17,8 @@ async function createDeepInfraEmbeddingProvider(options) {
 		provider: createRemoteEmbeddingProvider({
 			id: "deepinfra",
 			client,
-			errorPrefix: "DeepInfra embeddings API error"
+			errorPrefix: "DeepInfra embeddings API error",
+			batchQueryInputs: true
 		}),
 		client
 	};

package/dist/image-generation-provider.js

@@ -1,6 +1,6 @@
-import { DEEPINFRA_BASE_URL } from "./provider-models.js";
-import { DEEPINFRA_IMAGE_FALLBACK_MODELS, normalizeDeepInfraBaseUrl, normalizeDeepInfraModelRef } from "./media-models.js";
+import { DEEPINFRA_BASE_URL, DEEPINFRA_IMAGE_FALLBACK_MODELS, normalizeDeepInfraBaseUrl, normalizeDeepInfraModelRef } from "./media-models.js";
 import { normalizeOptionalString } from "openclaw/plugin-sdk/string-coerce-runtime";
+import { bufferToBlobPart } from "openclaw/plugin-sdk/blob-runtime";
 import { createOpenAiCompatibleImageGenerationProvider, imageSourceUploadFileName } from "openclaw/plugin-sdk/image-generation";
 //#region extensions/deepinfra/image-generation-provider.ts
 const DEEPINFRA_IMAGE_SIZES = [
@@ -12,10 +12,11 @@ const DEEPINFRA_IMAGE_SIZES = [
 const MAX_DEEPINFRA_INPUT_IMAGES = 1;
 function buildDeepInfraImageGenerationProvider(options) {
 	const ids = options?.imageGenModels && options.imageGenModels.length > 0 ? options.imageGenModels.map((model) => model.id) : [...DEEPINFRA_IMAGE_FALLBACK_MODELS];
+	const defaultModel = ids[0] ?? DEEPINFRA_IMAGE_FALLBACK_MODELS[0];
 	return createOpenAiCompatibleImageGenerationProvider({
 		id: "deepinfra",
 		label: "DeepInfra",
-		defaultModel: ids[0] ?? DEEPINFRA_IMAGE_FALLBACK_MODELS[0],
+		defaultModel,
 		models: ids,
 		capabilities: {
 			generate: {
@@ -60,7 +61,7 @@ function buildDeepInfraImageGenerationProvider(options) {
 			form.set("size", normalizeOptionalString(req.size) ?? "1024x1024");
 			form.set("response_format", "b64_json");
 			const mimeType = normalizeOptionalString(image.mimeType) ?? "image/png";
-			form.append("image", new Blob([new Uint8Array(image.buffer)], { type: mimeType }), imageSourceUploadFileName({
+			form.append("image", new Blob([bufferToBlobPart(image.buffer)], { type: mimeType }), imageSourceUploadFileName({
 				image,
 				index: 0
 			}));

package/dist/index.js

@@ -1,8 +1,8 @@
 import { createDeepInfraAnthropicCacheWrapper } from "./cache-wrapper.js";
-import { DEEPINFRA_DEFAULT_MODEL_REF, discoverDeepInfraModels, getDeepInfraSurfaceFallbackCatalog, hasDeepInfraApiKey } from "./provider-models.js";
 import { buildDeepInfraEmbeddingAdapter } from "./embedding-adapter.js";
 import { buildDeepInfraImageGenerationProvider } from "./image-generation-provider.js";
 import { buildDeepInfraMediaUnderstandingProvider } from "./media-understanding-provider.js";
+import { DEEPINFRA_DEFAULT_MODEL_REF, discoverDeepInfraModels, getDeepInfraSurfaceFallbackCatalog, hasDeepInfraApiKey } from "./provider-models.js";
 import { applyDeepInfraConfig } from "./onboard.js";
 import { buildDeepInfraApiKeyCatalog, buildStaticDeepInfraProvider } from "./provider-catalog.js";
 import { buildDeepInfraSpeechProvider } from "./speech-provider.js";

package/dist/media-models.js

@@ -1,6 +1,7 @@
-import { DEEPINFRA_BASE_URL } from "./provider-models.js";
+import { t as modelCatalog } from "./openclaw.plugin-BRaVociw.js";
 import { normalizeOptionalString } from "openclaw/plugin-sdk/string-coerce-runtime";
 //#region extensions/deepinfra/media-models.ts
+const DEEPINFRA_BASE_URL = modelCatalog.providers.deepinfra.baseUrl;
 const DEFAULT_DEEPINFRA_IMAGE_SIZE = "1024x1024";
 const DEFAULT_DEEPINFRA_TTS_VOICE = "af_bella";
 const DEEPINFRA_VIDEO_ASPECT_RATIOS = [
@@ -17,12 +18,33 @@ const DEEPINFRA_IMAGE_FALLBACK_MODELS = [
 	"Qwen/Qwen-Image-Max",
 	"stabilityai/sdxl-turbo"
 ];
-const DEEPINFRA_TTS_FALLBACK_MODELS = [
-	"hexgrad/Kokoro-82M",
-	"Qwen/Qwen3-TTS",
-	"ResembleAI/chatterbox-turbo",
-	"sesame/csm-1b"
+const DEEPINFRA_TTS_FALLBACK_CATALOG = [
+	{
+		id: "hexgrad/Kokoro-82M",
+		name: "hexgrad/Kokoro-82M",
+		tags: ["tts"],
+		pricing: { input_characters: .65 }
+	},
+	{
+		id: "Qwen/Qwen3-TTS",
+		name: "Qwen/Qwen3-TTS",
+		tags: ["tts"],
+		pricing: { input_characters: .65 }
+	},
+	{
+		id: "ResembleAI/chatterbox-turbo",
+		name: "ResembleAI/chatterbox-turbo",
+		tags: ["tts"],
+		pricing: { input_characters: 1 }
+	},
+	{
+		id: "sesame/csm-1b",
+		name: "sesame/csm-1b",
+		tags: ["tts"],
+		pricing: { input_characters: 7 }
+	}
 ];
+const DEEPINFRA_TTS_FALLBACK_MODELS = DEEPINFRA_TTS_FALLBACK_CATALOG.map(({ id }) => id);
 const DEEPINFRA_VIDEO_FALLBACK_MODELS = [
 	"Pixverse/Pixverse-T2V",
 	"Pixverse/Pixverse-T2V-HD",
@@ -40,4 +62,4 @@ function normalizeDeepInfraBaseUrl(value, fallback = DEEPINFRA_BASE_URL) {
 	return (normalizeOptionalString(value) ?? fallback).replace(/\/+$/u, "");
 }
 //#endregion
-export { DEEPINFRA_BASE_URL, DEEPINFRA_EMBED_FALLBACK_MODELS, DEEPINFRA_IMAGE_FALLBACK_MODELS, DEEPINFRA_STT_FALLBACK_MODELS, DEEPINFRA_TTS_FALLBACK_MODELS, DEEPINFRA_VIDEO_ASPECT_RATIOS, DEEPINFRA_VIDEO_DURATIONS, DEEPINFRA_VIDEO_FALLBACK_MODELS, DEEPINFRA_VLM_FALLBACK_MODELS, DEFAULT_DEEPINFRA_IMAGE_SIZE, DEFAULT_DEEPINFRA_TTS_VOICE, normalizeDeepInfraBaseUrl, normalizeDeepInfraModelRef };
+export { DEEPINFRA_BASE_URL, DEEPINFRA_EMBED_FALLBACK_MODELS, DEEPINFRA_IMAGE_FALLBACK_MODELS, DEEPINFRA_STT_FALLBACK_MODELS, DEEPINFRA_TTS_FALLBACK_CATALOG, DEEPINFRA_TTS_FALLBACK_MODELS, DEEPINFRA_VIDEO_ASPECT_RATIOS, DEEPINFRA_VIDEO_DURATIONS, DEEPINFRA_VIDEO_FALLBACK_MODELS, DEEPINFRA_VLM_FALLBACK_MODELS, DEFAULT_DEEPINFRA_IMAGE_SIZE, DEFAULT_DEEPINFRA_TTS_VOICE, normalizeDeepInfraBaseUrl, normalizeDeepInfraModelRef };

package/dist/media-understanding-provider.js

@@ -1,5 +1,4 @@
-import { DEEPINFRA_BASE_URL } from "./provider-models.js";
-import { DEEPINFRA_STT_FALLBACK_MODELS, DEEPINFRA_VLM_FALLBACK_MODELS } from "./media-models.js";
+import { DEEPINFRA_BASE_URL, DEEPINFRA_STT_FALLBACK_MODELS, DEEPINFRA_VLM_FALLBACK_MODELS } from "./media-models.js";
 import { describeImageWithModel, describeImagesWithModel, transcribeOpenAiCompatibleAudio } from "openclaw/plugin-sdk/media-understanding";
 //#region extensions/deepinfra/media-understanding-provider.ts
 function resolveDefault(surfaceModels, fallback) {

package/dist/{openclaw.plugin-DomWf43f.js → openclaw.plugin-BRaVociw.js}

@@ -1,5 +1,6 @@
 //#region extensions/deepinfra/openclaw.plugin.json
 var modelCatalog = {
+	"modelsDev": { "deepinfra": "deepinfra" },
 	"providers": { "deepinfra": {
 		"baseUrl": "https://api.deepinfra.com/v1/openai",
 		"api": "openai-completions",
@@ -48,9 +49,9 @@ var modelCatalog = {
 				"contextWindow": 1048576,
 				"maxTokens": 1048576,
 				"cost": {
-					"input": .93,
-					"output": 3,
-					"cacheRead": .18,
+					"input": .4875,
+					"output": 1.56,
+					"cacheRead": .091000001625,
 					"cacheWrite": 0
 				},
 				"compat": {
@@ -81,9 +82,9 @@ var modelCatalog = {
 				"contextWindow": 262144,
 				"maxTokens": 262144,
 				"cost": {
-					"input": .74,
-					"output": 3.5,
-					"cacheRead": .15,
+					"input": .68,
+					"output": 3.4,
+					"cacheRead": .136,
 					"cacheWrite": 0
 				},
 				"compat": { "supportsUsageInStreaming": true }
@@ -180,9 +181,9 @@ var modelCatalog = {
 				"contextWindow": 262144,
 				"maxTokens": 262144,
 				"cost": {
-					"input": .1,
+					"input": .09,
 					"output": .3,
-					"cacheRead": .02,
+					"cacheRead": .0199999998,
 					"cacheWrite": 0
 				},
 				"compat": { "supportsUsageInStreaming": true }

package/dist/pricing-api.js

@@ -0,0 +1,51 @@
+import { asFiniteNumberInRange, asOptionalRecord, asPositiveSafeInteger } from "openclaw/plugin-sdk/string-coerce-runtime";
+import { normalizeModelPricingCatalog } from "openclaw/plugin-sdk/model-catalog-pricing";
+//#region extensions/deepinfra/pricing-api.ts
+function normalizeDeepInfraTokenPricing(value) {
+	const row = asOptionalRecord(value);
+	if (!row || row.type !== void 0 && row.type !== "tokens") return;
+	const discount = asFiniteNumberInRange(row.discount ?? 0, {
+		min: 0,
+		max: 1
+	});
+	const inputCents = asFiniteNumberInRange(row.cents_per_input_token, { min: 0 });
+	const outputCents = asFiniteNumberInRange(row.cents_per_output_token, { min: 0 });
+	const cacheRatio = asFiniteNumberInRange(row.rate_per_input_token_cached ?? 0, { min: 0 });
+	const table = row.table == null ? void 0 : asOptionalRecord(row.table);
+	const explicitWrites = row.rate_per_explicit_cache_write_token;
+	const writeRates = explicitWrites == null ? void 0 : asOptionalRecord(explicitWrites);
+	if (discount === void 0 || inputCents === void 0 || outputCents === void 0 || cacheRatio === void 0 || row.full != null && typeof row.full !== "string" || row.table != null && !table || row.discount_ends_at != null && !Number.isSafeInteger(row.discount_ends_at) || [
+		"rate_per_input_token_cache_write",
+		"rate_per_service_tier_priority",
+		"rate_per_service_tier_flex"
+	].some((key) => row[key] != null && asFiniteNumberInRange(row[key], { min: 0 }) === void 0) || explicitWrites != null && (!writeRates || Object.values(writeRates).some((rate) => asFiniteNumberInRange(rate, { min: 0 }) === void 0)) || row.explicit_cache_granularity_tokens != null && asPositiveSafeInteger(row.explicit_cache_granularity_tokens) === void 0) return;
+	const input = inputCents * 1e4 * (1 - discount);
+	const output = outputCents * 1e4 * (1 - discount);
+	const cacheRead = input * cacheRatio;
+	return [
+		input,
+		output,
+		cacheRead
+	].every(Number.isFinite) ? {
+		input,
+		output,
+		cacheRead,
+		cacheWrite: 0
+	} : void 0;
+}
+/** Pure native price owner shared by hosted publication and chat discovery. */
+function parseDeepInfraPricingCatalog(payload) {
+	return normalizeModelPricingCatalog(payload, normalizeDeepInfraTokenPricing, {
+		readModelId: (model) => model.model_name,
+		readPricing: (model) => {
+			const pricing = asOptionalRecord(model.pricing);
+			return typeof pricing?.type === "string" && pricing.type !== "tokens" ? void 0 : model.pricing;
+		},
+		isSupportedPricing: (value) => {
+			const row = asOptionalRecord(value);
+			return row !== void 0 && !row.full && Object.keys(asOptionalRecord(row.table) ?? {}).length === 0 && row.discount_ends_at == null && row.rate_per_input_token_cache_write == null;
+		}
+	});
+}
+//#endregion
+export { parseDeepInfraPricingCatalog };

package/dist/provider-catalog.js

@@ -1,4 +1,5 @@
-import { DEEPINFRA_BASE_URL, DEEPINFRA_MODEL_CATALOG, buildDeepInfraModelDefinition, discoverDeepInfraModels } from "./provider-models.js";
+import { DEEPINFRA_BASE_URL } from "./media-models.js";
+import { DEEPINFRA_MODEL_CATALOG, buildDeepInfraModelDefinition, discoverDeepInfraModels } from "./provider-models.js";
 import { buildSingleProviderApiKeyCatalog } from "openclaw/plugin-sdk/provider-catalog-shared";
 //#region extensions/deepinfra/provider-catalog.ts
 function buildStaticDeepInfraProvider() {
@@ -9,10 +10,11 @@ function buildStaticDeepInfraProvider() {
 	};
 }
 async function buildDeepInfraProvider(options) {
+	const models = await discoverDeepInfraModels(options);
 	return {
 		baseUrl: DEEPINFRA_BASE_URL,
 		api: "openai-completions",
-		models: await discoverDeepInfraModels(options)
+		models
 	};
 }
 function buildDeepInfraApiKeyCatalog(ctx) {

package/dist/provider-models.js

@@ -1,9 +1,11 @@
-import { t as modelCatalog } from "./openclaw.plugin-DomWf43f.js";
-import { buildManifestModelProviderConfig } from "openclaw/plugin-sdk/provider-catalog-shared";
+import { t as modelCatalog } from "./openclaw.plugin-BRaVociw.js";
+import { DEEPINFRA_BASE_URL, DEEPINFRA_TTS_FALLBACK_CATALOG } from "./media-models.js";
+import { parseDeepInfraPricingCatalog } from "./pricing-api.js";
+import { buildManifestModelProviderConfig, getCachedLiveCatalogValue } from "openclaw/plugin-sdk/provider-catalog-shared";
 import { asPositiveSafeInteger } from "openclaw/plugin-sdk/string-coerce-runtime";
 import { withTrustedEnvProxyGuardedFetchMode } from "openclaw/plugin-sdk/fetch-runtime";
 import { isProviderApiKeyConfigured } from "openclaw/plugin-sdk/provider-auth";
-import { LiveModelCatalogHttpError, getCachedLiveProviderModelRows } from "openclaw/plugin-sdk/provider-catalog-live-runtime";
+import { LiveModelCatalogHttpError, fetchLiveProviderModelRows, getCachedLiveProviderModelRows } from "openclaw/plugin-sdk/provider-catalog-live-runtime";
 import { createSubsystemLogger } from "openclaw/plugin-sdk/runtime-env";
 import { hasConfiguredSecretInput } from "openclaw/plugin-sdk/secret-input";
 import { fetchWithSsrFGuard } from "openclaw/plugin-sdk/ssrf-runtime";
@@ -13,14 +15,14 @@ const DEEPINFRA_MANIFEST_PROVIDER = buildManifestModelProviderConfig({
 	providerId: "deepinfra",
 	catalog: modelCatalog.providers.deepinfra
 });
-const DEEPINFRA_BASE_URL = DEEPINFRA_MANIFEST_PROVIDER.baseUrl;
 const DEEPINFRA_MODELS_URL = `${DEEPINFRA_BASE_URL}/models?sort_by=openclaw&filter=with_meta`;
+const DEEPINFRA_PRICING_URL = "https://api.deepinfra.com/models/list";
 const DEEPINFRA_DEFAULT_MODEL_REF = `deepinfra/deepseek-ai/DeepSeek-V4-Flash`;
 const DEEPINFRA_DEFAULT_CONTEXT_WINDOW = 128e3;
 const DEEPINFRA_DEFAULT_MAX_TOKENS = 8192;
 const DEEPINFRA_MODEL_CATALOG = DEEPINFRA_MANIFEST_PROVIDER.models;
 const DISCOVERY_TIMEOUT_MS = 5e3;
-const DISCOVERY_CACHE_TTL_MS = 300 * 1e3;
+const DISCOVERY_CACHE_TTL_MS = 3e5;
 const SURFACE_FOR_TAG = {
 	chat: "chat",
 	vlm: "vlm",
@@ -140,30 +142,7 @@ const STATIC_NON_CHAT_FALLBACK = [
 		defaultHeight: 1024,
 		defaultIterations: 4
 	},
-	{
-		id: "hexgrad/Kokoro-82M",
-		name: "hexgrad/Kokoro-82M",
-		tags: ["tts"],
-		pricing: { input_characters: .65 }
-	},
-	{
-		id: "Qwen/Qwen3-TTS",
-		name: "Qwen/Qwen3-TTS",
-		tags: ["tts"],
-		pricing: { input_characters: .65 }
-	},
-	{
-		id: "ResembleAI/chatterbox-turbo",
-		name: "ResembleAI/chatterbox-turbo",
-		tags: ["tts"],
-		pricing: { input_characters: 1 }
-	},
-	{
-		id: "sesame/csm-1b",
-		name: "sesame/csm-1b",
-		tags: ["tts"],
-		pricing: { input_characters: 7 }
-	},
+	...DEEPINFRA_TTS_FALLBACK_CATALOG,
 	{
 		id: "openai/whisper-large-v3-turbo",
 		name: "openai/whisper-large-v3-turbo",
@@ -205,21 +184,15 @@ function chatSurfaceModelToModelDefinition(model) {
 	const manifestModel = DEEPINFRA_MODEL_CATALOG.find((entry) => entry.id === model.id);
 	const input = model.tags.includes("vlm") ? ["text", "image"] : ["text"];
 	const reasoning = model.tags.includes("reasoning") || model.tags.includes("reasoning_effort");
-	return buildDeepInfraModelDefinition({
+	return {
 		id: model.id,
 		name: model.name,
 		reasoning: manifestModel?.reasoning ?? reasoning,
 		input,
 		...manifestModel?.compat ? { compat: manifestModel.compat } : {},
 		contextWindow: model.contextWindow ?? DEEPINFRA_DEFAULT_CONTEXT_WINDOW,
-		maxTokens: model.maxTokens ?? DEEPINFRA_DEFAULT_MAX_TOKENS,
-		cost: {
-			input: model.pricing.input_tokens ?? 0,
-			output: model.pricing.output_tokens ?? 0,
-			cacheRead: model.pricing.cache_read_tokens ?? 0,
-			cacheWrite: 0
-		}
-	});
+		maxTokens: model.maxTokens ?? DEEPINFRA_DEFAULT_MAX_TOKENS
+	};
 }
 function hasDeepInfraApiKey(options) {
 	const fromEnv = (options?.env ?? process.env).DEEPINFRA_API_KEY;
@@ -231,13 +204,18 @@ function hasDeepInfraApiKey(options) {
 		agentDir: options?.agentDir
 	});
 }
-async function discoverDeepInfraSurfaces(options) {
+function canDiscoverDeepInfra(options) {
 	const env = options?.env ?? process.env;
-	if (env.NODE_ENV === "test" || env.VITEST) return manifestFallbackCatalog();
-	if (!(options?.hasApiKey ?? hasDeepInfraApiKey({
+	if (env.NODE_ENV === "test" || env.VITEST) return false;
+	return options?.hasApiKey ?? hasDeepInfraApiKey({
 		env,
 		agentDir: options?.agentDir
-	}))) return manifestFallbackCatalog();
+	});
+}
+async function discoverDeepInfraSurfaces(options) {
+	return canDiscoverDeepInfra(options) ? loadDeepInfraSurfaces() : manifestFallbackCatalog();
+}
+async function loadDeepInfraSurfaces() {
 	try {
 		const data = await getCachedLiveProviderModelRows({
 			providerId: "deepinfra",
@@ -272,19 +250,62 @@ async function discoverDeepInfraSurfaces(options) {
 		return manifestFallbackCatalog();
 	}
 }
+async function discoverDeepInfraPricing() {
+	try {
+		return await getCachedLiveCatalogValue({
+			keyParts: [
+				"deepinfra",
+				"native-pricing",
+				DEEPINFRA_PRICING_URL
+			],
+			ttlMs: DISCOVERY_CACHE_TTL_MS,
+			load: async () => {
+				const rows = await fetchLiveProviderModelRows({
+					providerId: "deepinfra",
+					endpoint: DEEPINFRA_PRICING_URL,
+					timeoutMs: DISCOVERY_TIMEOUT_MS,
+					buildRequestHeaders: () => ({ Accept: "application/json" }),
+					auditContext: "deepinfra-pricing-discovery",
+					fetchGuard: (params) => fetchWithSsrFGuard(withTrustedEnvProxyGuardedFetchMode(params)),
+					readRows: (body) => {
+						if (!Array.isArray(body)) throw new Error("Native DeepInfra pricing response must be an array");
+						return body;
+					}
+				});
+				const prices = parseDeepInfraPricingCatalog(rows);
+				if (!prices) throw new Error("Native DeepInfra pricing is malformed or has no usable schedules");
+				return prices;
+			}
+		});
+	} catch (error) {
+		log.warn(`Native pricing unavailable: ${String(error)}. Keeping model metadata with unknown estimates; retry discovery or configure explicit model costs.`);
+		return;
+	}
+}
 async function discoverDeepInfraModels(options) {
-	const catalog = await discoverDeepInfraSurfaces(options);
-	if (!catalog.live) return DEEPINFRA_MODEL_CATALOG.map(buildDeepInfraModelDefinition);
+	if (!canDiscoverDeepInfra(options)) return DEEPINFRA_MODEL_CATALOG.map(buildDeepInfraModelDefinition);
+	const [catalog, prices] = await Promise.all([loadDeepInfraSurfaces(), discoverDeepInfraPricing()]);
+	if (!catalog.live && !prices) return DEEPINFRA_MODEL_CATALOG.map(buildDeepInfraModelDefinition);
 	const chatModels = catalog.chat.length > 0 ? catalog.chat : [...catalog.chat, ...catalog.vlm];
-	if (chatModels.length === 0) return DEEPINFRA_MODEL_CATALOG.map(buildDeepInfraModelDefinition);
-	const liveModels = chatModels.map(chatSurfaceModelToModelDefinition);
+	const liveModels = catalog.live ? chatModels.map(chatSurfaceModelToModelDefinition) : [];
 	const seen = new Set(liveModels.map((model) => model.id));
 	const manifestModels = DEEPINFRA_MODEL_CATALOG.filter((model) => {
 		const unseen = !seen.has(model.id);
 		seen.add(model.id);
 		return unseen;
-	}).map(buildDeepInfraModelDefinition);
-	return [...liveModels, ...manifestModels];
+	});
+	const models = [...liveModels, ...manifestModels];
+	const unknownPrices = models.filter((model) => !prices?.has(model.id)).length;
+	if (prices && unknownPrices > 0) log.warn(`Native pricing unavailable or qualified for ${unknownPrices} models; keeping metadata with unknown estimates. Configure explicit model costs if needed.`);
+	return models.map((model) => buildDeepInfraModelDefinition({
+		...model,
+		cost: prices?.get(model.id) ?? {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0
+		}
+	}));
 }
 //#endregion
-export { DEEPINFRA_BASE_URL, DEEPINFRA_DEFAULT_MODEL_REF, DEEPINFRA_MODEL_CATALOG, buildDeepInfraModelDefinition, discoverDeepInfraModels, discoverDeepInfraSurfaces, getDeepInfraSurfaceFallbackCatalog, hasDeepInfraApiKey };
+export { DEEPINFRA_DEFAULT_MODEL_REF, DEEPINFRA_MODEL_CATALOG, buildDeepInfraModelDefinition, discoverDeepInfraModels, discoverDeepInfraSurfaces, getDeepInfraSurfaceFallbackCatalog, hasDeepInfraApiKey };

package/dist/speech-provider.js

@@ -1,7 +1,6 @@
-import { DEEPINFRA_BASE_URL } from "./provider-models.js";
-import { DEEPINFRA_TTS_FALLBACK_MODELS, DEFAULT_DEEPINFRA_TTS_VOICE, normalizeDeepInfraModelRef } from "./media-models.js";
+import { DEEPINFRA_BASE_URL, DEEPINFRA_TTS_FALLBACK_CATALOG, DEEPINFRA_TTS_FALLBACK_MODELS, DEFAULT_DEEPINFRA_TTS_VOICE, normalizeDeepInfraModelRef } from "./media-models.js";
 import { asOptionalRecord } from "openclaw/plugin-sdk/string-coerce-runtime";
-import { createOpenAiCompatibleSpeechProvider } from "openclaw/plugin-sdk/speech";
+import { createOpenAiCompatibleSpeechProvider } from "openclaw/plugin-sdk/speech-provider";
 //#region extensions/deepinfra/speech-provider.ts
 const DEEPINFRA_TTS_RESPONSE_FORMATS = [
 	"mp3",
@@ -12,7 +11,7 @@ const DEEPINFRA_TTS_RESPONSE_FORMATS = [
 ];
 function buildDeepInfraSpeechProvider(options) {
 	const ids = options?.ttsModels && options.ttsModels.length > 0 ? options.ttsModels.map((model) => model.id) : [...DEEPINFRA_TTS_FALLBACK_MODELS];
-	const defaultModel = ids[0] ?? DEEPINFRA_TTS_FALLBACK_MODELS[0];
+	const defaultModel = ids[0] ?? DEEPINFRA_TTS_FALLBACK_CATALOG[0].id;
 	return createOpenAiCompatibleSpeechProvider({
 		id: "deepinfra",
 		label: "DeepInfra",

package/dist/surface-model-catalogs.js

@@ -1,5 +1,5 @@
-import { discoverDeepInfraSurfaces } from "./provider-models.js";
 import { DEEPINFRA_VIDEO_ASPECT_RATIOS, DEEPINFRA_VIDEO_DURATIONS } from "./media-models.js";
+import { discoverDeepInfraSurfaces } from "./provider-models.js";
 //#region extensions/deepinfra/surface-model-catalogs.ts
 const PROVIDER_ID = "deepinfra";
 function surfaceModelToImageGenEntry(model) {

package/dist/video-generation-provider.js

@@ -1,12 +1,11 @@
-import { DEEPINFRA_BASE_URL } from "./provider-models.js";
-import { DEEPINFRA_VIDEO_ASPECT_RATIOS, DEEPINFRA_VIDEO_DURATIONS, DEEPINFRA_VIDEO_FALLBACK_MODELS, normalizeDeepInfraBaseUrl, normalizeDeepInfraModelRef } from "./media-models.js";
+import { DEEPINFRA_BASE_URL, DEEPINFRA_VIDEO_ASPECT_RATIOS, DEEPINFRA_VIDEO_DURATIONS, DEEPINFRA_VIDEO_FALLBACK_MODELS, normalizeDeepInfraBaseUrl, normalizeDeepInfraModelRef } from "./media-models.js";
 import { resolveDeepInfraVideoModelCapabilities } from "./surface-model-catalogs.js";
 import { asSafeIntegerInRange, normalizeOptionalString } from "openclaw/plugin-sdk/string-coerce-runtime";
 import { isProviderApiKeyConfigured } from "openclaw/plugin-sdk/provider-auth";
 import { extensionForMime } from "openclaw/plugin-sdk/media-mime";
 import { canonicalizeBase64 } from "openclaw/plugin-sdk/media-runtime";
 import { resolveApiKeyForProvider } from "openclaw/plugin-sdk/provider-auth-runtime";
-import { assertOkOrThrowHttpError, createProviderOperationDeadline, pollProviderOperationJson, postJsonRequest, readProviderJsonResponse, resolveProviderHttpRequestConfig, resolveProviderOperationTimeoutMs } from "openclaw/plugin-sdk/provider-http";
+import { assertOkOrThrowHttpError, createProviderOperationDeadline, pollProviderOperationJson, postJsonRequest, readProviderJsonResponse, resolveProviderHttpRequestConfig, resolveProviderOperationTimeoutMs, sanitizeConfiguredModelProviderRequest } from "openclaw/plugin-sdk/provider-http";
 //#region extensions/deepinfra/video-generation-provider.ts
 const DEFAULT_HTTP_TIMEOUT_MS = 6e4;
 const POLL_INTERVAL_MS = 5e3;
@@ -128,14 +127,14 @@ function buildDeepInfraVideoGenerationProvider(options) {
 			const { baseUrl, allowPrivateNetwork, headers, dispatcherPolicy } = resolveProviderHttpRequestConfig({
 				baseUrl: resolveDeepInfraVideoBaseUrl(req),
 				defaultBaseUrl: DEEPINFRA_BASE_URL,
-				allowPrivateNetwork: false,
 				defaultHeaders: {
 					Authorization: `Bearer ${auth.apiKey}`,
 					"Content-Type": "application/json"
 				},
 				provider: "deepinfra",
 				capability: "video",
-				transport: "http"
+				transport: "http",
+				request: sanitizeConfiguredModelProviderRequest(req.cfg?.models?.providers?.deepinfra?.request)
 			});
 			const { response, release } = await postJsonRequest({
 				url: `${baseUrl}/videos`,

package/openclaw.plugin.json

@@ -1,5 +1,6 @@
 {
   "id": "deepinfra",
+  "capabilityCatalogEntry": "./dist/capability-catalog.js",
   "doctorContract": {
     "configRepair": true
   },
@@ -7,12 +8,16 @@
     "onStartup": false
   },
   "enabledByDefault": true,
-  "providers": ["deepinfra"],
-  "providerCatalogEntry": "./provider-discovery.ts",
+  "providers": [
+    "deepinfra"
+  ],
+  "providerCatalogEntry": "./dist/provider-discovery.js",
   "providerEndpoints": [
     {
       "endpointClass": "deepinfra-native",
-      "hosts": ["api.deepinfra.com"]
+      "hosts": [
+        "api.deepinfra.com"
+      ]
     }
   ],
   "providerRequest": {
@@ -26,12 +31,19 @@
     "providers": [
       {
         "id": "deepinfra",
-        "authMethods": ["api-key"],
-        "envVars": ["DEEPINFRA_API_KEY"]
+        "authMethods": [
+          "api-key"
+        ],
+        "envVars": [
+          "DEEPINFRA_API_KEY"
+        ]
       }
     ]
   },
   "modelCatalog": {
+    "modelsDev": {
+      "deepinfra": "deepinfra"
+    },
     "providers": {
       "deepinfra": {
         "baseUrl": "https://api.deepinfra.com/v1/openai",
@@ -41,7 +53,9 @@
             "id": "deepseek-ai/DeepSeek-V4-Flash",
             "name": "DeepSeek V4 Flash",
             "reasoning": true,
-            "input": ["text"],
+            "input": [
+              "text"
+            ],
             "contextWindow": 1048576,
             "maxTokens": 1048576,
             "cost": {
@@ -59,7 +73,9 @@
             "id": "deepseek-ai/DeepSeek-V4-Pro",
             "name": "DeepSeek V4 Pro",
             "reasoning": true,
-            "input": ["text"],
+            "input": [
+              "text"
+            ],
             "contextWindow": 1048576,
             "maxTokens": 1048576,
             "cost": {
@@ -77,13 +93,15 @@
             "id": "zai-org/GLM-5.2",
             "name": "GLM-5.2",
             "reasoning": true,
-            "input": ["text"],
+            "input": [
+              "text"
+            ],
             "contextWindow": 1048576,
             "maxTokens": 1048576,
             "cost": {
-              "input": 0.93,
-              "output": 3,
-              "cacheRead": 0.18,
+              "input": 0.4875,
+              "output": 1.56,
+              "cacheRead": 0.091000001625,
               "cacheWrite": 0
             },
             "compat": {
@@ -95,7 +113,10 @@
             "id": "stepfun-ai/Step-3.7-Flash",
             "name": "Step 3.7 Flash",
             "reasoning": true,
-            "input": ["text", "image"],
+            "input": [
+              "text",
+              "image"
+            ],
             "contextWindow": 262144,
             "maxTokens": 262144,
             "cost": {
@@ -112,13 +133,16 @@
             "id": "moonshotai/Kimi-K2.7-Code",
             "name": "Kimi K2.7 Code",
             "reasoning": true,
-            "input": ["text", "image"],
+            "input": [
+              "text",
+              "image"
+            ],
             "contextWindow": 262144,
             "maxTokens": 262144,
             "cost": {
-              "input": 0.74,
-              "output": 3.5,
-              "cacheRead": 0.15,
+              "input": 0.68,
+              "output": 3.4,
+              "cacheRead": 0.136,
               "cacheWrite": 0
             },
             "compat": {
@@ -129,7 +153,10 @@
             "id": "moonshotai/Kimi-K2.6",
             "name": "Kimi K2.6",
             "reasoning": true,
-            "input": ["text", "image"],
+            "input": [
+              "text",
+              "image"
+            ],
             "contextWindow": 262144,
             "maxTokens": 262144,
             "cost": {
@@ -146,7 +173,10 @@
             "id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
             "name": "NVIDIA Nemotron 3 Ultra 550B A55B",
             "reasoning": true,
-            "input": ["text", "image"],
+            "input": [
+              "text",
+              "image"
+            ],
             "contextWindow": 262144,
             "maxTokens": 262144,
             "cost": {
@@ -163,7 +193,9 @@
             "id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B",
             "name": "NVIDIA Nemotron 3 Super 120B A12B",
             "reasoning": true,
-            "input": ["text"],
+            "input": [
+              "text"
+            ],
             "contextWindow": 262144,
             "maxTokens": 262144,
             "cost": {
@@ -182,7 +214,9 @@
             "status": "deprecated",
             "replacedBy": "deepseek-ai/DeepSeek-V4-Pro",
             "reasoning": false,
-            "input": ["text"],
+            "input": [
+              "text"
+            ],
             "contextWindow": 163840,
             "maxTokens": 163840,
             "cost": {
@@ -201,7 +235,9 @@
             "status": "deprecated",
             "replacedBy": "zai-org/GLM-5.2",
             "reasoning": true,
-            "input": ["text"],
+            "input": [
+              "text"
+            ],
             "contextWindow": 202752,
             "maxTokens": 202752,
             "cost": {
@@ -221,13 +257,15 @@
             "status": "deprecated",
             "replacedBy": "stepfun-ai/Step-3.7-Flash",
             "reasoning": true,
-            "input": ["text"],
+            "input": [
+              "text"
+            ],
             "contextWindow": 262144,
             "maxTokens": 262144,
             "cost": {
-              "input": 0.1,
+              "input": 0.09,
               "output": 0.3,
-              "cacheRead": 0.02,
+              "cacheRead": 0.0199999998,
               "cacheWrite": 0
             },
             "compat": {
@@ -240,7 +278,10 @@
             "status": "deprecated",
             "replacedBy": "moonshotai/Kimi-K2.6",
             "reasoning": true,
-            "input": ["text", "image"],
+            "input": [
+              "text",
+              "image"
+            ],
             "contextWindow": 262144,
             "maxTokens": 262144,
             "cost": {
@@ -260,6 +301,15 @@
       "deepinfra": "refreshable"
     }
   },
+  "modelPricing": {
+    "providers": {
+      "deepinfra": {
+        "deepinfra": {
+          "provider": "deepinfra"
+        }
+      }
+    }
+  },
   "providerAuthChoices": [
     {
       "provider": "deepinfra",
@@ -278,15 +328,28 @@
     }
   ],
   "contracts": {
-    "mediaUnderstandingProviders": ["deepinfra"],
-    "embeddingProviders": ["deepinfra"],
-    "imageGenerationProviders": ["deepinfra"],
-    "speechProviders": ["deepinfra"],
-    "videoGenerationProviders": ["deepinfra"]
+    "mediaUnderstandingProviders": [
+      "deepinfra"
+    ],
+    "embeddingProviders": [
+      "deepinfra"
+    ],
+    "imageGenerationProviders": [
+      "deepinfra"
+    ],
+    "speechProviders": [
+      "deepinfra"
+    ],
+    "videoGenerationProviders": [
+      "deepinfra"
+    ]
   },
   "mediaUnderstandingProviderMetadata": {
     "deepinfra": {
-      "capabilities": ["image", "audio"],
+      "capabilities": [
+        "image",
+        "audio"
+      ],
       "defaultModels": {
         "image": "moonshotai/Kimi-K2.5",
         "audio": "openai/whisper-large-v3-turbo"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@openclaw/deepinfra-provider",
-  "version": "2026.9.1-beta.1",
+  "version": "2026.9.2",
   "description": "OpenClaw DeepInfra provider plugin.",
   "repository": {
     "type": "git",
@@ -18,10 +18,10 @@
       "minHostVersion": ">=2026.6.8"
     },
     "compat": {
-      "pluginApi": ">=2026.9.1-beta.1"
+      "pluginApi": ">=2026.9.2"
     },
     "build": {
-      "openclawVersion": "2026.9.1-beta.1",
+      "openclawVersion": "2026.9.2",
       "bundledDist": false
     },
     "release": {
@@ -38,7 +38,7 @@
     "README.md"
   ],
   "peerDependencies": {
-    "openclaw": ">=2026.9.1-beta.1"
+    "openclaw": ">=2026.9.2"
   },
   "peerDependenciesMeta": {
     "openclaw": {