Add NVIDIA NIM provider

packages/ai/CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## [Unreleased]
 
+### Added
+
+- Added NVIDIA NIM as a built-in OpenAI-compatible provider, exposing public NIM models that support tool use.
+
 ### Fixed
 
 - Fixed Amazon Bedrock requests to replace blank required user/tool-result text with a placeholder and skip blank replay text blocks ([#4975](https://github.com/earendil-works/pi/issues/4975)).

packages/ai/README.md

@@ -54,6 +54,7 @@ Unified LLM API with automatic model discovery, provider configuration, token an
 - **Azure OpenAI (Responses)**
 - **OpenAI Codex** (ChatGPT Plus/Pro subscription, requires OAuth, see below)
 - **DeepSeek**
+- **NVIDIA NIM**
 - **Anthropic**
 - **Google**
 - **Vertex AI** (Gemini via Vertex AI)
@@ -801,7 +802,7 @@ A **provider** offers models through a specific API. For example:
 - **Google** models use the `google-generative-ai` API
 - **OpenAI** models use the `openai-responses` API
 - **Mistral** models use the `mistral-conversations` API
-- **xAI, Cerebras, Groq, Together AI, etc.** models use the `openai-completions` API (OpenAI-compatible)
+- **xAI, Cerebras, Groq, NVIDIA NIM, Together AI, etc.** models use the `openai-completions` API (OpenAI-compatible)
 
 ### Querying Providers and Models
 
@@ -923,7 +924,7 @@ const ollamaReasoningModel: Model<'openai-completions'> = {
 
 ### OpenAI Compatibility Settings
 
-The `openai-completions` API is implemented by many providers with minor differences. By default, the library auto-detects compatibility settings based on `baseUrl` for a small set of known OpenAI-compatible providers (Cerebras, xAI, Chutes, DeepSeek, Together AI, zAi, OpenCode, Cloudflare Workers AI, etc.). For custom proxies or unknown endpoints, you can override these settings via the `compat` field. For `openai-responses` models, the compat field only supports Responses-specific flags.
+The `openai-completions` API is implemented by many providers with minor differences. By default, the library auto-detects compatibility settings based on `baseUrl` for a small set of known OpenAI-compatible providers (Cerebras, xAI, Chutes, DeepSeek, NVIDIA NIM, Together AI, zAi, OpenCode, Cloudflare Workers AI, etc.). For custom proxies or unknown endpoints, you can override these settings via the `compat` field. For `openai-responses` models, the compat field only supports Responses-specific flags.
 
 ```typescript
 interface OpenAICompletionsCompat {
@@ -1102,6 +1103,7 @@ In Node.js environments, you can set environment variables to avoid passing API
 | Azure OpenAI | `AZURE_OPENAI_API_KEY` + `AZURE_OPENAI_BASE_URL` (e.g. `https://{resource}.openai.azure.com`) or `AZURE_OPENAI_RESOURCE_NAME`. Supports `*.openai.azure.com` and `*.cognitiveservices.azure.com`; root endpoints auto-normalize to `/openai/v1`. Optional: `AZURE_OPENAI_API_VERSION` (default `v1`), `AZURE_OPENAI_DEPLOYMENT_NAME_MAP`. |
 | Anthropic | `ANTHROPIC_API_KEY` or `ANTHROPIC_OAUTH_TOKEN` |
 | DeepSeek | `DEEPSEEK_API_KEY` |
+| NVIDIA NIM | `NVIDIA_API_KEY` |
 | Google | `GEMINI_API_KEY` |
 | Vertex AI | `GOOGLE_CLOUD_API_KEY` or `GOOGLE_CLOUD_PROJECT` (or `GCLOUD_PROJECT`) + `GOOGLE_CLOUD_LOCATION` + ADC |
 | Mistral | `MISTRAL_API_KEY` |

packages/ai/scripts/generate-models.ts

@@ -32,12 +32,17 @@ interface ModelsDevModel {
 	};
 	modalities?: {
 		input?: string[];
+		output?: string[];
 	};
 	provider?: {
 		npm?: string;
 	};
 }
 
+interface NvidiaNimModelListItem {
+	id: string;
+}
+
 interface AiGatewayModel {
 	id: string;
 	name?: string;
@@ -117,6 +122,39 @@ const TOGETHER_TOGGLE_REASONING_LEVEL_MAP = {
 
 const AI_GATEWAY_MODELS_URL = "https://ai-gateway.vercel.sh/v1";
 const AI_GATEWAY_BASE_URL = "https://ai-gateway.vercel.sh";
+const NVIDIA_BASE_URL = "https://integrate.api.nvidia.com/v1";
+const NVIDIA_HEADERS = {
+	"NVCF-POLL-SECONDS": "3600",
+} as const;
+const NVIDIA_OPENAI_COMPAT: OpenAICompletionsCompat = {
+	supportsStore: false,
+	supportsDeveloperRole: false,
+	supportsReasoningEffort: false,
+	maxTokensField: "max_tokens",
+	supportsStrictMode: false,
+	supportsLongCacheRetention: false,
+};
+const NVIDIA_NIM_UNSUPPORTED_MODELS = new Set([
+	"abacusai/dracarys-llama-3.1-70b-instruct",
+	"bytedance/seed-oss-36b-instruct",
+	"deepseek-ai/deepseek-v4-flash",
+	"deepseek-ai/deepseek-v4-pro",
+	"google/gemma-2-2b-it",
+	"google/gemma-3n-e2b-it",
+	"google/gemma-3n-e4b-it",
+	"google/gemma-4-31b-it",
+	"meta/llama-3.2-1b-instruct",
+	"meta/llama-4-maverick-17b-128e-instruct",
+	"microsoft/phi-4-mini-instruct",
+	"minimaxai/minimax-m2.7",
+	"mistralai/mistral-nemotron",
+	"nvidia/nemotron-mini-4b-instruct",
+	"qwen/qwen3-next-80b-a3b-instruct",
+	"qwen/qwen3.5-122b-a10b",
+	"qwen/qwen3.5-397b-a17b",
+	"sarvamai/sarvam-m",
+	"upstage/solar-10.7b-instruct",
+]);
 const ZAI_TOOL_STREAM_UNSUPPORTED_MODELS = new Set(["glm-4.5", "glm-4.5-air", "glm-4.5-flash", "glm-4.5v"]);
 const EAGER_TOOL_INPUT_STREAMING_UNSUPPORTED_ANTHROPIC_MODELS = new Set([
 	"github-copilot:claude-haiku-4.5",
@@ -312,6 +350,30 @@ function getBedrockBaseUrl(modelId: string): string {
 		: "https://bedrock-runtime.us-east-1.amazonaws.com";
 }
 
+function normalizeNvidiaModelId(modelId: string): string {
+	return modelId.toLowerCase().replaceAll("_", ".");
+}
+
+async function fetchNvidiaNimModelIds(): Promise<Map<string, string>> {
+	try {
+		console.log("Fetching models from NVIDIA NIM API...");
+		const response = await fetch(`${NVIDIA_BASE_URL}/models`);
+		const data = (await response.json()) as { data?: NvidiaNimModelListItem[] };
+		const modelIds = new Map<string, string>();
+
+		for (const model of data.data ?? []) {
+			modelIds.set(model.id, model.id);
+			modelIds.set(normalizeNvidiaModelId(model.id), model.id);
+		}
+
+		console.log(`Fetched ${data.data?.length ?? 0} model IDs from NVIDIA NIM`);
+		return modelIds;
+	} catch (error) {
+		console.error("Failed to fetch NVIDIA NIM models:", error);
+		return new Map();
+	}
+}
+
 async function fetchOpenRouterModels(): Promise<Model<any>[]> {
 	try {
 		console.log("Fetching models from OpenRouter API...");
@@ -435,6 +497,7 @@ async function loadModelsDevData(): Promise<Model<any>[]> {
 		const data = await response.json();
 
 		const models: Model<any>[] = [];
+		const nvidiaNimModelIds = data.nvidia?.models ? await fetchNvidiaNimModelIds() : new Map<string, string>();
 
 		// Process Amazon Bedrock models
 		if (data["amazon-bedrock"]?.models) {
@@ -836,6 +899,40 @@ async function loadModelsDevData(): Promise<Model<any>[]> {
 			}
 		}
 
+		// Process NVIDIA NIM models
+		if (data.nvidia?.models) {
+			for (const [modelId, model] of Object.entries(data.nvidia.models)) {
+				const m = model as ModelsDevModel;
+				if (m.tool_call !== true) continue;
+				if (!m.modalities?.input?.includes("text")) continue;
+				if (!m.modalities?.output?.includes("text")) continue;
+
+				const liveModelId = nvidiaNimModelIds.get(modelId) ?? nvidiaNimModelIds.get(normalizeNvidiaModelId(modelId));
+				if (!liveModelId) continue;
+				if (NVIDIA_NIM_UNSUPPORTED_MODELS.has(liveModelId)) continue;
+
+				models.push({
+					id: liveModelId,
+					name: m.name || liveModelId,
+					api: "openai-completions",
+					provider: "nvidia",
+					baseUrl: NVIDIA_BASE_URL,
+					headers: { ...NVIDIA_HEADERS },
+					reasoning: m.reasoning === true,
+					input: m.modalities.input.includes("image") ? ["text", "image"] : ["text"],
+					cost: {
+						input: m.cost?.input || 0,
+						output: m.cost?.output || 0,
+						cacheRead: m.cost?.cache_read || 0,
+						cacheWrite: m.cost?.cache_write || 0,
+					},
+					compat: NVIDIA_OPENAI_COMPAT,
+					contextWindow: m.limit?.context || 4096,
+					maxTokens: m.limit?.output || 4096,
+				});
+			}
+		}
+
 		// Process Together AI models
 		const togetherProvider = data.together ?? data.togetherai ?? data["together-ai"];
 		if (togetherProvider?.models) {

packages/ai/src/env-api-keys.ts

@@ -101,6 +101,7 @@ function getApiKeyEnvVars(provider: string): readonly string[] | undefined {
 	const envMap: Record<string, string> = {
 		openai: "OPENAI_API_KEY",
 		"azure-openai-responses": "AZURE_OPENAI_API_KEY",
+		nvidia: "NVIDIA_API_KEY",
 		deepseek: "DEEPSEEK_API_KEY",
 		google: "GEMINI_API_KEY",
 		"google-vertex": "GOOGLE_CLOUD_API_KEY",

packages/ai/src/models.generated.ts

@@ -6335,6 +6335,369 @@ export const MODELS = {
 			maxTokens: 262144,
 		} satisfies Model<"openai-completions">,
 	},
+	"nvidia": {
+		"meta/llama-3.1-70b-instruct": {
+			id: "meta/llama-3.1-70b-instruct",
+			name: "Llama 3.1 70b Instruct",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: false,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 128000,
+			maxTokens: 4096,
+		} satisfies Model<"openai-completions">,
+		"meta/llama-3.1-8b-instruct": {
+			id: "meta/llama-3.1-8b-instruct",
+			name: "Llama 3.1 8B Instruct",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: false,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 16000,
+			maxTokens: 4096,
+		} satisfies Model<"openai-completions">,
+		"meta/llama-3.2-11b-vision-instruct": {
+			id: "meta/llama-3.2-11b-vision-instruct",
+			name: "Llama 3.2 11b Vision Instruct",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: false,
+			input: ["text", "image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 128000,
+			maxTokens: 4096,
+		} satisfies Model<"openai-completions">,
+		"meta/llama-3.2-90b-vision-instruct": {
+			id: "meta/llama-3.2-90b-vision-instruct",
+			name: "Llama-3.2-90B-Vision-Instruct",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: false,
+			input: ["text", "image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 128000,
+			maxTokens: 8192,
+		} satisfies Model<"openai-completions">,
+		"meta/llama-3.3-70b-instruct": {
+			id: "meta/llama-3.3-70b-instruct",
+			name: "Llama 3.3 70b Instruct",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: false,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 128000,
+			maxTokens: 4096,
+		} satisfies Model<"openai-completions">,
+		"mistralai/mistral-large-3-675b-instruct-2512": {
+			id: "mistralai/mistral-large-3-675b-instruct-2512",
+			name: "Mistral Large 3 675B Instruct 2512",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: false,
+			input: ["text", "image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 262144,
+			maxTokens: 262144,
+		} satisfies Model<"openai-completions">,
+		"mistralai/mistral-small-4-119b-2603": {
+			id: "mistralai/mistral-small-4-119b-2603",
+			name: "mistral-small-4-119b-2603",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: false,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 128000,
+			maxTokens: 8192,
+		} satisfies Model<"openai-completions">,
+		"moonshotai/kimi-k2.6": {
+			id: "moonshotai/kimi-k2.6",
+			name: "Kimi K2.6",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: true,
+			input: ["text", "image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 262144,
+			maxTokens: 262144,
+		} satisfies Model<"openai-completions">,
+		"nvidia/llama-3.3-nemotron-super-49b-v1": {
+			id: "nvidia/llama-3.3-nemotron-super-49b-v1",
+			name: "Llama 3.3 Nemotron Super 49B v1",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 131072,
+			maxTokens: 131072,
+		} satisfies Model<"openai-completions">,
+		"nvidia/llama-3.3-nemotron-super-49b-v1.5": {
+			id: "nvidia/llama-3.3-nemotron-super-49b-v1.5",
+			name: "Llama 3.3 Nemotron Super 49B v1.5",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 131072,
+			maxTokens: 131072,
+		} satisfies Model<"openai-completions">,
+		"nvidia/nemotron-3-nano-30b-a3b": {
+			id: "nvidia/nemotron-3-nano-30b-a3b",
+			name: "nemotron-3-nano-30b-a3b",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 131072,
+			maxTokens: 131072,
+		} satisfies Model<"openai-completions">,
+		"nvidia/nemotron-3-nano-omni-30b-a3b-reasoning": {
+			id: "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning",
+			name: "Nemotron 3 Nano Omni",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: true,
+			input: ["text", "image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 256000,
+			maxTokens: 65536,
+		} satisfies Model<"openai-completions">,
+		"nvidia/nemotron-3-super-120b-a12b": {
+			id: "nvidia/nemotron-3-super-120b-a12b",
+			name: "Nemotron 3 Super",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0.2,
+				output: 0.8,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 262144,
+			maxTokens: 262144,
+		} satisfies Model<"openai-completions">,
+		"nvidia/nvidia-nemotron-nano-9b-v2": {
+			id: "nvidia/nvidia-nemotron-nano-9b-v2",
+			name: "nvidia-nemotron-nano-9b-v2",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 131072,
+			maxTokens: 131072,
+		} satisfies Model<"openai-completions">,
+		"openai/gpt-oss-20b": {
+			id: "openai/gpt-oss-20b",
+			name: "GPT OSS 20B",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 131072,
+			maxTokens: 32768,
+		} satisfies Model<"openai-completions">,
+		"qwen/qwen3-coder-480b-a35b-instruct": {
+			id: "qwen/qwen3-coder-480b-a35b-instruct",
+			name: "Qwen3 Coder 480B A35B Instruct",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: false,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 262144,
+			maxTokens: 66536,
+		} satisfies Model<"openai-completions">,
+		"stepfun-ai/step-3.5-flash": {
+			id: "stepfun-ai/step-3.5-flash",
+			name: "Step 3.5 Flash",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 256000,
+			maxTokens: 16384,
+		} satisfies Model<"openai-completions">,
+		"stepfun-ai/step-3.7-flash": {
+			id: "stepfun-ai/step-3.7-flash",
+			name: "Step 3.7 Flash",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: true,
+			input: ["text", "image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 256000,
+			maxTokens: 16384,
+		} satisfies Model<"openai-completions">,
+		"z-ai/glm-5.1": {
+			id: "z-ai/glm-5.1",
+			name: "GLM-5.1",
+			api: "openai-completions",
+			provider: "nvidia",
+			baseUrl: "https://integrate.api.nvidia.com/v1",
+			headers: {"NVCF-POLL-SECONDS":"3600"},
+			compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"supportsLongCacheRetention":false},
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 131072,
+			maxTokens: 131072,
+		} satisfies Model<"openai-completions">,
+	},
 	"openai": {
 		"gpt-4": {
 			id: "gpt-4",

packages/ai/src/providers/openai-completions.ts

@@ -1078,8 +1078,10 @@ function detectCompat(model: Model<"openai-completions">): ResolvedOpenAIComplet
 	const isOpenRouter = provider === "openrouter" || baseUrl.includes("openrouter.ai");
 	const isCloudflareWorkersAI = provider === "cloudflare-workers-ai" || baseUrl.includes("api.cloudflare.com");
 	const isCloudflareAiGateway = provider === "cloudflare-ai-gateway" || baseUrl.includes("gateway.ai.cloudflare.com");
+	const isNvidia = provider === "nvidia" || baseUrl.includes("integrate.api.nvidia.com");
 
 	const isNonStandard =
+		isNvidia ||
 		provider === "cerebras" ||
 		baseUrl.includes("cerebras.ai") ||
 		provider === "xai" ||
@@ -1094,7 +1096,7 @@ function detectCompat(model: Model<"openai-completions">): ResolvedOpenAIComplet
 		isCloudflareWorkersAI ||
 		isCloudflareAiGateway;
 
-	const useMaxTokens = baseUrl.includes("chutes.ai") || isMoonshot || isCloudflareAiGateway || isTogether;
+	const useMaxTokens = baseUrl.includes("chutes.ai") || isMoonshot || isCloudflareAiGateway || isTogether || isNvidia;
 
 	const isGrok = provider === "xai" || baseUrl.includes("api.x.ai");
 	const isDeepSeek = provider === "deepseek" || baseUrl.includes("deepseek.com");
@@ -1103,7 +1105,7 @@ function detectCompat(model: Model<"openai-completions">): ResolvedOpenAIComplet
 	return {
 		supportsStore: !isNonStandard,
 		supportsDeveloperRole: !isNonStandard && !isOpenRouter,
-		supportsReasoningEffort: !isGrok && !isZai && !isMoonshot && !isTogether && !isCloudflareAiGateway,
+		supportsReasoningEffort: !isGrok && !isZai && !isMoonshot && !isTogether && !isCloudflareAiGateway && !isNvidia,
 		supportsUsageInStreaming: true,
 		maxTokensField: useMaxTokens ? "max_tokens" : "max_completion_tokens",
 		requiresToolResultName: false,
@@ -1122,10 +1124,10 @@ function detectCompat(model: Model<"openai-completions">): ResolvedOpenAIComplet
 		openRouterRouting: {},
 		vercelGatewayRouting: {},
 		zaiToolStream: false,
-		supportsStrictMode: !isMoonshot && !isTogether && !isCloudflareAiGateway,
+		supportsStrictMode: !isMoonshot && !isTogether && !isCloudflareAiGateway && !isNvidia,
 		cacheControlFormat,
 		sendSessionAffinityHeaders: false,
-		supportsLongCacheRetention: !(isTogether || isCloudflareWorkersAI || isCloudflareAiGateway),
+		supportsLongCacheRetention: !(isTogether || isCloudflareWorkersAI || isCloudflareAiGateway || isNvidia),
 	};
 }
 

packages/ai/src/types.ts

@@ -28,6 +28,7 @@ export type KnownProvider =
 	| "openai"
 	| "azure-openai-responses"
 	| "openai-codex"
+	| "nvidia"
 	| "deepseek"
 	| "github-copilot"
 	| "xai"

packages/ai/test/stream.test.ts

@@ -787,6 +787,30 @@ describe("Generate E2E Tests", () => {
 		});
 	});
 
+	describe.skipIf(!process.env.NVIDIA_API_KEY)("NVIDIA NIM Provider (Nemotron 3 Super via OpenAI Completions)", () => {
+		const llm = getModel("nvidia", "nvidia/nemotron-3-super-120b-a12b");
+
+		it("should complete basic text generation", { retry: 3 }, async () => {
+			await basicTextGeneration(llm);
+		});
+
+		it("should handle tool calling", { retry: 3 }, async () => {
+			await handleToolCall(llm);
+		});
+
+		it("should handle streaming", { retry: 3 }, async () => {
+			await handleStreaming(llm);
+		});
+
+		it("should handle thinking mode", { retry: 3 }, async () => {
+			await handleThinking(llm, { reasoningEffort: "high" });
+		});
+
+		it("should handle multi-turn with thinking and tools", { retry: 3 }, async () => {
+			await multiTurn(llm, { reasoningEffort: "high" });
+		});
+	});
+
 	describe.skipIf(!process.env.OPENROUTER_API_KEY)("OpenRouter Provider (glm-4.5v via OpenAI Completions)", () => {
 		const llm = getModel("openrouter", "z-ai/glm-4.5v");
 

packages/coding-agent/CHANGELOG.md

@@ -4,6 +4,7 @@
 
 ### Added
 
+- Added NVIDIA NIM provider selection, setup documentation, and direct NIM request attribution headers.
 - Added `ctx.mode` to extension contexts so extensions can distinguish TUI, RPC, JSON, and print mode.
 - Added `ctx.getSystemPromptOptions()` for extension commands to inspect the current base system prompt inputs.
 

packages/coding-agent/README.md

@@ -113,6 +113,7 @@ For each built-in provider, pi maintains a list of tool-capable models, updated
 - OpenAI
 - Azure OpenAI
 - DeepSeek
+- NVIDIA NIM
 - Google Gemini
 - Google Vertex
 - Amazon Bedrock
@@ -290,7 +291,7 @@ See [docs/settings.md](docs/settings.md) for all options.
 Pi has two separate startup features:
 
 - **Update check:** fetches `https://pi.dev/api/latest-version` to check whether a newer Pi version exists. Disable it with `PI_SKIP_VERSION_CHECK=1`. Disabling update checks only turns off this check.
-- **Install/update telemetry:** after first install or a changelog-detected update, sends an anonymous version ping to `https://pi.dev/api/report-install`. Opt out by setting `enableInstallTelemetry` to `false` in `settings.json`, or by setting `PI_TELEMETRY=0`. This does not disable update checks; Pi may still contact `pi.dev` for the latest version unless update checks are disabled or offline mode is enabled.
+- **Install/update telemetry:** after first install or a changelog-detected update, sends an anonymous version ping to `https://pi.dev/api/report-install`. This setting also controls optional provider attribution headers for OpenRouter, Cloudflare, and direct NVIDIA NIM requests. Opt out by setting `enableInstallTelemetry` to `false` in `settings.json`, or by setting `PI_TELEMETRY=0`. This does not disable update checks; Pi may still contact `pi.dev` for the latest version unless update checks are disabled or offline mode is enabled.
 
 Use `--offline` or `PI_OFFLINE=1` to disable all startup network operations described here, including update checks, package update checks, and install/update telemetry.
 
@@ -641,7 +642,7 @@ pi --thinking high "Solve this complex problem"
 | `PI_PACKAGE_DIR` | Override package directory (useful for Nix/Guix where store paths tokenize poorly) |
 | `PI_OFFLINE` | Disable startup network operations, including update checks, package update checks, and install/update telemetry |
 | `PI_SKIP_VERSION_CHECK` | Skip the Pi version update check at startup. This prevents the `pi.dev` latest-version request |
-| `PI_TELEMETRY` | Override install/update telemetry. Use `1`/`true`/`yes` to enable or `0`/`false`/`no` to disable. This does not disable update checks |
+| `PI_TELEMETRY` | Override install/update telemetry and provider attribution headers. Use `1`/`true`/`yes` to enable or `0`/`false`/`no` to disable. This does not disable update checks |
 | `PI_CACHE_RETENTION` | Set to `long` for extended prompt cache (Anthropic: 1h, OpenAI: 24h) |
 | `VISUAL`, `EDITOR` | External editor for Ctrl+G |
 

packages/coding-agent/docs/providers.md

@@ -52,6 +52,7 @@ pi
 | Azure OpenAI Responses | `AZURE_OPENAI_API_KEY` | `azure-openai-responses` |
 | OpenAI | `OPENAI_API_KEY` | `openai` |
 | DeepSeek | `DEEPSEEK_API_KEY` | `deepseek` |
+| NVIDIA NIM | `NVIDIA_API_KEY` | `nvidia` |
 | Google Gemini | `GEMINI_API_KEY` | `google` |
 | Mistral | `MISTRAL_API_KEY` | `mistral` |
 | Groq | `GROQ_API_KEY` | `groq` |
@@ -86,6 +87,7 @@ Store credentials in `~/.pi/agent/auth.json`:
   "anthropic": { "type": "api_key", "key": "sk-ant-..." },
   "openai": { "type": "api_key", "key": "sk-..." },
   "deepseek": { "type": "api_key", "key": "sk-..." },
+  "nvidia": { "type": "api_key", "key": "nvapi-..." },
   "google": { "type": "api_key", "key": "..." },
   "opencode": { "type": "api_key", "key": "..." },
   "opencode-go": { "type": "api_key", "key": "..." },

packages/coding-agent/docs/usage.md

@@ -275,7 +275,7 @@ pi --exclude-tools ask_question
 | `PI_PACKAGE_DIR` | Override package directory, useful for Nix/Guix store paths |
 | `PI_OFFLINE` | Disable startup network operations, including update checks, package update checks, and install/update telemetry |
 | `PI_SKIP_VERSION_CHECK` | Skip the Pi version update check at startup. This prevents the `pi.dev` latest-version request |
-| `PI_TELEMETRY` | Override install/update telemetry: `1`/`true`/`yes` or `0`/`false`/`no`. This does not disable update checks |
+| `PI_TELEMETRY` | Override install/update telemetry and provider attribution headers: `1`/`true`/`yes` or `0`/`false`/`no`. This does not disable update checks |
 | `PI_CACHE_RETENTION` | Set to `long` for extended prompt cache where supported |
 | `VISUAL`, `EDITOR` | External editor for Ctrl+G |
 

packages/coding-agent/src/cli/args.ts

@@ -335,6 +335,7 @@ ${chalk.bold("Environment Variables:")}
   AZURE_OPENAI_API_VERSION         - Azure OpenAI API version (default: v1)
   AZURE_OPENAI_DEPLOYMENT_NAME_MAP - Azure OpenAI model=deployment map (comma-separated)
   DEEPSEEK_API_KEY                 - DeepSeek API key
+  NVIDIA_API_KEY                   - NVIDIA NIM API key
   GEMINI_API_KEY                   - Google Gemini API key
   GROQ_API_KEY                     - Groq API key
   CEREBRAS_API_KEY                 - Cerebras API key

packages/coding-agent/src/core/agent-session.ts

@@ -2787,6 +2787,7 @@ export class AgentSession {
 					customInstructions,
 					replaceInstructions,
 					reserveTokens: branchSummarySettings.reserveTokens,
+					streamFn: this.agent.streamFn,
 				});
 				if (result.aborted) {
 					return { cancelled: true, aborted: true };

packages/coding-agent/src/core/compaction/branch-summarization.ts

@@ -5,8 +5,8 @@
  * a summary of the branch being left so context isn't lost.
  */
 
-import type { AgentMessage } from "@earendil-works/pi-agent-core";
+import type { AgentMessage, StreamFn } from "@earendil-works/pi-agent-core";
-import type { Model } from "@earendil-works/pi-ai";
+import type { Model, SimpleStreamOptions } from "@earendil-works/pi-ai";
 import { completeSimple } from "@earendil-works/pi-ai";
 import {
 	convertToLlm,
@@ -77,6 +77,8 @@ export interface GenerateBranchSummaryOptions {
 	replaceInstructions?: boolean;
 	/** Tokens reserved for prompt + LLM response (default 16384) */
 	reserveTokens?: number;
+	/** Optional session stream function. Used to preserve SDK request behavior without mutating agent state. */
+	streamFn?: StreamFn;
 }
 
 // ============================================================================
@@ -284,7 +286,16 @@ export async function generateBranchSummary(
 	entries: SessionEntry[],
 	options: GenerateBranchSummaryOptions,
 ): Promise<BranchSummaryResult> {
-	const { model, apiKey, headers, signal, customInstructions, replaceInstructions, reserveTokens = 16384 } = options;
+	const {
+		model,
+		apiKey,
+		headers,
+		signal,
+		customInstructions,
+		replaceInstructions,
+		reserveTokens = 16384,
+		streamFn,
+	} = options;
 
 	// Token budget = context window minus reserved space for prompt + response
 	const contextWindow = model.contextWindow || 128000;
@@ -320,12 +331,14 @@ export async function generateBranchSummary(
 		},
 	];
 
-	// Call LLM for summarization
+	// Call LLM for summarization. Prefer the session stream function so SDK
-	const response = await completeSimple(
+	// request behavior (timeouts, retries, attribution headers) stays consistent
-		model,
+	// without running through agent state/events.
-		{ systemPrompt: SUMMARIZATION_SYSTEM_PROMPT, messages: summarizationMessages },
+	const context = { systemPrompt: SUMMARIZATION_SYSTEM_PROMPT, messages: summarizationMessages };
-		{ apiKey, headers, signal, maxTokens: 2048 },
+	const requestOptions: SimpleStreamOptions = { apiKey, headers, signal, maxTokens: 2048 };
-	);
+	const response = streamFn
+		? await (await streamFn(model, context, requestOptions)).result()
+		: await completeSimple(model, context, requestOptions);
 
 	// Check if aborted or errored
 	if (response.stopReason === "aborted") {

packages/coding-agent/src/core/model-resolver.ts

@@ -17,6 +17,7 @@ export const defaultModelPerProvider: Record<KnownProvider, string> = {
 	openai: "gpt-5.4",
 	"azure-openai-responses": "gpt-5.4",
 	"openai-codex": "gpt-5.5",
+	nvidia: "nvidia/nemotron-3-super-120b-a12b",
 	deepseek: "deepseek-v4-pro",
 	google: "gemini-3.1-pro-preview",
 	"google-vertex": "gemini-3.1-pro-preview",

packages/coding-agent/src/core/provider-attribution.ts

@@ -0,0 +1,97 @@
+import type { Api, Model } from "@earendil-works/pi-ai";
+import type { SettingsManager } from "./settings-manager.ts";
+import { isInstallTelemetryEnabled } from "./telemetry.ts";
+
+const OPENROUTER_HOST = "openrouter.ai";
+const NVIDIA_NIM_HOST = "integrate.api.nvidia.com";
+const CLOUDFLARE_API_HOST = "api.cloudflare.com";
+const CLOUDFLARE_AI_GATEWAY_HOST = "gateway.ai.cloudflare.com";
+const OPENCODE_HOST = "opencode.ai";
+
+function matchesHost(baseUrl: string, expectedHost: string): boolean {
+	try {
+		return new URL(baseUrl).hostname === expectedHost;
+	} catch {
+		return false;
+	}
+}
+
+function isOpenRouterModel(model: Model<Api>): boolean {
+	return model.provider === "openrouter" || model.baseUrl.includes(OPENROUTER_HOST);
+}
+
+function isNvidiaNimModel(model: Model<Api>): boolean {
+	return model.provider === "nvidia" || matchesHost(model.baseUrl, NVIDIA_NIM_HOST);
+}
+
+function isCloudflareModel(model: Model<Api>): boolean {
+	return (
+		model.provider === "cloudflare-workers-ai" ||
+		model.provider === "cloudflare-ai-gateway" ||
+		matchesHost(model.baseUrl, CLOUDFLARE_API_HOST) ||
+		matchesHost(model.baseUrl, CLOUDFLARE_AI_GATEWAY_HOST)
+	);
+}
+
+function getDefaultAttributionHeaders(
+	model: Model<Api>,
+	settingsManager: SettingsManager,
+): Record<string, string> | undefined {
+	if (!isInstallTelemetryEnabled(settingsManager)) {
+		return undefined;
+	}
+
+	if (isOpenRouterModel(model)) {
+		return {
+			"HTTP-Referer": "https://pi.dev",
+			"X-OpenRouter-Title": "pi",
+			"X-OpenRouter-Categories": "cli-agent",
+		};
+	}
+
+	if (isNvidiaNimModel(model)) {
+		return {
+			"X-BILLING-INVOKE-ORIGIN": "Pi",
+		};
+	}
+
+	if (isCloudflareModel(model)) {
+		return {
+			"User-Agent": "pi-coding-agent",
+		};
+	}
+
+	return undefined;
+}
+
+function getSessionHeaders(model: Model<Api>, sessionId: string | undefined): Record<string, string> | undefined {
+	if (!sessionId) return undefined;
+	if (
+		model.provider !== "opencode" &&
+		model.provider !== "opencode-go" &&
+		!matchesHost(model.baseUrl, OPENCODE_HOST)
+	) {
+		return undefined;
+	}
+	return { "x-opencode-session": sessionId, "x-opencode-client": "pi" };
+}
+
+export function mergeProviderAttributionHeaders(
+	model: Model<Api>,
+	settingsManager: SettingsManager,
+	sessionId: string | undefined,
+	...headerSources: Array<Record<string, string> | undefined>
+): Record<string, string> | undefined {
+	const merged = {
+		...getSessionHeaders(model, sessionId),
+		...getDefaultAttributionHeaders(model, settingsManager),
+	};
+
+	for (const headers of headerSources) {
+		if (headers) {
+			Object.assign(merged, headers);
+		}
+	}
+
+	return Object.keys(merged).length > 0 ? merged : undefined;
+}

packages/coding-agent/src/core/provider-display-names.ts

@@ -17,6 +17,7 @@ export const BUILT_IN_PROVIDER_DISPLAY_NAMES: Record<string, string> = {
 	"minimax-cn": "MiniMax (China)",
 	moonshotai: "Moonshot AI",
 	"moonshotai-cn": "Moonshot AI (China)",
+	nvidia: "NVIDIA NIM",
 	opencode: "OpenCode Zen",
 	"opencode-go": "OpenCode Go",
 	openai: "OpenAI",

packages/coding-agent/src/core/sdk.ts

@@ -11,11 +11,11 @@ import type { ExtensionRunner, LoadExtensionsResult, SessionStartEvent, ToolDefi
 import { convertToLlm } from "./messages.ts";
 import { ModelRegistry } from "./model-registry.ts";
 import { findInitialModel } from "./model-resolver.ts";
+import { mergeProviderAttributionHeaders } from "./provider-attribution.ts";
 import type { ResourceLoader } from "./resource-loader.ts";
 import { DefaultResourceLoader } from "./resource-loader.ts";
 import { getDefaultSessionDir, SessionManager } from "./session-manager.ts";
 import { SettingsManager } from "./settings-manager.ts";
-import { isInstallTelemetryEnabled } from "./telemetry.ts";
 import { time } from "./timings.ts";
 import {
 	createBashTool,
@@ -128,44 +128,6 @@ function getDefaultAgentDir(): string {
 	return getAgentDir();
 }
 
-function getAttributionHeaders(
-	model: Model<any>,
-	settingsManager: SettingsManager,
-	sessionId?: string,
-): Record<string, string> | undefined {
-	if (
-		sessionId &&
-		(model.provider === "opencode" || model.provider === "opencode-go" || model.baseUrl.includes("opencode.ai"))
-	) {
-		return { "x-opencode-session": sessionId, "x-opencode-client": "pi" };
-	}
-
-	if (!isInstallTelemetryEnabled(settingsManager)) {
-		return undefined;
-	}
-
-	if (model.provider === "openrouter" || model.baseUrl.includes("openrouter.ai")) {
-		return {
-			"HTTP-Referer": "https://pi.dev",
-			"X-OpenRouter-Title": "pi",
-			"X-OpenRouter-Categories": "cli-agent",
-		};
-	}
-
-	if (
-		model.provider === "cloudflare-workers-ai" ||
-		model.provider === "cloudflare-ai-gateway" ||
-		model.baseUrl.includes("api.cloudflare.com") ||
-		model.baseUrl.includes("gateway.ai.cloudflare.com")
-	) {
-		return {
-			"User-Agent": "pi-coding-agent",
-		};
-	}
-
-	return undefined;
-}
-
 /**
  * Create an AgentSession with the specified options.
  *
@@ -349,7 +311,6 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			const timeoutMs = options?.timeoutMs ?? providerRetrySettings.timeoutMs ?? effectiveTimeoutMs;
 			const websocketConnectTimeoutMs =
 				options?.websocketConnectTimeoutMs ?? settingsManager.getWebSocketConnectTimeoutMs();
-			const attributionHeaders = getAttributionHeaders(model, settingsManager, options?.sessionId);
 			return streamSimple(model, context, {
 				...options,
 				apiKey: auth.apiKey,
@@ -357,10 +318,13 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 				websocketConnectTimeoutMs,
 				maxRetries: options?.maxRetries ?? providerRetrySettings.maxRetries,
 				maxRetryDelayMs: options?.maxRetryDelayMs ?? providerRetrySettings.maxRetryDelayMs,
-				headers:
+				headers: mergeProviderAttributionHeaders(
-					attributionHeaders || auth.headers || options?.headers
+					model,
-						? { ...attributionHeaders, ...auth.headers, ...options?.headers }
+					settingsManager,
-						: undefined,
+					options?.sessionId,
+					auth.headers,
+					options?.headers,
+				),
 			});
 		},
 		onPayload: async (payload, _model) => {

packages/coding-agent/test/sdk-openrouter-attribution.test.ts

@@ -15,14 +15,14 @@ import { createAgentSession } from "../src/core/sdk.ts";
 import { SessionManager } from "../src/core/session-manager.ts";
 import { SettingsManager } from "../src/core/settings-manager.ts";
 
-describe("createAgentSession OpenRouter attribution headers", () => {
+describe("createAgentSession provider attribution headers", () => {
 	let tempDir: string;
 	let cwd: string;
 	let agentDir: string;
 	let originalTelemetryEnv: string | undefined;
 
 	beforeEach(() => {
-		tempDir = join(tmpdir(), `pi-sdk-openrouter-test-${Date.now()}-${Math.random().toString(36).slice(2)}`);
+		tempDir = join(tmpdir(), `pi-sdk-attribution-test-${Date.now()}-${Math.random().toString(36).slice(2)}`);
 		cwd = join(tempDir, "project");
 		agentDir = join(tempDir, "agent");
 		mkdirSync(cwd, { recursive: true });
@@ -42,9 +42,9 @@ describe("createAgentSession OpenRouter attribution headers", () => {
 		}
 	});
 
-	function createModel(provider: string, baseUrl: string): Model<Api> {
+	function createModel(provider: string, baseUrl: string, id = `${provider}-test-model`): Model<Api> {
 		return {
-			id: `${provider}-test-model`,
+			id,
 			name: `${provider} Test Model`,
 			api: "openai-completions",
 			provider,
@@ -172,6 +172,14 @@ describe("createAgentSession OpenRouter attribution headers", () => {
 		expect(headers?.["X-OpenRouter-Categories"]).toBe("cli-agent");
 	});
 
+	it("preserves legacy OpenRouter base URL substring attribution matching", async () => {
+		const headers = await captureHeaders(createModel("custom-openrouter", "not-a-url-openrouter.ai"));
+
+		expect(headers?.["HTTP-Referer"]).toBe("https://pi.dev");
+		expect(headers?.["X-OpenRouter-Title"]).toBe("pi");
+		expect(headers?.["X-OpenRouter-Categories"]).toBe("cli-agent");
+	});
+
 	it("lets provider and request headers override the defaults", async () => {
 		const headers = await captureHeaders(createModel("openrouter", "https://openrouter.ai/api/v1"), {
 			providerHeaders: {
@@ -188,6 +196,56 @@ describe("createAgentSession OpenRouter attribution headers", () => {
 		expect(headers?.["X-OpenRouter-Categories"]).toBe("provider-category");
 	});
 
+	it("adds default attribution headers for direct NVIDIA NIM endpoints", async () => {
+		const headers = await captureHeaders(createModel("custom-nim", "https://integrate.api.nvidia.com/v1"));
+
+		expect(headers?.["X-BILLING-INVOKE-ORIGIN"]).toBe("Pi");
+	});
+
+	it("adds default attribution headers for the NVIDIA provider", async () => {
+		const headers = await captureHeaders(createModel("nvidia", "https://example.test/v1"));
+
+		expect(headers?.["X-BILLING-INVOKE-ORIGIN"]).toBe("Pi");
+	});
+
+	it("does not add NVIDIA NIM attribution headers when telemetry is disabled", async () => {
+		const headers = await captureHeaders(createModel("nvidia", "https://integrate.api.nvidia.com/v1"), {
+			telemetryEnabled: false,
+		});
+
+		expect(headers?.["X-BILLING-INVOKE-ORIGIN"]).toBeUndefined();
+	});
+
+	it("lets provider and request headers override NVIDIA NIM defaults", async () => {
+		const headers = await captureHeaders(createModel("nvidia", "https://integrate.api.nvidia.com/v1"), {
+			providerHeaders: {
+				"X-BILLING-INVOKE-ORIGIN": "Provider",
+			},
+			requestHeaders: {
+				"X-BILLING-INVOKE-ORIGIN": "Request",
+			},
+		});
+
+		expect(headers?.["X-BILLING-INVOKE-ORIGIN"]).toBe("Request");
+	});
+
+	it("does not add NVIDIA NIM attribution headers for NVIDIA models routed through OpenRouter", async () => {
+		const headers = await captureHeaders(
+			createModel("openrouter", "https://openrouter.ai/api/v1", "nvidia/nemotron-3-super-120b-a12b"),
+		);
+
+		expect(headers?.["HTTP-Referer"]).toBe("https://pi.dev");
+		expect(headers?.["X-BILLING-INVOKE-ORIGIN"]).toBeUndefined();
+	});
+
+	it("does not add NVIDIA NIM attribution headers for NVIDIA models routed through Vercel AI Gateway", async () => {
+		const headers = await captureHeaders(
+			createModel("vercel-ai-gateway", "https://ai-gateway.vercel.sh/v1", "nvidia/nemotron-3-super-120b-a12b"),
+		);
+
+		expect(headers?.["X-BILLING-INVOKE-ORIGIN"]).toBeUndefined();
+	});
+
 	it("adds OpenCode session headers", async () => {
 		const headers = await captureHeaders(createModel("opencode", "https://opencode.ai/zen/v1"), {
 			sessionId: "opencode-session",