--- updatedAt: 2026-08-06T09:59:50.000Z --- # Request response from the model Invokes inference using the model chat parameters. # OpenAPI definition ```json { "openapi": "3.1.0", "info": { "title": "NVIDIA NIM API for google/diffusiongemma-26b-a4b-it", "description": "The NVIDIA NIM REST API. Please see https://docs.api.nvidia.com/nim/reference/google-diffusiongemma-26b-a4b-it for more details.", "version": "1.0.0", "termsOfService": "https://nvidia.com/legal/terms-of-use", "contact": { "name": "NVIDIA Support", "url": "https://help.nvidia.com/" }, "license": { "name": "Gemma Terms of Use", "url": "https://ai.google.dev/gemma/terms" } }, "servers": [ { "url": "https://integrate.api.nvidia.com/v1" } ], "tags": [ { "name": "Multimodal API", "description": "This API performs inference using visual language understanding models" } ], "paths": { "/chat/completions": { "post": { "tags": [ "Multimodal API" ], "summary": "Request response from the model", "description": "Invokes inference using the model chat parameters.", "operationId": "invokeFunction", "requestBody": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/NIMLLMChatCompletionRequest" } } }, "required": true }, "responses": { "200": { "description": "Invocation is fulfilled", "content": { "application/json": { "schema": { "$ref": "#/components/schemas/ChatCompletionResponse" } }, "text/event-stream": { "schema": { "$ref": "#/components/schemas/ChatCompletionStreamResponse" } } } }, "202": { "description": "Result is pending. Client should poll using the requestId.\n", "content": { "application/json": { "example": {}, "schema": {} } }, "headers": { "NVCF-REQID": { "description": "requestId required for pooling", "schema": { "type": "string", "format": "uuid", "maxLength": 36 } }, "NVCF-STATUS": { "description": "Invocation status", "schema": { "type": "string", "format": "^[a-zA-Z-]{1,64}$", "maxLength": 64 } } } }, "422": { "description": "Validation failed, provided entity could not be processed.", "content": { "application/json": { "schema": { "$ref": "#/components/schemas/ErrorResponse" }, "example": { "type": "urn:nvcf-worker-service:problem-details:unprocessable-entity", "title": "Unprocessable Entity", "status": 422, "detail": "string", "instance": "/v2/nvcf/pexec/functions/e598bfc1-b058-41af-869d-556d3c7e1b48", "requestId": "3fa85f64-5717-4562-b3fc-2c963f66afa6" } } } }, "500": { "description": "The invocation ended with an error.", "content": { "application/json": { "schema": { "$ref": "#/components/schemas/ErrorResponse" }, "example": { "type": "urn:nvcf-worker-service:problem-details:internal-server-error", "title": "Internal Server Error", "status": 500, "detail": "string", "instance": "/v2/nvcf/pexec/functions/e598bfc1-b058-41af-869d-556d3c7e1b48", "requestId": "3fa85f64-5717-4562-b3fc-2c963f66afa6" } } } } }, "x-nvai-meta": { "name": "Create chat completion", "returns": "Returns a [chat completion](/docs/api-reference/chat/object) object, or a streamed sequence of [chat completion chunk](/docs/api-reference/chat/streaming) objects if the request is streamed.\n", "path": "create", "examples": [ { "name": "dogs.jpeg", "input": { "text": "What is in this image?", "images": [ "https://assets.ngc.nvidia.com/products/api-catalog/phi-3-5-vision/example1b.jpg" ] }, "output": { "text": "The image showcases a serene landscape featuring a wooden boardwalk that cuts through a vast, grassy area. The boardwalk is the central focus, inviting viewers to imagine strolling through this peaceful environment. The grass surrounding the boardwalk is very tall, giving a sense of immersion in nature. In the distance, a variety of trees can be seen, their varied foliage adding complexity to the scene." }, "requestJson": "{ \"model\": \"google/diffusiongemma-26b-a4b-it\", \"messages\": [ { \"role\": \"user\", \"content\": [ { \"type\": \"text\", \"text\": \"<%- input.text %>\" }, { \"type\": \"image_url\", \"image_url\": { \"url\": \"<%- input.images[0] %>\" } } ] } ], \"max_tokens\": 1024, \"temperature\": 1.0, \"top_p\": 0.95, \"stream\": false }" }, { "name": "video.mp4", "input": { "text": "Describe what you see in this video.", "videos": [ "https://assets.ngc.nvidia.com/products/api-catalog/nvidia-nemotron-nano-12b-v2-vl/nvidia-studio-itns-wk53-scene-in-omniverse-1280w.mp4" ] }, "output": { "text": "The video displays a 3D model of a small, rustic cabin with a thatched roof, situated in a snowy landscape. The cabin is constructed with wooden planks and has a stone chimney. Surrounding the cabin are various objects such as logs, a wooden table, and a barrel." }, "requestJson": "{ \"model\": \"google/diffusiongemma-26b-a4b-it\", \"messages\": [ { \"role\": \"user\", \"content\": [ { \"type\": \"text\", \"text\": \"<%- input.text %>\" }, { \"type\": \"video_url\", \"video_url\": { \"url\": \"<%- input.videos[0] %>\" } } ] } ], \"max_tokens\": 1024, \"temperature\": 1.0, \"top_p\": 0.95, \"stream\": false }" } ], "templates": [ { "title": "Default", "response": "{\n \"choices\": [{\n \"index\": 0,\n \"model\": \"google/diffusiongemma-26b-a4b-it\",\n \"message\": {\n \"role\": \"assistant\",\n \"content\": \"...\",\n },\n \"finish_reason\": \"stop\"\n }],\n \"usage\": {\n \"prompt_tokens\": 9,\n \"completion_tokens\": 12,\n \"total_tokens\": 21\n }\n}\n", "requestEjs": { "python": "<%\n // Catalog ejs locals replace JSON with {stringify} only (spread skips non-enumerable parse).\n const deepClone = (value) => {\n if (Array.isArray(value)) return value.map(deepClone);\n if (value && typeof value === \"object\") {\n const out = {};\n for (const key of Object.keys(value)) out[key] = deepClone(value[key]);\n return out;\n }\n return value;\n };\n const mediaConfig = {\"image\": {\"mimeTypes\": {\".png\": \"image/png\", \".jpg\": \"image/jpeg\", \".jpeg\": \"image/jpeg\", \".webp\": \"image/webp\"}}, \"video\": {\"mimeTypes\": {\".mp4\": \"video/mp4\"}}, \"audio\": {\"mimeTypes\": {}, \"defaultFormat\": \"wav\"}};\n const stripNulls = (value) => {\n if (Array.isArray(value)) return value.map(stripNulls);\n if (value && typeof value === \"object\") {\n const out = {};\n for (const [key, item] of Object.entries(value)) {\n if (item === null || item === undefined) continue;\n out[key] = stripNulls(item);\n }\n return out;\n }\n return value;\n };\n const payloadForSnippet = stripNulls(deepClone(request || {}));\n const hasStream = payloadForSnippet.stream !== undefined && payloadForSnippet.stream !== null;\n const pyStream = payloadForSnippet.stream ? \"True\" : \"False\";\n if (hasStream) payloadForSnippet.stream = \"__STREAM__\";\n\n const hasMediaKind = (kind) => Object.keys(mediaConfig[kind].mimeTypes || {}).length > 0;\n const isInlineDataUrl = (url, kind) => (\n typeof url === \"string\" && new RegExp(`^data:${kind}/[^;]+;base64,`, \"i\").test(url)\n );\n\n const mediaUrl = (part, key) => {\n const slot = part?.[key];\n return typeof slot === \"string\" ? slot : slot?.url;\n };\n const setMediaUrl = (part, key, url) => {\n if (typeof part[key] === \"string\") part[key] = url;\n else {\n if (!part[key] || typeof part[key] !== \"object\") part[key] = {};\n part[key].url = url;\n }\n };\n\n let imageCount = 0;\n let videoCount = 0;\n let audioCount = 0;\n\n (payloadForSnippet.messages || []).forEach((message) => {\n if (!Array.isArray(message?.content)) return;\n message.content.forEach((part) => {\n const imageUrl = mediaUrl(part, \"image_url\");\n const videoUrl = mediaUrl(part, \"video_url\");\n const audioUrl = mediaUrl(part, \"audio_url\");\n // File helpers only for inline base64; http(s) URLs stay as-is in the payload.\n if (hasMediaKind(\"image\") && part?.type === \"image_url\" && isInlineDataUrl(imageUrl, \"image\")) {\n setMediaUrl(part, \"image_url\", `__IMAGE_DATA_URL_${imageCount}__`);\n imageCount += 1;\n } else if (hasMediaKind(\"video\") && part?.type === \"video_url\" && isInlineDataUrl(videoUrl, \"video\")) {\n setMediaUrl(part, \"video_url\", `__VIDEO_DATA_URL_${videoCount}__`);\n videoCount += 1;\n } else if (hasMediaKind(\"audio\") && part?.type === \"audio_url\" && isInlineDataUrl(audioUrl, \"audio\")) {\n setMediaUrl(part, \"audio_url\", `__AUDIO_DATA_URL_${audioCount}__`);\n audioCount += 1;\n } else if (hasMediaKind(\"audio\") && part?.type === \"input_audio\" && part.input_audio?.data) {\n part.input_audio.data = `__AUDIO_B64_${audioCount}__`;\n part.input_audio.format = part.input_audio.format || mediaConfig.audio.defaultFormat;\n audioCount += 1;\n }\n });\n });\n\n const pyRepr = (value, level = 0) => {\n if (value === null) return \"None\";\n if (typeof value === \"boolean\") return value ? \"True\" : \"False\";\n if (typeof value === \"number\") return String(value);\n if (typeof value === \"string\") return JSON.stringify(value);\n const pad = \" \".repeat(level);\n const innerPad = \" \".repeat(level + 2);\n if (Array.isArray(value)) {\n if (value.length === 0) return \"[]\";\n return \"[\\n\" + value.map((item) => `${innerPad}${pyRepr(item, level + 2)}`).join(\",\\n\") + \"\\n\" + pad + \"]\";\n }\n if (typeof value === \"object\") {\n const entries = Object.entries(value);\n if (entries.length === 0) return \"{}\";\n return \"{\\n\" + entries.map(([key, item]) => `${innerPad}${JSON.stringify(key)}: ${pyRepr(item, level + 2)}`).join(\",\\n\") + \"\\n\" + pad + \"}\";\n }\n return \"None\";\n };\n\n const replaceAll = (s, a, b) => s.split(a).join(b);\n let renderedPayload = pyRepr(payloadForSnippet);\n if (hasStream) renderedPayload = replaceAll(renderedPayload, '\"__STREAM__\"', \"stream\");\n for (let i = 0; i < imageCount; i++) {\n renderedPayload = replaceAll(renderedPayload, `\"__IMAGE_DATA_URL_${i}__\"`, `image_data_urls[${i}]`);\n }\n for (let i = 0; i < videoCount; i++) {\n renderedPayload = replaceAll(renderedPayload, `\"__VIDEO_DATA_URL_${i}__\"`, `video_data_urls[${i}]`);\n }\n for (let i = 0; i < audioCount; i++) {\n renderedPayload = replaceAll(renderedPayload, `\"__AUDIO_DATA_URL_${i}__\"`, `audio_data_urls[${i}]`);\n renderedPayload = replaceAll(renderedPayload, `\"__AUDIO_B64_${i}__\"`, `audio_b64s[${i}]`);\n }\n\n // Either file helpers XOR inline base64 - never emit unused helpers.\n const usesFileHelpers =\n imageCount + videoCount + audioCount > 0 &&\n [...Array(imageCount).keys()].every((i) => renderedPayload.includes(`image_data_urls[${i}]`)) &&\n [...Array(videoCount).keys()].every((i) => renderedPayload.includes(`video_data_urls[${i}]`)) &&\n [...Array(audioCount).keys()].every(\n (i) => renderedPayload.includes(`audio_data_urls[${i}]`) || renderedPayload.includes(`audio_b64s[${i}]`)\n );\n if (!usesFileHelpers) {\n imageCount = 0;\n videoCount = 0;\n audioCount = 0;\n const restored = stripNulls(deepClone(request || {}));\n const restoredStream = restored.stream !== undefined && restored.stream !== null;\n if (restoredStream) restored.stream = \"__STREAM__\";\n renderedPayload = pyRepr(restored);\n if (restoredStream) renderedPayload = replaceAll(renderedPayload, '\"__STREAM__\"', \"stream\");\n }\n const hasMedia = usesFileHelpers;\n const mimePy = (mimes) =>\n \"{\" + Object.entries(mimes).map(([k, v]) => `${JSON.stringify(k)}: ${JSON.stringify(v)}`).join(\", \") + \"}\";\n%>\n<% if (hasMedia) { %>import base64\nfrom pathlib import Path\n<% } %>import requests\n\ninvoke_url = \"https://integrate.api.nvidia.com/v1/chat/completions\"\nstream = <%- pyStream %>\n<% if (hasMedia) { %>\ndef media_data_url(kind, index, mime_types):\n for suffix, mime in mime_types.items():\n path = Path(f\"{kind}_{index}{suffix}\")\n if path.exists():\n return f\"data:{mime};base64,{base64.b64encode(path.read_bytes()).decode()}\"\n raise FileNotFoundError(f\"Expected {kind}_{index} with suffix in {list(mime_types)}\")\n<% if (imageCount > 0) { %>\nimage_data_urls = [media_data_url(\"image\", i, <%- mimePy(mediaConfig.image.mimeTypes) %>) for i in <%- imageCount === 1 ? \"[1]\" : (\"range(1, \" + (imageCount + 1) + \")\") %>]\n<% } %><% if (videoCount > 0) { %>\nvideo_data_urls = [media_data_url(\"video\", i, <%- mimePy(mediaConfig.video.mimeTypes) %>) for i in <%- videoCount === 1 ? \"[1]\" : (\"range(1, \" + (videoCount + 1) + \")\") %>]\n<% } %><% if (audioCount > 0) { %>\naudio_data_urls = [media_data_url(\"audio\", i, <%- mimePy(mediaConfig.audio.mimeTypes) %>) for i in <%- audioCount === 1 ? \"[1]\" : (\"range(1, \" + (audioCount + 1) + \")\") %>]\naudio_b64s = [url.split(\",\", 1)[1] for url in audio_data_urls]\n<% } %><% } %>\nheaders = {\n \"Authorization\": \"Bearer $NVIDIA_API_KEY\",\n \"Accept\": \"text/event-stream\" if stream else \"application/json\",\n}\n\npayload = <%- renderedPayload %>\n\nresponse = requests.post(invoke_url, headers=headers, json=payload, stream=stream)\nif stream:\n for line in response.iter_lines():\n if line:\n print(line.decode(\"utf-8\"))\nelse:\n print(response.json())\n", "langChain": "<%\n const deepClone = (value) => {\n if (Array.isArray(value)) return value.map(deepClone);\n if (value && typeof value === \"object\") {\n const out = {};\n for (const key of Object.keys(value)) out[key] = deepClone(value[key]);\n return out;\n }\n return value;\n };\n const mediaConfig = {\"image\": {\"mimeTypes\": {\".png\": \"image/png\", \".jpg\": \"image/jpeg\", \".jpeg\": \"image/jpeg\", \".webp\": \"image/webp\"}}, \"video\": {\"mimeTypes\": {\".mp4\": \"video/mp4\"}}, \"audio\": {\"mimeTypes\": {}, \"defaultFormat\": \"wav\"}};\n const payloadForSnippet = deepClone(request || {});\n const messagesForSnippet = deepClone(payloadForSnippet.messages || []);\n const hasMediaKind = (kind) => Object.keys(mediaConfig[kind].mimeTypes || {}).length > 0;\n const isInlineDataUrl = (url, kind) => (\n typeof url === \"string\" && new RegExp(`^data:${kind}/[^;]+;base64,`, \"i\").test(url)\n );\n\n const mediaUrl = (part, key) => {\n const slot = part?.[key];\n return typeof slot === \"string\" ? slot : slot?.url;\n };\n const setMediaUrl = (part, key, url) => {\n if (typeof part[key] === \"string\") part[key] = url;\n else {\n if (!part[key] || typeof part[key] !== \"object\") part[key] = {};\n part[key].url = url;\n }\n };\n\n let imageCount = 0;\n let hasUnsupportedMedia = false;\n messagesForSnippet.forEach((message) => {\n if (!Array.isArray(message?.content)) return;\n message.content.forEach((part) => {\n const imageUrl = mediaUrl(part, \"image_url\");\n if (hasMediaKind(\"image\") && part?.type === \"image_url\" && isInlineDataUrl(imageUrl, \"image\")) {\n setMediaUrl(part, \"image_url\", `__IMAGE_DATA_URL_${imageCount}__`);\n imageCount += 1;\n } else if (part?.type === \"video_url\" && isInlineDataUrl(mediaUrl(part, \"video_url\"), \"video\")) {\n // LangChain cannot ship local file helpers for video; reject inline only.\n hasUnsupportedMedia = true;\n setMediaUrl(part, \"video_url\", \"__UNSUPPORTED_MEDIA_DATA__\");\n } else if (part?.type === \"audio_url\" && isInlineDataUrl(mediaUrl(part, \"audio_url\"), \"audio\")) {\n hasUnsupportedMedia = true;\n setMediaUrl(part, \"audio_url\", \"__UNSUPPORTED_MEDIA_DATA__\");\n } else if (part?.type === \"input_audio\" && part.input_audio?.data) {\n hasUnsupportedMedia = true;\n part.input_audio.data = \"__UNSUPPORTED_MEDIA_DATA__\";\n }\n });\n });\n\n const pyRepr = (value, level = 0) => {\n if (value === null) return \"None\";\n if (typeof value === \"boolean\") return value ? \"True\" : \"False\";\n if (typeof value === \"number\") return String(value);\n if (typeof value === \"string\") return JSON.stringify(value);\n const pad = \" \".repeat(level);\n const innerPad = \" \".repeat(level + 2);\n if (Array.isArray(value)) {\n if (value.length === 0) return \"[]\";\n return \"[\\n\" + value.map((item) => `${innerPad}${pyRepr(item, level + 2)},`).join(\"\\n\") + \"\\n\" + pad + \"]\";\n }\n if (typeof value === \"object\") {\n const entries = Object.entries(value);\n if (entries.length === 0) return \"{}\";\n return \"{\\n\" + entries.map(([key, item]) => `${innerPad}${JSON.stringify(key)}: ${pyRepr(item, level + 2)},`).join(\"\\n\") + \"\\n\" + pad + \"}\";\n }\n return \"None\";\n };\n\n const replaceAll = (s, a, b) => s.split(a).join(b);\n let renderedMessages = pyRepr(messagesForSnippet);\n for (let i = 0; i < imageCount; i++) {\n renderedMessages = replaceAll(renderedMessages, `\"__IMAGE_DATA_URL_${i}__\"`, `image_data_urls[${i}]`);\n }\n // Either file helpers XOR inline base64 - never emit unused helpers.\n const usesFileHelpers =\n imageCount > 0 &&\n [...Array(imageCount).keys()].every((i) => renderedMessages.includes(`image_data_urls[${i}]`));\n if (!usesFileHelpers) {\n imageCount = 0;\n renderedMessages = pyRepr(deepClone((request || {}).messages || []));\n }\n%>\nfrom langchain_nvidia_ai_endpoints import ChatNVIDIA\n<% if (imageCount > 0) { %>from pathlib import Path\nimport base64\n<% } %>\n\nclient = ChatNVIDIA(\n model=\"<%- request.model %>\",\n api_key=\"$NVIDIA_API_KEY\",\n temperature=<%- request.temperature %>,\n top_p=<%- request.top_p %>,\n max_completion_tokens=<%- request.max_tokens %>,\n)\n<% if (imageCount > 0) { %>\ndef find_image_path(index):\n image_mime_types = <%- pyRepr(mediaConfig.image.mimeTypes) %>\n for suffix in image_mime_types:\n path = Path(f\"image_{index}{suffix}\")\n if path.exists():\n return path\n expected = \"/\".join(f\".{suffix.lstrip('.')}\" for suffix in image_mime_types)\n raise FileNotFoundError(f\"Expected image_{index}{expected}\")\n\ndef read_image_data_url(path):\n image_mime_types = <%- pyRepr(mediaConfig.image.mimeTypes) %>\n with open(path, \"rb\") as f:\n image_b64 = base64.b64encode(f.read()).decode()\n return f\"data:{image_mime_types[path.suffix.lower()]};base64,{image_b64}\"\n\nimage_data_urls = [read_image_data_url(find_image_path(i)) for i in range(1, <%- imageCount + 1 %>)]\n<% } %><% if (hasUnsupportedMedia) { %>\nraise ValueError(\"LangChain snippet currently supports text and image input. Use Python, Node.js, or Shell for video or audio input.\")\n<% } %>\nlc_messages = <%- renderedMessages %>\n<% if (request.stream) { %>\nfor chunk in client.stream(lc_messages<%= request.chat_template_kwargs ? ', chat_template_kwargs=' + pyRepr(request.chat_template_kwargs) : '' %>):\n if chunk.additional_kwargs and \"reasoning_content\" in chunk.additional_kwargs:\n print(chunk.additional_kwargs[\"reasoning_content\"], end=\"\")\n print(chunk.content, end=\"\")\n<% } else { %>\nresponse = client.invoke(lc_messages<%= request.chat_template_kwargs ? ', chat_template_kwargs=' + pyRepr(request.chat_template_kwargs) : '' %>)\nif response.additional_kwargs and \"reasoning_content\" in response.additional_kwargs:\n print(response.additional_kwargs[\"reasoning_content\"])\nprint(response.content)\n<% } %>\n", "node.js": "<%\n const deepClone = (value) => {\n if (Array.isArray(value)) return value.map(deepClone);\n if (value && typeof value === \"object\") {\n const out = {};\n for (const key of Object.keys(value)) out[key] = deepClone(value[key]);\n return out;\n }\n return value;\n };\n const mediaConfig = {\"image\": {\"mimeTypes\": {\".png\": \"image/png\", \".jpg\": \"image/jpeg\", \".jpeg\": \"image/jpeg\", \".webp\": \"image/webp\"}}, \"video\": {\"mimeTypes\": {\".mp4\": \"video/mp4\"}}, \"audio\": {\"mimeTypes\": {}, \"defaultFormat\": \"wav\"}};\n const payloadForSnippet = deepClone(request || {});\n const hasStream = payloadForSnippet.stream !== undefined && payloadForSnippet.stream !== null;\n if (hasStream) payloadForSnippet.stream = \"__STREAM__\";\n\n const hasMediaKind = (kind) => Object.keys(mediaConfig[kind].mimeTypes || {}).length > 0;\n const isInlineDataUrl = (url, kind) => (\n typeof url === \"string\" && new RegExp(`^data:${kind}/[^;]+;base64,`, \"i\").test(url)\n );\n\n let imageCount = 0;\n let videoCount = 0;\n let audioCount = 0;\n (payloadForSnippet.messages || []).forEach((message) => {\n if (!Array.isArray(message?.content)) return;\n message.content.forEach((part) => {\n if (hasMediaKind(\"image\") && part?.type === \"image_url\" && isInlineDataUrl(part.image_url?.url, \"image\")) {\n part.image_url.url = `__IMAGE_DATA_URL_${imageCount}__`;\n imageCount += 1;\n } else if (hasMediaKind(\"video\") && part?.type === \"video_url\" && isInlineDataUrl(part.video_url?.url, \"video\")) {\n part.video_url.url = `__VIDEO_DATA_URL_${videoCount}__`;\n videoCount += 1;\n } else if (hasMediaKind(\"audio\") && part?.type === \"audio_url\" && isInlineDataUrl(part.audio_url?.url, \"audio\")) {\n part.audio_url.url = `__AUDIO_DATA_URL_${audioCount}__`;\n audioCount += 1;\n } else if (hasMediaKind(\"audio\") && part?.type === \"input_audio\" && part.input_audio?.data) {\n part.input_audio.data = `__AUDIO_B64_${audioCount}__`;\n part.input_audio.format = part.input_audio.format || mediaConfig.audio.defaultFormat;\n audioCount += 1;\n }\n });\n });\n\n const replaceAll = (s, a, b) => s.split(a).join(b);\n let renderedPayload = JSON.stringify(payloadForSnippet, null, 2);\n if (hasStream) renderedPayload = replaceAll(renderedPayload, '\"__STREAM__\"', \"stream\");\n for (let i = 0; i < imageCount; i++) {\n renderedPayload = replaceAll(renderedPayload, `\"__IMAGE_DATA_URL_${i}__\"`, `imageDataUrls[${i}]`);\n }\n for (let i = 0; i < videoCount; i++) {\n renderedPayload = replaceAll(renderedPayload, `\"__VIDEO_DATA_URL_${i}__\"`, `videoDataUrls[${i}]`);\n }\n for (let i = 0; i < audioCount; i++) {\n renderedPayload = replaceAll(renderedPayload, `\"__AUDIO_DATA_URL_${i}__\"`, `audioDataUrls[${i}]`);\n renderedPayload = replaceAll(renderedPayload, `\"__AUDIO_B64_${i}__\"`, `audioB64s[${i}]`);\n }\n // Either file helpers XOR inline base64 - never emit unused helpers.\n const usesFileHelpers =\n imageCount + videoCount + audioCount > 0 &&\n [...Array(imageCount).keys()].every((i) => renderedPayload.includes(`imageDataUrls[${i}]`)) &&\n [...Array(videoCount).keys()].every((i) => renderedPayload.includes(`videoDataUrls[${i}]`)) &&\n [...Array(audioCount).keys()].every(\n (i) => renderedPayload.includes(`audioDataUrls[${i}]`) || renderedPayload.includes(`audioB64s[${i}]`)\n );\n if (!usesFileHelpers) {\n imageCount = 0;\n videoCount = 0;\n audioCount = 0;\n const restored = deepClone(request || {});\n const restoredStream = restored.stream !== undefined && restored.stream !== null;\n if (restoredStream) restored.stream = \"__STREAM__\";\n renderedPayload = JSON.stringify(restored, null, 2);\n if (restoredStream) renderedPayload = replaceAll(renderedPayload, '\"__STREAM__\"', \"stream\");\n }\n const hasMedia = usesFileHelpers;\n%>\nimport axios from 'axios';\n<% if (hasMedia) { %>import { existsSync } from 'node:fs';\nimport { readFile } from 'node:fs/promises';\nimport { extname } from 'node:path';\n<% } %>\n\nconst invokeUrl = \"https://integrate.api.nvidia.com/v1/chat/completions\";\nconst stream = <%- request.stream %>;\n\nconst headers = {\n \"Authorization\": \"Bearer $NVIDIA_API_KEY\",\n \"Accept\": stream ? \"text/event-stream\" : \"application/json\"\n};\n<% if (hasMedia) { %>\nfunction findMediaPath(kind, index, mimeTypes) {\n for (const suffix of Object.keys(mimeTypes)) {\n const path = `${kind}_${index}${suffix}`;\n if (existsSync(path)) {\n return path;\n }\n }\n const expected = Object.keys(mimeTypes).join(\"/\");\n throw new Error(`Expected ${kind}_${index}${expected}`);\n}\n\nasync function readMediaB64(path) {\n return Buffer.from(await readFile(path)).toString('base64');\n}\n\nfunction mediaDataUrl(path, mediaB64, mimeTypes) {\n return `data:${mimeTypes[extname(path).toLowerCase()]};base64,${mediaB64}`;\n}\n<% } %>\nasync function main() {\n<% if (imageCount > 0) { %> const imageMimeTypes = <%- JSON.stringify(mediaConfig.image.mimeTypes, null, 2) %>;\n const imagePaths = [<% for (let i = 1; i <= imageCount; i++) { %>findMediaPath(\"image\", <%- i %>, imageMimeTypes)<%- i < imageCount ? \", \" : \"\" %><% } %>];\n const imageB64s = await Promise.all(imagePaths.map(readMediaB64));\n const imageDataUrls = imagePaths.map((path, index) => mediaDataUrl(path, imageB64s[index], imageMimeTypes));\n<% } %><% if (videoCount > 0) { %> const videoMimeTypes = <%- JSON.stringify(mediaConfig.video.mimeTypes, null, 2) %>;\n const videoPaths = [<% for (let i = 1; i <= videoCount; i++) { %>findMediaPath(\"video\", <%- i %>, videoMimeTypes)<%- i < videoCount ? \", \" : \"\" %><% } %>];\n const videoB64s = await Promise.all(videoPaths.map(readMediaB64));\n const videoDataUrls = videoPaths.map((path, index) => mediaDataUrl(path, videoB64s[index], videoMimeTypes));\n<% } %><% if (audioCount > 0) { %> const audioMimeTypes = <%- JSON.stringify(mediaConfig.audio.mimeTypes, null, 2) %>;\n const audioPaths = [<% for (let i = 1; i <= audioCount; i++) { %>findMediaPath(\"audio\", <%- i %>, audioMimeTypes)<%- i < audioCount ? \", \" : \"\" %><% } %>];\n const audioB64s = await Promise.all(audioPaths.map(readMediaB64));\n const audioDataUrls = audioPaths.map((path, index) => mediaDataUrl(path, audioB64s[index], audioMimeTypes));\n<% } %> const payload = <%- renderedPayload %>;\n\n const response = await axios.post(invokeUrl, payload, {\n headers: headers,\n responseType: stream ? 'stream' : 'json'\n });\n\n if (stream) {\n response.data.on('data', (chunk) => {\n console.log(chunk.toString());\n });\n } else {\n console.log(JSON.stringify(response.data));\n }\n}\n\nmain().catch(error => {\n if (error.response) {\n console.error(`HTTP ${error.response.status}`);\n if (error.response.data?.on) {\n error.response.data.on('data', (chunk) => console.error(chunk.toString()));\n } else {\n console.error(error.response.data);\n }\n } else {\n console.error(error);\n }\n});\n", "curl": "<%\n const deepClone = (value) => {\n if (Array.isArray(value)) return value.map(deepClone);\n if (value && typeof value === \"object\") {\n const out = {};\n for (const key of Object.keys(value)) out[key] = deepClone(value[key]);\n return out;\n }\n return value;\n };\n const mediaConfig = {\"image\": {\"mimeTypes\": {\".png\": \"image/png\", \".jpg\": \"image/jpeg\", \".jpeg\": \"image/jpeg\", \".webp\": \"image/webp\"}}, \"video\": {\"mimeTypes\": {\".mp4\": \"video/mp4\"}}, \"audio\": {\"mimeTypes\": {}, \"defaultFormat\": \"wav\"}};\n const payloadForSnippet = deepClone(request || {});\n const hasMediaKind = (kind) => Object.keys(mediaConfig[kind].mimeTypes || {}).length > 0;\n const isInlineDataUrl = (url, kind) => (\n typeof url === \"string\" && new RegExp(`^data:${kind}/[^;]+;base64,`, \"i\").test(url)\n );\n\n let imageCount = 0;\n let videoCount = 0;\n let audioCount = 0;\n (payloadForSnippet.messages || []).forEach((message) => {\n if (!Array.isArray(message?.content)) return;\n message.content.forEach((part) => {\n if (hasMediaKind(\"image\") && part?.type === \"image_url\" && isInlineDataUrl(part.image_url?.url, \"image\")) {\n part.image_url.url = `__IMAGE_DATA_URL_${imageCount}__`;\n imageCount += 1;\n } else if (hasMediaKind(\"video\") && part?.type === \"video_url\" && isInlineDataUrl(part.video_url?.url, \"video\")) {\n part.video_url.url = `__VIDEO_DATA_URL_${videoCount}__`;\n videoCount += 1;\n } else if (hasMediaKind(\"audio\") && part?.type === \"audio_url\" && isInlineDataUrl(part.audio_url?.url, \"audio\")) {\n part.audio_url.url = `__AUDIO_DATA_URL_${audioCount}__`;\n audioCount += 1;\n } else if (hasMediaKind(\"audio\") && part?.type === \"input_audio\" && part.input_audio?.data) {\n part.input_audio.data = `__AUDIO_B64_${audioCount}__`;\n part.input_audio.format = part.input_audio.format || mediaConfig.audio.defaultFormat;\n audioCount += 1;\n }\n });\n });\n\n const replaceAll = (s, a, b) => s.split(a).join(b);\n let renderedPayload = JSON.stringify(payloadForSnippet, null, 2);\n for (let i = 0; i < imageCount; i++) {\n renderedPayload = replaceAll(renderedPayload, `\"__IMAGE_DATA_URL_${i}__\"`, `\"${\"${\"}image_${i + 1}_data_url}\"`);\n }\n for (let i = 0; i < videoCount; i++) {\n renderedPayload = replaceAll(renderedPayload, `\"__VIDEO_DATA_URL_${i}__\"`, `\"${\"${\"}video_${i + 1}_data_url}\"`);\n }\n for (let i = 0; i < audioCount; i++) {\n renderedPayload = replaceAll(renderedPayload, `\"__AUDIO_DATA_URL_${i}__\"`, `\"${\"${\"}audio_${i + 1}_data_url}\"`);\n renderedPayload = replaceAll(renderedPayload, `\"__AUDIO_B64_${i}__\"`, `\"${\"${\"}audio_${i + 1}_b64}\"`);\n }\n // Either file helpers XOR inline base64 - never emit unused helpers.\n const usesFileHelpers =\n imageCount + videoCount + audioCount > 0 &&\n [...Array(imageCount).keys()].every((i) => renderedPayload.includes(`image_${i + 1}_data_url`)) &&\n [...Array(videoCount).keys()].every((i) => renderedPayload.includes(`video_${i + 1}_data_url`)) &&\n [...Array(audioCount).keys()].every(\n (i) => renderedPayload.includes(`audio_${i + 1}_data_url`) || renderedPayload.includes(`audio_${i + 1}_b64`)\n );\n if (!usesFileHelpers) {\n imageCount = 0;\n videoCount = 0;\n audioCount = 0;\n renderedPayload = JSON.stringify(deepClone(request || {}), null, 2);\n }\n const hasMedia = usesFileHelpers;\n%>\nstream=<%- request.stream %>\nif [ \"$stream\" = true ]; then\n accept_header='Accept: text/event-stream'\nelse\n accept_header='Accept: application/json'\nfi\n<% if (hasMedia) { %>\nmime_type_for() {\n case \"$1\" in\n<% for (const [suffix, mimeType] of Object.entries({...mediaConfig.image.mimeTypes, ...mediaConfig.video.mimeTypes, ...mediaConfig.audio.mimeTypes})) { %> *<%- suffix %>) echo \"<%- mimeType %>\" ;;\n<% } %> *) echo \"Unsupported media file: $1\" >&2; return 1 ;;\n esac\n}\n\nfind_media_path() {\n local kind=\"$1\"\n local index=\"$2\"\n shift 2\n local ext path\n for ext in \"$@\"; do\n path=\"${kind}_${index}.${ext}\"\n if [ -f \"$path\" ]; then\n echo \"$path\"\n return 0\n fi\n done\n echo \"Expected ${kind}_${index}.$*\" >&2\n return 1\n}\n\nmedia_b64() {\n base64 < \"$1\" | tr -d '\\n'\n}\n\nmedia_data_url() {\n local path=\"$1\"\n local mime_type media_base64\n mime_type=$(mime_type_for \"$path\") || return 1\n media_base64=$(media_b64 \"$path\") || return 1\n printf 'data:%s;base64,%s' \"$mime_type\" \"$media_base64\"\n}\n<% for (let i = 1; i <= imageCount; i++) { %>\nimage_<%- i %>_path=$(find_media_path image <%- i %><% for (const suffix of Object.keys(mediaConfig.image.mimeTypes)) { %> <%- suffix.slice(1) %><% } %>)\nimage_<%- i %>_data_url=$(media_data_url \"$image_<%- i %>_path\")\n<% } %><% for (let i = 1; i <= videoCount; i++) { %>\nvideo_<%- i %>_path=$(find_media_path video <%- i %><% for (const suffix of Object.keys(mediaConfig.video.mimeTypes)) { %> <%- suffix.slice(1) %><% } %>)\nvideo_<%- i %>_data_url=$(media_data_url \"$video_<%- i %>_path\")\n<% } %><% for (let i = 1; i <= audioCount; i++) { %>\naudio_<%- i %>_path=$(find_media_path audio <%- i %><% for (const suffix of Object.keys(mediaConfig.audio.mimeTypes)) { %> <%- suffix.slice(1) %><% } %>)\naudio_<%- i %>_b64=$(media_b64 \"$audio_<%- i %>_path\")\naudio_<%- i %>_data_url=$(media_data_url \"$audio_<%- i %>_path\")\n<% } %><% } %>\ncat > payload.json <\nJSON\n\ncurl https://integrate.api.nvidia.com/v1/chat/completions \\\n -H \"Authorization: Bearer $NVIDIA_API_KEY\" \\\n -H \"Content-Type: application/json\" \\\n -H \"$accept_header\" \\\n -d @payload.json" } } ] } } } }, "security": [ { "Token": [] } ], "components": { "securitySchemes": { "Token": { "type": "http", "scheme": "bearer" } }, "schemas": { "ChatCompletionResponse": { "properties": { "id": { "type": "string", "title": "Id" }, "object": { "type": "string", "enum": [ "chat.completion" ], "const": "chat.completion", "title": "Object", "default": "chat.completion" }, "created": { "type": "integer", "title": "Created" }, "model": { "type": "string", "title": "Model" }, "choices": { "items": { "$ref": "#/components/schemas/ChatCompletionResponseChoice" }, "type": "array", "title": "Choices" }, "usage": { "$ref": "#/components/schemas/UsageInfo" } }, "additionalProperties": false, "type": "object", "required": [ "model", "choices", "usage" ], "title": "ChatCompletionResponse" }, "ChatCompletionResponseChoice": { "properties": { "index": { "type": "integer", "title": "Index" }, "message": { "$ref": "#/components/schemas/ChatMessage" }, "finish_reason": { "anyOf": [ { "type": "string", "enum": [ "stop", "length" ] }, { "type": "null" } ], "title": "Finish Reason" }, "stop_reason": { "anyOf": [ { "type": "integer" }, { "type": "string" }, { "type": "null" } ], "title": "Stop Reason" } }, "additionalProperties": false, "type": "object", "required": [ "index", "message" ], "title": "ChatCompletionResponseChoice" }, "ChatCompletionResponseStreamChoice": { "properties": { "index": { "type": "integer", "title": "Index" }, "delta": { "$ref": "#/components/schemas/DeltaMessage" }, "finish_reason": { "anyOf": [ { "type": "string", "enum": [ "stop", "length" ] }, { "type": "null" } ], "title": "Finish Reason" }, "stop_reason": { "anyOf": [ { "type": "integer" }, { "type": "string" }, { "type": "null" } ], "title": "Stop Reason" } }, "additionalProperties": false, "type": "object", "required": [ "index", "delta" ], "title": "ChatCompletionResponseStreamChoice" }, "ChatCompletionStreamResponse": { "properties": { "id": { "type": "string", "title": "Id" }, "object": { "type": "string", "enum": [ "chat.completion.chunk" ], "const": "chat.completion.chunk", "title": "Object", "default": "chat.completion.chunk" }, "created": { "type": "integer", "title": "Created" }, "model": { "type": "string", "title": "Model" }, "choices": { "items": { "$ref": "#/components/schemas/ChatCompletionResponseStreamChoice" }, "type": "array", "title": "Choices" }, "usage": { "anyOf": [ { "$ref": "#/components/schemas/UsageInfo" }, { "type": "null" } ] } }, "additionalProperties": false, "type": "object", "required": [ "model", "choices" ], "title": "ChatCompletionStreamResponse" }, "ChatMessage": { "properties": { "role": { "type": "string", "title": "Role" }, "content": { "anyOf": [ { "type": "string" }, { "type": "null" } ], "title": "Content" } }, "additionalProperties": false, "type": "object", "required": [ "role" ], "title": "ChatMessage" }, "DeltaMessage": { "properties": { "role": { "anyOf": [ { "type": "string" }, { "type": "null" } ], "title": "Role" }, "content": { "anyOf": [ { "type": "string" }, { "type": "null" } ], "title": "Content" } }, "additionalProperties": false, "type": "object", "title": "DeltaMessage" }, "ErrorResponse": { "properties": { "object": { "type": "string", "title": "Object", "default": "error" }, "message": { "type": "string", "title": "Message" }, "type": { "type": "string", "title": "Type" }, "param": { "anyOf": [ { "type": "string" }, { "type": "null" } ], "title": "Param" }, "code": { "type": "integer", "title": "Code" } }, "additionalProperties": false, "type": "object", "required": [ "message", "type", "code" ], "title": "ErrorResponse" }, "ImageURL": { "properties": { "url": { "type": "string", "title": "Url" } }, "type": "object", "required": [ "url" ], "title": "ImageURL" }, "VideoURL": { "properties": { "url": { "type": "string", "title": "Url" } }, "type": "object", "required": [ "url" ], "title": "VideoURL" }, "NIMVLMChatCompletionContentPartVideo": { "additionalProperties": false, "properties": { "video_url": { "description": "Video url", "allOf": [ { "$ref": "#/components/schemas/VideoURL" } ] }, "type": { "type": "string", "description": "The type of the content part.", "enum": [ "video_url" ], "const": "video_url", "title": "Type" } }, "required": [ "type", "video_url" ], "type": "object", "title": "NIMVLMChatCompletionContentPartVideo" }, "NIMLLMChatCompletionMessage": { "properties": { "role": { "allOf": [ { "$ref": "#/components/schemas/Role" } ], "description": "The role of the message's author." }, "content": { "anyOf": [ { "type": "string" }, { "items": { "anyOf": [ { "$ref": "#/components/schemas/NIMVLMChatCompletionContentPartImage" }, { "$ref": "#/components/schemas/NIMVLMChatCompletionContentPartVideo" }, { "$ref": "#/components/schemas/NIMVLMChatCompletionContentPartText" } ] }, "type": "array" } ], "title": "Content", "description": "The contents of the message.\n
To pass media (only with role=`user`):\n
- Use content parts. Text is passed with `type`=`text`.\n
- Images are passed with `type`=`image_url`; set `image_url.url` to an image URL or a base64 data URI like `data:image/{format};base64,{base64encodedimage}`.\n
- Videos are passed with `type`=`video_url`; set `video_url.url` to a video URL or a base64 data URI like `data:video/{format};base64,{base64encodedvideo}`.\n
For `system` and `assistant` roles, the object list format is not supported.\n" } }, "type": "object", "required": [ "role", "content" ], "title": "NIMLLMChatCompletionMessage" }, "NIMLLMChatCompletionRequest": { "properties": { "messages": { "items": { "$ref": "#/components/schemas/NIMLLMChatCompletionMessage" }, "type": "array", "minItems": 1, "title": "Messages", "description": "A list of messages comprising the conversation so far.", "example": [ { "role": "user", "content": [ { "type": "text", "text": "What is in this image?" }, { "type": "image_url", "image_url": { "url": "https://assets.ngc.nvidia.com/products/api-catalog/phi-3-5-vision/example1b.jpg" } } ] } ] }, "model": { "type": "string", "title": "Model", "description": "The model to use.", "default": "google/diffusiongemma-26b-a4b-it" }, "chat_template_kwargs": { "type": "object", "properties": { "enable_thinking": { "type": "boolean", "default": true, "description": "Enable or disable the model's thinking/reasoning mode." } }, "title": "Chat Template Kwargs", "description": "Additional keyword arguments to pass to the chat template. Use {\"enable_thinking\": true} to enable reasoning or {\"enable_thinking\": false} to disable it.", "default": { "enable_thinking": true } }, "max_tokens": { "anyOf": [ { "type": "integer", "maximum": 32768, "minimum": 1 }, { "type": "null" } ], "title": "Max Tokens", "description": "The maximum number of tokens that can be generated.", "default": 4096 }, "seed": { "anyOf": [ { "type": "integer", "maximum": 9223372036854776000, "minimum": -9223372036854776000 }, { "type": "null" } ], "title": "Seed", "description": "Changing the seed will produce a different response with similar characteristics. Fixing the seed will reproduce the same results if all other parameters are also kept constant." }, "stream": { "anyOf": [ { "type": "boolean" }, { "type": "null" } ], "title": "Stream", "description": "If set, partial message deltas will be sent, like in ChatGPT. Tokens will be sent as data-only [server-sent events](https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events#Event_stream_format) as they become available, with the stream terminated by a `data: [DONE]`", "default": false }, "temperature": { "anyOf": [ { "type": "number", "maximum": 1, "minimum": 0 }, { "type": "null" } ], "title": "Temperature", "description": "What sampling temperature to use, between 0 and 2. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic.", "default": 1 }, "top_p": { "anyOf": [ { "type": "number", "maximum": 1, "exclusiveMinimum": 0 }, { "type": "null" } ], "title": "Top P", "description": "An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both.", "default": 0.95 } }, "additionalProperties": false, "type": "object", "required": [ "messages", "model" ], "title": "NIMLLMChatCompletionRequest" }, "NIMVLMChatCompletionContentPartImage": { "properties": { "image_url": { "$ref": "#/components/schemas/ImageURL" }, "type": { "type": "string", "enum": [ "image_url" ], "const": "image_url", "title": "Type" } }, "type": "object", "required": [ "image_url", "type" ], "title": "NIMVLMChatCompletionContentPartImage" }, "NIMVLMChatCompletionContentPartText": { "properties": { "text": { "type": "string", "title": "Text" }, "type": { "type": "string", "enum": [ "text" ], "const": "text", "title": "Type" } }, "type": "object", "required": [ "text", "type" ], "title": "NIMVLMChatCompletionContentPartText" }, "Role": { "type": "string", "enum": [ "user", "assistant" ], "title": "Role" }, "UsageInfo": { "properties": { "prompt_tokens": { "type": "integer", "title": "Prompt Tokens", "default": 0 }, "total_tokens": { "type": "integer", "title": "Total Tokens", "default": 0 }, "completion_tokens": { "anyOf": [ { "type": "integer" }, { "type": "null" } ], "title": "Completion Tokens", "default": 0 } }, "additionalProperties": false, "type": "object", "title": "UsageInfo" } } }, "x-readme": { "explorer-enabled": true, "proxy-enabled": true } } ```