{"openapi":"3.0.0","info":{"description":"NVIDIA NIM (NVIDIA Inference Microservices) provides access to top open-source AI models through a unified, OpenAI-compatible API: chat completions, completions, responses, embeddings, ranking, and NIM model functions.","version":[{"name":"v1","value":"v1"}]},"servers":[{"description":"NVIDIA NIM (NVIDIA Inference Microservices) provides access to top open-source AI models through a unified, OpenAI-compatible API: chat completions, completions, responses, embeddings, ranking, and NIM model functions.","url":"https://api.apirails.com/nvidia"}],"paths":{"/v1/chat/completions":{"post":{"tags":["Chat"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/llm-apis","api_endpoint":"https://integrate.api.nvidia.com/v1/chat/completions"},"summary":"Create chat completion","description":"Creates a model response for the given chat conversation. NVIDIA NIM provides access to 100+ top open-source models through a unified, OpenAI-compatible API.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"frequency_penalty":{"type":"number","description":"Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim"},"logit_bias":{"type":"object","description":"Modify the likelihood of specified tokens appearing in the completion. Maps tokens (specified by their token ID as a string) to a bias value from -100 to 100","properties":{"token_id":{"type":"number","description":"Token ID string mapped to bias value (-100 to 100)"}}},"logprobs":{"type":"boolean","description":"Whether to return log probabilities of the output tokens or not","enum":[true,false]},"max_tokens":{"type":"number","description":"The maximum number of tokens that can be generated in the chat completion. The total length of input tokens and generated tokens is limited by the model's context length"},"messages":{"type":"array","description":"A list of messages comprising the conversation so far. Each message has a role and content. Content can be a string or an array of multimodal content objects (for vision-language models)","required":true,"items":{"type":"object","properties":{"content":{"type":["string","array"],"description":"Content of the message. Can be a string (plain text) or an array of multimodal content objects for vision inputs. Each object has type ('text' or 'image_url') and corresponding data","required":true},"role":{"type":"string","description":"Role of the message sender","required":true,"enum":["system","user","assistant","context"]}}}},"model":{"type":"string","description":"The ID of the model to use for chat completion. NVIDIA NIM supports 100+ models including Meta Llama, Mistral AI, Google Gemma, and more","required":true},"n":{"type":"number","description":"How many chat completion choices to generate for each input message. Note that you will be charged based on the number of generated tokens across all choices"},"presence_penalty":{"type":"number","description":"Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the model's likelihood to talk about new topics"},"response_format":{"type":"object","description":"An object specifying the format that the model must output. Setting type to json_object enables JSON mode; json_schema enables structured outputs with a defined schema","properties":{"type":{"type":"string","description":"Type of response format","enum":["text","json_object","json_schema"]},"json_schema":{"type":"object","description":"JSON schema definition when type is json_schema (OpenAI-compatible only)"}}},"seed":{"type":"number","description":"If specified, the system will make a best effort to sample deterministically, such that repeated requests with the same seed and parameters should return the same result"},"stop":{"type":["string","array"],"description":"Up to 4 sequences where the API will stop generating further tokens. The returned text will not contain the stop sequence. Accepts string or array of strings"},"stream":{"type":"boolean","description":"If set, partial message deltas will be sent, like in ChatGPT. Tokens will be sent as data-only server-sent events as they become available, with the stream terminated by a data: [DONE] message","enum":[true,false]},"temperature":{"type":"number","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"},"top_logprobs":{"type":"number","description":"An integer between 0 and 20 specifying the number of most likely tokens to return at each token position, each with a log probability. Requires logprobs to be set to true"},"top_p":{"type":"number","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"},"user":{"type":"string","description":"A unique identifier representing your end-user, which can help NVIDIA to monitor and detect abuse"}}},"example":{"frequency_penalty":[0,0.5],"logit_bias":[{"1234":-50},{"5678":100}],"logprobs":[true,false],"max_tokens":[128,256,1024,4096],"messages":[{"content":"You are a helpful assistant.","role":"system"},{"content":"What is the capital of France?","role":"user"},{"content":[{"text":"Describe this image:","type":"text"},{"type":"image_url","image_url":{"url":"data:image/png;base64,iVBORw0KGgo..."}}],"role":"user"}],"model":"meta/llama-3.1-70b-instruct","n":[1,3],"presence_penalty":[0,0.6],"response_format":[{"type":"text"},{"type":"json_object"},{"type":"json_schema","json_schema":{"name":"response_schema","schema":{"type":"object","properties":{"age":{"type":"number"},"name":{"type":"string"}}}}}],"seed":[42,12345],"stop":["\n\n",["\n\nHuman:","\n\nAI:"]],"stream":[true,false],"temperature":0.7,"top_logprobs":[0,5,10],"top_p":[0.9,1],"user":["user_abc123","tenant_xyz"]}}}}}},"/v1/models":{"get":{"tags":["Models"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/llm-apis","api_endpoint":"https://integrate.api.nvidia.com/v1/models"},"summary":"List models","description":"Lists the currently available NVIDIA NIM models, and provides basic information about each one such as the owner and availability. Includes 100+ models from Meta, Mistral AI, Google, NVIDIA, and more.","security":[{"bearerAuth":[]}]}},"/v1/embeddings":{"post":{"tags":["Embeddings"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/retrieval-apis","api_endpoint":"https://integrate.api.nvidia.com/v1/embeddings"},"summary":"Create embedding","description":"Creates an embedding vector representing the input text. NVIDIA NIM supports various high-performance embedding models for retrieval, classification, clustering, and semantic similarity tasks.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"dimensions":{"type":"number","description":"The number of dimensions the resulting output embeddings should have. Only supported in some models (e.g., nv-embed-v1). Value is model-dependent"},"encoding_format":{"type":"string","description":"The format to return the embeddings in. Can be either float or base64","enum":["float","base64"]},"input":{"type":"array","description":"The input text to embed. Can be a string or an array of strings (up to 2048 inputs per request depending on model)","required":true,"items":{"type":"string"}},"input_type":{"type":"string","description":"The type of input text for specialized embedding models. Helps the model understand the context of the text","enum":["query","document"]},"model":{"type":"string","description":"ID of the model to use for generating embeddings. NVIDIA NIM supports various embedding models for different use cases","required":true},"truncate":{"type":"string","description":"How to handle inputs longer than the maximum token length. Options are NONE (raises error) or END (truncates)","enum":["NONE","END"]},"user":{"type":"string","description":"A unique identifier representing your end-user, which can help NVIDIA to monitor and detect abuse"}}},"example":{"dimensions":[256,512,1024,2048,4096],"encoding_format":["float","base64"],"input":[["Your text string goes here"],["First document","Second document"]],"input_type":["query","document"],"model":"nvidia/nv-embed-v1","truncate":["NONE","END"],"user":["user_abc123"]}}}}}},"/v1/ranking":{"post":{"tags":["Reranking"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.nvidia.com/nim/nemo-retriever/text-reranking/latest/reference.html","api_endpoint":"https://integrate.api.nvidia.com/v1/ranking"},"summary":"Create ranking","description":"Reranks passages by their relevance to a query. NVIDIA NIM reranking models are designed to improve retrieval accuracy in RAG (Retrieval-Augmented Generation) applications by scoring and reordering search results.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"model":{"type":"string","description":"ID of the reranking model to use. NVIDIA NIM supports various reranking models for improving retrieval accuracy in RAG applications","required":true},"passages":{"type":"array","description":"An array of passage objects, each containing a text field. The passages will be ranked by relevance to the query (max 512 passages)","required":true,"items":{"type":"object","properties":{"text":{"type":"string","description":"Passage text to be ranked against the query","required":true}}}},"query":{"type":"object","description":"The query object containing the text field with the search query to rank passages against","required":true,"properties":{"text":{"type":"string","description":"Search query text to rank passages against","required":true}}},"truncate":{"type":"string","description":"How to handle inputs longer than the maximum token length. Options are NONE (raises error) or END (truncates)","enum":["NONE","END"]},"user":{"type":"string","description":"A unique identifier representing your end-user, which can help NVIDIA to monitor and detect abuse"}}},"example":{"model":"nvidia/llama-3.2-nemoretriever-500m-rerank-v2","passages":[[{"text":"two roads diverged in a yellow wood"},{"text":"and sorry I could not travel both"},{"text":"and be one traveler long I stood"}]],"query":[{"text":"which way should i go?"}],"truncate":["NONE","END"],"user":["user_abc123"]}}}}}},"/v1/cv/nvidia/nv-dinov2":{"post":{"tags":["Visual Models"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-nv-dinov2","api_endpoint":"https://ai.api.nvidia.com/v1/cv/nvidia/nv-dinov2"},"summary":"Run DINOv2 inference","description":"Runs DINOv2 inference on the input image to extract feature embeddings. NVIDIA DINOv2 is a vision transformer model trained using self-supervised learning to generate rich visual features for images.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"embedding_type":{"type":"string","description":"Type of embedding to extract from the image","enum":["cluster","projection"]},"image":{"type":"string","description":"The input image encoded as a Base64 string or specified using the file identifier of an uploaded file","required":true}}},"example":{"embedding_type":["cluster","projection"],"image":["data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAUA...","nvcf_asset_id_abc123"]}}}}}},"/v1/cv/nvidia/nv-grounding-dino":{"post":{"tags":["Visual Models"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-nv-grounding-dino","api_endpoint":"https://ai.api.nvidia.com/v1/cv/nvidia/nv-grounding-dino"},"summary":"Run Grounding DINO inference","description":"Runs Grounding DINO inference on the input image for open-vocabulary object detection using text prompts. Detects and localizes objects specified in natural language text.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"box_threshold":{"type":"number","description":"Confidence threshold for bounding box detection. Values between 0 and 1 where higher values return fewer but more confident detections"},"image":{"type":"string","description":"The input image encoded as a Base64 string or specified using the file identifier of an uploaded file","required":true},"prompt":{"type":"string","description":"Text prompt describing the objects to detect in the image. Separate multiple objects with periods","required":true}}},"example":{"box_threshold":[0.25,0.35,0.5],"image":["data:image/jpeg;base64,/9j/4AAQSkZJRgABAQAA...","nvcf_asset_id_def456"],"prompt":["person . car . dog .","cat . dog . bird ."]}}}}}},"/v1/cv/nvidia/ocdrnet":{"post":{"tags":["Visual Models"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-ocdrnet","api_endpoint":"https://ai.api.nvidia.com/v1/cv/nvidia/ocdrnet"},"summary":"Run Ocrdnet inference","description":"Runs OCRdnet inference on the input document image to extract text. NVIDIA OCRdnet is optimized for document text recognition and OCR tasks.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"image":{"type":"string","description":"The input document image encoded as a Base64 string or specified using the file identifier of an uploaded file","required":true}}},"example":{"image":["data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAU...","nvcf_asset_id_doc001"]}}}}}},"/v1/cv/nvidia/retail-object-detection":{"post":{"tags":["Visual Models"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-retail-object-detection","api_endpoint":"https://ai.api.nvidia.com/v1/cv/nvidia/retail-object-detection"},"summary":"Run retail object detection","description":"Runs retail object detection inference on the input image. NVIDIA retail object detection model is optimized for detecting retail products and items on shelves.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"image":{"type":"string","description":"The input image encoded as a Base64 string or specified using the file identifier of an uploaded file","required":true},"num_detections":{"type":"number","description":"Maximum number of objects to detect in the image. Must be a positive integer"}}},"example":{"image":["data:image/jpeg;base64,/9j/4AAQSkZJRgABAQAA...","nvcf_asset_id_retail001"],"num_detections":[5,10,50]}}}}}},"/v1/cv/nvidia/visual-changenet":{"post":{"tags":["Visual Models"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-visual-changenet","api_endpoint":"https://ai.api.nvidia.com/v1/cv/nvidia/visual-changenet"},"summary":"Run Visual ChangeNet inference","description":"Runs Visual ChangeNet inference to detect and visualize changes between two images. NVIDIA Visual ChangeNet is designed for change detection in satellite imagery, construction monitoring, and other comparison tasks.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"image_a":{"type":"string","description":"The first (before) image encoded as a Base64 string or specified using the file identifier of an uploaded file","required":true},"image_b":{"type":"string","description":"The second (after) image encoded as a Base64 string or specified using the file identifier of an uploaded file","required":true}}},"example":{"image_a":["data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAU...","nvcf_asset_id_before001"],"image_b":["data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAU...","nvcf_asset_id_after001"]}}}}}},"/v1/v2/nvcf/pexec/status/{requestId}":{"get":{"tags":["Visual Models"],"parameters":[{"name":"requestId","in":"path","description":"The unique request identifier returned by the inference endpoint when status 202 is returned","required":true,"schema":{"type":"string","example":""}}],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/cloud-functions/reference/statuspolling","api_endpoint":"https://ai.api.nvidia.com/v1/v2/nvcf/pexec/status/{requestId}"},"summary":"Get inference status","description":"Gets the result of an earlier function invocation request that returned a status of 202. Use this endpoint to poll for async inference results from visual models and other NVIDIA NIM endpoints.","security":[{"bearerAuth":[]}]}},"/v1/vlm/nvidia/neva-22b":{"post":{"tags":["Multimodal"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-neva-22b-infer","api_endpoint":"https://ai.api.nvidia.com/v1/vlm/nvidia/neva-22b"},"summary":"Run NeVA 22B inference","description":"Runs NeVA 22B inference for vision-language tasks. NeVA (NVIDIA Enhanced Vision Assistant) is a multimodal model that can understand and generate content from both images and text.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"max_tokens":{"type":"number","description":"The maximum number of tokens that can be generated in the chat completion. Model-dependent upper limit"},"messages":{"type":"array","description":"A list of messages comprising the conversation so far. For multimodal inputs, content can include image URLs or base64 data as content objects","required":true,"items":{"type":"object","properties":{"content":{"type":["string","array"],"description":"Content of the message. For multimodal inputs, can include image URLs or base64 data as content objects with type 'text' or 'image_url'","required":true},"role":{"type":"string","description":"Role of the message sender","required":true,"enum":["system","user","assistant","context"]}}}},"model":{"type":"string","description":"The model identifier for NeVA 22B multimodal model","required":true},"stream":{"type":"boolean","description":"If set, partial message deltas will be sent as server-sent events","enum":[true,false]},"temperature":{"type":"number","description":"What sampling temperature to use, between 0 and 2. Higher values make output more random, lower values more focused"},"top_p":{"type":"number","description":"An alternative to sampling with temperature, called nucleus sampling. Only tokens comprising top_p probability mass are considered"}}},"example":{"max_tokens":[256,512,1024],"messages":[{"content":"You are a helpful vision assistant.","role":"system"},{"content":[{"text":"Describe what you see in this image:","type":"text"},{"type":"image_url","image_url":{"url":"data:image/jpeg;base64,/9j/4AAQ..."}}],"role":"user"}],"model":"nvidia/neva-22b","stream":[false,true],"temperature":0.7,"top_p":[0.9,1]}}}}}},"/v1/vlm/nvidia/vila":{"post":{"tags":["Multimodal"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-vila-infer","api_endpoint":"https://ai.api.nvidia.com/v1/vlm/nvidia/vila"},"summary":"Run VILA inference","description":"Runs VILA inference for vision-language tasks. VILA (Vision-Language-Agnostic) is a multimodal model optimized for understanding and generating content from images and text.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"max_tokens":{"type":"number","description":"The maximum number of tokens that can be generated in the chat completion. Model-dependent upper limit"},"messages":{"type":"array","description":"A list of messages comprising the conversation so far. For multimodal inputs, content can include image URLs or base64 data as content objects","required":true,"items":{"type":"object","properties":{"content":{"type":["string","array"],"description":"Content of the message. For multimodal inputs, can include image URLs or base64 data as content objects with type 'text' or 'image_url'","required":true},"role":{"type":"string","description":"Role of the message sender","required":true,"enum":["system","user","assistant","context"]}}}},"model":{"type":"string","description":"The model identifier for VILA multimodal model","required":true},"stream":{"type":"boolean","description":"If set, partial message deltas will be sent as server-sent events","enum":[true,false]},"temperature":{"type":"number","description":"What sampling temperature to use, between 0 and 2. Higher values make output more random, lower values more focused"},"top_p":{"type":"number","description":"An alternative to sampling with temperature, called nucleus sampling. Only tokens comprising top_p probability mass are considered"}}},"example":{"max_tokens":[256,512,1024],"messages":[{"content":[{"text":"What animals are in this image?","type":"text"},{"type":"image_url","image_url":{"url":"data:image/png;base64,iVBORw0KGgo..."}}],"role":"user"}],"model":"nvidia/vila","stream":[false,true],"temperature":0.7,"top_p":[0.9,1]}}}}}},"/v1/vlm/nvidia/llama-3.1-nemotron-nano-vl-8b-v1":{"post":{"tags":["Multimodal"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/multimodal-apis","api_endpoint":"https://ai.api.nvidia.com/v1/vlm/nvidia/llama-3.1-nemotron-nano-vl-8b-v1"},"summary":"Run Llama 3.1 Nemotron Nano VL 8B inference","description":"Runs Llama 3.1 Nemotron Nano VL 8B inference for vision-language tasks. This compact multimodal model provides efficient vision-language understanding for edge deployment and resource-constrained environments.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"max_tokens":{"type":"number","description":"The maximum number of tokens that can be generated in the chat completion. Model-dependent upper limit"},"messages":{"type":"array","description":"A list of messages comprising the conversation so far. For multimodal inputs, content can include image URLs or base64 data as content objects","required":true,"items":{"type":"object","properties":{"content":{"type":["string","array"],"description":"Content of the message. For multimodal inputs, can include image URLs or base64 data as content objects with type 'text' or 'image_url'","required":true},"role":{"type":"string","description":"Role of the message sender","required":true,"enum":["system","user","assistant","context"]}}}},"model":{"type":"string","description":"The model identifier for Llama 3.1 Nemotron Nano VL 8B multimodal model","required":true},"stream":{"type":"boolean","description":"If set, partial message deltas will be sent as server-sent events","enum":[true,false]},"temperature":{"type":"number","description":"What sampling temperature to use, between 0 and 2. Higher values make output more random, lower values more focused"},"top_p":{"type":"number","description":"An alternative to sampling with temperature, called nucleus sampling. Only tokens comprising top_p probability mass are considered"}}},"example":{"max_tokens":[128,256,512],"messages":[{"content":"You are a compact vision-language assistant.","role":"system"},{"content":[{"text":"Examine this image:","type":"text"},{"type":"image_url","image_url":{"url":"data:image/jpeg;base64,/9j/4AAQ..."}}],"role":"user"}],"model":"nvidia/llama-3.1-nemotron-nano-vl-8b-v1","stream":[false,true],"temperature":0.7,"top_p":[0.9,1]}}}}}},"/v1/health/nvidia/deepvariant":{"post":{"tags":["Healthcare"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-deepvariant","api_endpoint":"https://ai.api.nvidia.com/v1/health/nvidia/deepvariant"},"summary":"Run Parabricks Universal Variant Calling","description":"Runs DeepVariant inference to identify variants in short- and long-read sequencing datasets. NVIDIA DeepVariant is a deep learning model for universal variant calling that supports Illumina, Oxford Nanopore, and Pacific Biosciences read sets.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"input":{"type":"object","description":"Input data containing reference genome tarball, BAM file, and BAM Index (BAI) file for variant calling","required":true,"properties":{"bai_file":{"type":"string","description":"Path or URL to BAM index file (.bai)"},"bam_file":{"type":"string","description":"Path or URL to BAM alignment file"},"ref_tarball":{"type":"string","description":"Path or URL to reference genome tarball (.tar.gz)"}}}}},"example":{"input":[{"bai_file":"gs://bucket/sample.bai","bam_file":"gs://bucket/sample.bam","ref_tarball":"gs://bucket/ref.tar.gz"}]}}}}}},"/v1/health/nvidia/fq2bam":{"post":{"tags":["Healthcare"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-fq2bam","api_endpoint":"https://ai.api.nvidia.com/v1/health/nvidia/fq2bam"},"summary":"Run Parabricks fq2bam sequence alignment","description":"Runs Parabricks fq2bam to generate BAM/CRAM output using BWA-MEM and GATK best practices. Performs BWA-MEM alignment, coordinate sorting, mark duplicates, and BQSR on pair-ended FASTQ files.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"input":{"type":"object","description":"Input data containing reference genome tarball and FASTQ files for sequence alignment","required":true,"properties":{"fastq_files":{"type":"array","description":"List of FASTQ file paths or URLs for paired-end reads"},"ref_tarball":{"type":"string","description":"Path or URL to reference genome tarball (.tar.gz)"}}}}},"example":{"input":[{"fastq_files":["gs://bucket/sample_R1.fastq.gz","gs://bucket/sample_R2.fastq.gz"],"ref_tarball":"gs://bucket/ref.tar.gz"}]}}}}}},"/v1/health/nvidia/genmol":{"post":{"tags":["Healthcare"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-genmol","api_endpoint":"https://ai.api.nvidia.com/v1/health/nvidia/genmol"},"summary":"Generate molecules with GenMol","description":"Runs GenMol inference for fragment-based molecule generation. GenMol is a masked diffusion model trained on SAFE representations for De Novo generation, linker design, motif extension, scaffold decoration, hit generation, and lead optimization.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"input":{"type":"object","description":"Input data containing molecular sequences in SAFE format for fragment-based molecule generation","required":true,"properties":{"sequences":{"type":"array","description":"Molecular sequences in SAFE format used as fragments for generation"}}},"num_molecules":{"type":"number","description":"Number of molecules to generate"},"sample_num":{"type":"number","description":"Number of samples to generate per molecule"},"temperature":{"type":"number","description":"SoftMax temperature scaling factor for generation randomness. Higher values increase diversity, lower values increase determinism"}}},"example":{"input":[{"sequences":["*C.*C.*C.*C(=O)*","*CC(=O)N*"]}],"num_molecules":[5,10,20],"sample_num":[1,5],"temperature":[0.8,1,1.5]}}}}}},"/v1/health/nvidia/maisi":{"post":{"tags":["Healthcare"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-maisi","api_endpoint":"https://ai.api.nvidia.com/v1/health/nvidia/maisi"},"summary":"Generate synthetic CT images with MAISI","description":"Generates high-quality synthetic 3D CT images with optional anatomical annotations using NVIDIA MAISI (Medical AI for Synthetic Imaging). Supports variable voxel sizes, up to 127 anatomical classes, and controllable anatomy sizes for research purposes.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"anatomy_list":{"type":"array","description":"Optional list of up to 127 anatomical class names or indices to annotate","items":{"type":"string"}},"body_region":{"type":"array","description":"Target body regions for the generated CT image","required":true,"enum":["head","chest","thorax","abdomen","pelvis","lower"],"items":{"type":"string"}},"controllable_anatomy_size":{"type":"array","description":"List of tuples (organ_name, size_value) for organ size control. Max 10 anatomies. Size: 0.0-1.0 or -1 to remove. Organs: liver, gallbladder, stomach, pancreas, colon, lung tumor, bone lesion, hepatic tumor, colon cancer primaries, pancreatic tumor","items":{"type":"object","properties":{"organ_name":{"type":"string","description":"Name of the anatomical organ","enum":["liver","gallbladder","stomach","pancreas","colon","lung tumor","bone lesion","hepatic tumor","colon cancer primaries","pancreatic tumor"]},"size_value":{"type":"number","description":"Size value: 0.0 to 1.0 for control, or -1 to remove the anatomy"}}}},"num_output_samples":{"type":"number","description":"Number of synthetic CT images to generate"},"output_size":{"type":"array","description":"Image dimensions as a list of 3 integers (x, y, z). x,y: 256-512, z: 128-768","items":{"type":"number"}},"spacing":{"type":"array","description":"Voxel spacing as a list of 3 floats in mm. Each value: 0.5 to 5.0","items":{"type":"number"}}}},"example":{"anatomy_list":[["liver","lung","kidney","heart"]],"body_region":[["head","chest","abdomen","pelvis"]],"controllable_anatomy_size":[[{"organ_name":"liver","size_value":0.8},{"organ_name":"hepatic_tumor","size_value":0.5}]],"num_output_samples":[1,4],"output_size":[[256,256,384],[512,512,768]],"spacing":[[1,1,1],[1.5,1.5,2]]}}}}}},"/v1/health/nvidia/molmim":{"post":{"tags":["Healthcare"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-molmim","api_endpoint":"https://ai.api.nvidia.com/v1/health/nvidia/molmim"},"summary":"Generate molecules with MolMIM","description":"Runs MolMIM inference for molecule generation and optimization. MolMIM generates new molecules in SMILES format by sampling from the latent space around a seed molecule, with optional optimization using the CMA-ES algorithm.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"input":{"type":"string","description":"Seed molecule in SMILES format for sampling and optimization","required":true},"num_samples":{"type":"number","description":"Number of molecules to generate from the latent space around the seed molecule"},"optimize":{"type":"boolean","description":"Whether to perform optimization with CMA-ES algorithm. When true, uses property_name scoring function to guide molecule generation toward desired properties","enum":[true,false]},"property_name":{"type":"string","description":"Scoring function name for optimization. QED (Quantitative Estimate of Drug-likeness) or LogP (partition coefficient). Only used when optimize is true","enum":["QED","LogP"]}}},"example":{"input":["CC(C)Cc1ccc(cc1)C(C)C","c1ccccc1","CCO"],"num_samples":[5,10,25],"optimize":[true,false],"property_name":["QED","LogP"]}}}}}},"/v1/health/nvidia/vista3d":{"post":{"tags":["Healthcare"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-vista3d","api_endpoint":"https://ai.api.nvidia.com/v1/health/nvidia/vista3d"},"summary":"Run VISTA-3D medical imaging segmentation","description":"Runs VISTA-3D inference for 3D medical imaging segmentation. VISTA-3D is a specialized interactive foundation model that provides accurate and adaptable segmentation analysis across anatomies and modalities using a multi-head architecture.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"class_info":{"type":"array","description":"Array of class/point information for targeted segmentation. Contains label names or indices for class prompts. Supports up to 127 anatomical classes","items":{"type":"string"}},"image":{"type":"string","description":"Input CT image in NIfTI format encoded as Base64 string or file identifier","required":true},"point_prompt":{"type":"array","description":"Click-based point prompts for interactive segmentation. Maps label names to coordinate tuples (x, y, z)","items":{"type":"object","properties":{"label":{"type":"string","description":"Label name or index for the anatomy"},"points":{"type":"array","description":"List of coordinate tuples (x, y, z) for user click points"}}}}}},"example":{"class_info":[["liver","kidney","spleen","pancreas"]],"image":["data:application/octet-stream;base64,eJzD...==","nvcf_asset_id_ctscan001"],"point_prompt":[[{"label":"liver","points":[[128,200,45]]},{"label":"tumor","points":[[140,210,50],[135,205,48]]}]]}}}}}},"/v1/climate/nvidia/corrdiff":{"post":{"tags":["Climate Simulation"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-corrdiff","api_endpoint":"https://ai.api.nvidia.com/v1/climate/nvidia/corrdiff"},"summary":"Run CorrDiff weather downscaling","description":"Runs CorrDiff inference for weather downscaling. Corrector Diffusion (CorrDiff) US GEFS-HRRR model down-scales several surface and atmospheric variables from 25-km resolution forecast data to 3-km resolution using Patch-Based Corrector Diffusion architecture.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"input":{"type":"object","description":"Input data containing 25-km resolution forecast data with 38 surface and atmospheric variables","required":true,"properties":{"forecast_data":{"type":"string","description":"URL or path to the forecast input data file"},"timestamp":{"type":"string","description":"Forecast initialization timestamp"},"variables":{"type":"array","description":"List of variable names included in the forecast data"}}},"lead_time":{"type":"string","description":"Forecast lead time for downscaling. Specifies how many hours ahead to downscale (typically F03, F06, ..., F24)","required":true}}},"example":{"input":[{"forecast_data":"gs://bucket/gefs-hrrr-forecast.nc","timestamp":"2024-06-15T12:00:00Z","variables":["t2m","r2","u10","v10"]}],"lead_time":["F06","F12","F18","F24"]}}}}}},"/v1/climate/nvidia/fourcastnet":{"post":{"tags":["Climate Simulation"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-fourcastnet","api_endpoint":"https://ai.api.nvidia.com/v1/climate/nvidia/fourcastnet"},"summary":"Run FourCastNet global weather forecasting","description":"Runs FourCastNet V2 inference for global atmospheric forecasting. FourCastNet V2 uses Spherical Fourier Neural Operator (SFNO) to predict a collection of surface and atmospheric variables with 6-hour time-delta forecasting capability.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"datetime":{"type":"string","description":"DateTime for the forecast initialization in ISO 8601 format","required":true},"input":{"type":"object","description":"Input data containing 73 surface and atmospheric variables for initial weather state","required":true,"properties":{"initial_state":{"type":"string","description":"URL or path to initial condition data file with 73 variables"},"resolution":{"type":"number","description":"Grid resolution in degrees (typically 0.25)"}}}}},"example":{"datetime":["2024-06-15T00:00:00Z","2024-06-15T12:00:00Z"],"input":[{"initial_state":"gs://bucket/fourcastnet-initial-state.nc","resolution":0.25}]}}}}}},"/v1/route/nvidia/cuopt":{"post":{"tags":["Route Optimization"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.api.nvidia.com/nim/reference/nvidia-cuopt","api_endpoint":"https://ai.api.nvidia.com/v1/route/nvidia/cuopt"},"summary":"Submit cuOpt routing problem","description":"Submits a routing optimization problem to NVIDIA cuOpt solver. cuOpt is an AI microservice that optimizes logistics routing to reduce costs, increase revenue, and lower carbon emissions. Supports Capacitated Vehicle Routing Problems (CVRP), CVRP with Time Windows (CVRTW), and Pickup and Delivery with Time Windows (PDPTW). Use cases include Last Mile Delivery, Dispatch Optimization, and Pickup and Delivery scenarios.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"input":{"type":"object","description":"Input data containing routing problem configuration in JSON format. Includes vehicle locations, task locations, fleet size, time windows, capacity constraints, and other problem-specific parameters","required":true,"properties":{"capacity_constraints":{"type":"array","description":"Maximum capacity per vehicle (weight, volume, etc.)"},"fleet_size":{"type":"number","description":"Number of vehicles available for routing"},"task_locations":{"type":"array","description":"Task/delivery location coordinates"},"time_windows":{"type":"array","description":"Time window constraints for each task location"},"vehicle_locations":{"type":"array","description":"Depot/vehicle start locations as coordinates"}}},"input_file":{"type":"string","description":"Optional file identifier for large input data files. Use NVCF large assets API to upload files and obtain the identifier"}}},"example":{"input":[{"capacity_constraints":[100],"fleet_size":3,"task_locations":[[1,2],[3,4]],"vehicle_locations":[[0,0]]}],"input_file":["nvcf_asset_id_cuopt_input_001"]}}}}}},"/v1/completions":{"post":{"tags":["Completions"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.nvidia.com/nim/large-language-models/latest/reference/api-reference.html","api_endpoint":"https://integrate.api.nvidia.com/v1/completions"},"summary":"Create completion","description":"Creates a completion for the provided prompt and parameters (OpenAI-compatible, single-turn). NVIDIA NIM serves this through the vLLM inference backend.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"frequency_penalty":{"type":"number","description":"Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model likelihood of repeating the same line verbatim"},"logit_bias":{"type":"object","description":"Modify the likelihood of specified tokens appearing in the completion. Accepts a JSON object that maps tokens (specified by their token ID as a string) to a bias value from -100 to 100","properties":{}},"logprobs":{"type":"boolean","description":"Whether to return log probabilities of the output tokens or not","enum":[true,false]},"max_tokens":{"type":"number","description":"The maximum number of tokens that can be generated in the completion. The total length of input tokens and generated tokens is limited by the model context length"},"model":{"type":"string","description":"ID of the model to use for this request","required":true},"n":{"type":"number","description":"How many completions to generate for each prompt"},"presence_penalty":{"type":"number","description":"Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the model likelihood to talk about new topics"},"prompt":{"type":"string","description":"The prompt(s) to generate completions for. May be a string, array of strings, or array of token arrays","required":true},"seed":{"type":"number","description":"If specified, the system will make a best effort to sample deterministically"},"stop":{"type":"string","description":"Up to 4 sequences where the API will stop generating further tokens. The returned text will not contain the stop sequence"},"stream":{"type":"boolean","description":"Whether to stream back partial progress via server-sent events","enum":[true,false]},"temperature":{"type":"number","description":"What sampling temperature to use, between 0 and 2. Higher values like 0.8 make the output more random, while lower values like 0.2 make it more focused and deterministic"},"top_p":{"type":"number","description":"An alternative to sampling with temperature, called nucleus sampling; the model considers the results of the tokens with top_p probability mass"}}},"example":{"frequency_penalty":"","logit_bias":{},"logprobs":"","max_tokens":[64,256,1024],"model":"meta/llama-3.1-8b-instruct","n":"","presence_penalty":"","prompt":"Once upon a time","seed":"","stop":"","stream":"","temperature":"","top_p":""}}}}}},"/v1/responses":{"post":{"tags":["Responses"],"parameters":[],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.nvidia.com/nim/large-language-models/latest/reference/api-reference.html","api_endpoint":"https://integrate.api.nvidia.com/v1/responses"},"summary":"Create model response","description":"Creates a model response using the OpenAI Responses API. Provide text, image, or file inputs to generate a response from a NVIDIA NIM-hosted model.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{"input":{"type":"string","description":"Text, image, or file inputs to the model, used to generate a response","required":true},"instructions":{"type":"string","description":"System (or developer) instructions inserted at the beginning of the model context"},"max_output_tokens":{"type":"number","description":"An upper bound for the number of tokens that can be generated for a response"},"metadata":{"type":"object","description":"Set of key-value pairs that can be attached to the response","properties":{}},"model":{"type":"string","description":"ID of the model to use for this request","required":true},"stream":{"type":"boolean","description":"Whether to stream back partial progress via server-sent events","enum":[true,false]},"temperature":{"type":"number","description":"What sampling temperature to use, between 0 and 2. Higher values make the output more random, lower values make it more focused and deterministic"},"top_p":{"type":"number","description":"An alternative to sampling with temperature, called nucleus sampling; the model considers the results of the tokens with top_p probability mass"}}},"example":{"input":"Explain the theory of relativity in one sentence.","instructions":"","max_output_tokens":"","metadata":{},"model":"meta/llama-3.1-8b-instruct","stream":"","temperature":"","top_p":""}}}}}},"/v1/responses/{response_id}":{"get":{"tags":["Responses"],"parameters":[{"name":"response_id","in":"path","description":"The ID of the response to retrieve","required":true,"schema":{"type":"string","example":""}}],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.nvidia.com/nim/large-language-models/latest/reference/api-reference.html","api_endpoint":"https://integrate.api.nvidia.com/v1/responses/{response_id}"},"summary":"Retrieve model response","description":"Retrieves a previously created model response with the given ID.","security":[{"bearerAuth":[]}]}},"/v1/responses/{response_id}/cancel":{"post":{"tags":["Responses"],"parameters":[{"name":"response_id","in":"path","description":"The ID of the response to cancel","required":true,"schema":{"type":"string","example":""}}],"responses":{"401":{"description":"Bearer token required (Authorization header)","content":{"application/json":{}}}},"externalDocs":{"api_ref":"https://docs.nvidia.com/nim/large-language-models/latest/reference/api-reference.html","api_endpoint":"https://integrate.api.nvidia.com/v1/responses/{response_id}/cancel"},"summary":"Cancel model response","description":"Cancels a streaming model response with the given ID.","security":[{"bearerAuth":[]}],"requestBody":{"description":"","required":true,"content":{"application/json":{"schema":{"type":"object","required":[],"properties":{}},"example":{}}}}}}},"components":{"securitySchemes":{"bearerAuth":{"type":"http","scheme":"bearer"}}},"security":[{"bearerAuth":[]}]}