Llama Nemotron Embed VL 1B v2 was developed by NVIDIA to retrieve material that helps answer a question. The query is text. Ailoot's current catalogue presents text input for the model: the question and material are encoded as text. The manufacturer also describes a document as a page image or an image together with extracted text. These visual variants require a separate method supporting images. With the appropriate input, the model represents tables, charts and infographics in a shared vector space.
This retrieval is suited to question-answering assistants over large document collections, internal reference material and research material. Encoding produces a numerical vector. The retrieved page or passage can then be supplied to a model that forms an answer: Embed VL itself performs the context-retrieval stage.
Query and document are encoded independently. A collection can therefore be processed in advance, with vectors stored alongside source pages, and only the query vector computed for a new question. Vector similarity helps select material by meaning. The text roles of queries and documents differ, which needs to be preserved when preparing data.
Each input becomes one vector of up to 2048 numbers. Text mode uses a budget of 8192 tokens. The manufacturer's visual method combining an image and text describes a shared budget of 10,240 tokens including both parts; this is not the text limit of Ailoot's method. The image is split into patches: the described configuration permits up to six patches and a downscaled copy of the whole page, totaling around 1792 visual tokens. Long text needs to be divided or shortened.
The language component is based on Llama 3.2 1B and the visual component on SigLIP2. They are combined into an encoder that averages input-token representations and returns a single vector. Contrastive training moves a question closer to the appropriate page and farther from irrelevant material. Retrieval remains probabilistic: vector proximity does not guarantee that the selected passage contains the correct answer.
", "aboutText": "### Finding the relevant page from a question\n\nLlama Nemotron Embed VL 1B v2 was developed by NVIDIA to retrieve material that helps answer a question. The query is text. Ailoot's current catalogue presents text input for the model: the question and material are encoded as text. The manufacturer also describes a document as a page image or an image together with extracted text. These visual variants require a separate method supporting images. With the appropriate input, the model represents tables, charts and infographics in a shared vector space.\n\nThis retrieval is suited to question-answering assistants over large document collections, internal reference material and research material. Encoding produces a numerical vector. The retrieved page or passage can then be supplied to a model that forms an answer: Embed VL itself performs the context-retrieval stage.\n\n### The manufacturer's document options\n\n- A text document: an individual section, passage or already recognized page text.\n- An image: a complete page with its text, tables and graphic elements.\n- An image with text: a page's visual appearance and textual content combined in one representation.\n\nQuery and document are encoded independently. A collection can therefore be processed in advance, with vectors stored alongside source pages, and only the query vector computed for a new question. Vector similarity helps select material by meaning. The text roles of queries and documents differ, which needs to be preserved when preparing data.\n\n### Representation size and input length\n\nEach input becomes one vector of up to 2048 numbers. Text mode uses a budget of 8192 tokens. The manufacturer's visual method combining an image and text describes a shared budget of 10,240 tokens including both parts; this is not the text limit of Ailoot's method. The image is split into patches: the described configuration permits up to six patches and a downscaled copy of the whole page, totaling around 1792 visual tokens. Long text needs to be divided or shortened.\n\nThe language component is based on Llama 3.2 1B and the visual component on SigLIP2. They are combined into an encoder that averages input-token representations and returns a single vector. Contrastive training moves a question closer to the appropriate page and farther from irrelevant material. Retrieval remains probabilistic: vector proximity does not guarantee that the selected passage contains the correct answer.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "dd6f005433d631a55a865cb3e4a14b67d0c9f09506790f68f0228a1a47f5d49b", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "nvidia/llama-nemotron-embed-vl-1b-v2Llama Nemotron Embed VL v2 creates vectors for semantic document search. Ailoot's current catalogue presents text input for queries and material.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Llama Nemotron Embed VL 1B v2 was developed by NVIDIA to retrieve material that helps answer a question. The query is text. Ailoot's current catalogue presents text input for the model: the question and material are encoded as text. The manufacturer also describes a document as a page image or an image together with extracted text. These visual variants require a separate method supporting images. With the appropriate input, the model represents tables, charts and infographics in a shared vector space.
This retrieval is suited to question-answering assistants over large document collections, internal reference material and research material. Encoding produces a numerical vector. The retrieved page or passage can then be supplied to a model that forms an answer: Embed VL itself performs the context-retrieval stage.
Query and document are encoded independently. A collection can therefore be processed in advance, with vectors stored alongside source pages, and only the query vector computed for a new question. Vector similarity helps select material by meaning. The text roles of queries and documents differ, which needs to be preserved when preparing data.
Each input becomes one vector of up to 2048 numbers. Text mode uses a budget of 8192 tokens. The manufacturer's visual method combining an image and text describes a shared budget of 10,240 tokens including both parts; this is not the text limit of Ailoot's method. The image is split into patches: the described configuration permits up to six patches and a downscaled copy of the whole page, totaling around 1792 visual tokens. Long text needs to be divided or shortened.
The language component is based on Llama 3.2 1B and the visual component on SigLIP2. They are combined into an encoder that averages input-token representations and returns a single vector. Contrastive training moves a question closer to the appropriate page and farther from irrelevant material. Retrieval remains probabilistic: vector proximity does not guarantee that the selected passage contains the correct answer.
POST /v1/embeddingstextembedding| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits nvidia/llama-nemotron-embed-vl-1b-v2 | Default— |
Fieldinput | Typestring|string[] | RequirementRequired | Values & limits Minimum items: 1 Conditions: Every string nonempty after trim; data URIs are rejected; accepted budget measures relevant UTF8 bytes. | Default— |
Fieldinput_type | Typestring | RequirementOptional | Values & limits query, passage | Defaultquery |
Fieldencoding_format | Typestring | RequirementOptional | Values & limits float, base64 | Defaultfloat |
Fieldtruncate | Typestring | RequirementOptional | Values & limits NONE, START, END | Default— |
Fielddimensions | Typeinteger | RequirementOptional | Values & limits Minimum: 1 | Default— |
Fielduser | Typestring | RequirementOptional | Values & limits | Default— |
text_to_embedding
| Parameter | Contract |
|---|---|
Parametermaterial_types | Contracttext |
Parameterrequired | Contractmodel, input |
Parameteroptional_controls | Contractnull |
Parametercontent_type | Contractapplication/json |
Lifecycle: exact retained published method
Method shapes and IDs are exact; supplier capability is separate from admission.
POST /v1/embeddingsAiloot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\n: "${AILOOT_TEXT_FILE:?Set your actual UTF-8 document}"\nexport AILOOT_TEXT_FILE\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model": "nvidia/llama-nemotron-embed-vl-1b-v2", "input": []}')\nf=Path(os.environ["AILOOT_TEXT_FILE"])\nif not f.is_file():raise SystemExit("Choose your actual UTF-8 source material")\ntext=f.read_text(encoding="utf-8")\nif not text.strip():raise SystemExit("Source material must contain text")\np["input"]=[text]\nPath("request.json").write_text(json.dumps(p,ensure_ascii=False))\nPY_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/embeddings \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \\\n --data-binary @request.json -D response-headers.txt -o result.json\npython3 - <<'PY_RESULT'\nimport json\nfrom pathlib import Path\nr=json.loads(Path("result.json").read_text())\nfor x in r.get("data",[]): print(json.dumps({"index":x.get("index"),"embedding":x.get("embedding")}))\nPY_RESULT\njson
Keep every complete returned response, file and content part.
Decode base64 vectors only when that format was requested; retain every vector and its original index.
No embedding job endpoint is introduced.
| Field | Type |
|---|---|
Fielddata[].index | Typeinteger |
Fielddata[].embedding | Typefinite float[]|base64 float32 vector |
Fieldmodel | Typestring |
Fieldusage | Typeobject |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionInvalid required fields, material representation or parameter value | Next stepCorrect the indicated fields before a new creation request. |
Ailoot nvidia/llama-nemotron-embed-vl-1b-v2, nemotron-embed-vl, nemotron-embed-vl-latest
Use the method paths shown above.
nvidia/llama-nemotron-embed-vl-1b-v2, nemotron-embed-vl, nemotron-embed-vl-latest
nemotron-embed-vlLlama Nemotron Embed VL v2 creates vectors for semantic document search. Ailoot's current catalogue presents text input for queries and material.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Llama Nemotron Embed VL 1B v2 was developed by NVIDIA to retrieve material that helps answer a question. The query is text. Ailoot's current catalogue presents text input for the model: the question and material are encoded as text. The manufacturer also describes a document as a page image or an image together with extracted text. These visual variants require a separate method supporting images. With the appropriate input, the model represents tables, charts and infographics in a shared vector space.
This retrieval is suited to question-answering assistants over large document collections, internal reference material and research material. Encoding produces a numerical vector. The retrieved page or passage can then be supplied to a model that forms an answer: Embed VL itself performs the context-retrieval stage.
Query and document are encoded independently. A collection can therefore be processed in advance, with vectors stored alongside source pages, and only the query vector computed for a new question. Vector similarity helps select material by meaning. The text roles of queries and documents differ, which needs to be preserved when preparing data.
Each input becomes one vector of up to 2048 numbers. Text mode uses a budget of 8192 tokens. The manufacturer's visual method combining an image and text describes a shared budget of 10,240 tokens including both parts; this is not the text limit of Ailoot's method. The image is split into patches: the described configuration permits up to six patches and a downscaled copy of the whole page, totaling around 1792 visual tokens. Long text needs to be divided or shortened.
The language component is based on Llama 3.2 1B and the visual component on SigLIP2. They are combined into an encoder that averages input-token representations and returns a single vector. Contrastive training moves a question closer to the appropriate page and farther from irrelevant material. Retrieval remains probabilistic: vector proximity does not guarantee that the selected passage contains the correct answer.
POST /v1/embeddingstextembedding| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits nemotron-embed-vl | Default— |
Fieldinput | Typestring|string[] | RequirementRequired | Values & limits Minimum items: 1 Conditions: Every string nonempty after trim; data URIs are rejected; accepted budget measures relevant UTF8 bytes. | Default— |
Fieldinput_type | Typestring | RequirementOptional | Values & limits query, passage | Defaultquery |
Fieldencoding_format | Typestring | RequirementOptional | Values & limits float, base64 | Defaultfloat |
Fieldtruncate | Typestring | RequirementOptional | Values & limits NONE, START, END | Default— |
Fielddimensions | Typeinteger | RequirementOptional | Values & limits Minimum: 1 | Default— |
Fielduser | Typestring | RequirementOptional | Values & limits | Default— |
text_to_embedding
| Parameter | Contract |
|---|---|
Parametermaterial_types | Contracttext |
Parameterrequired | Contractmodel, input |
Parameteroptional_controls | Contractnull |
Parametercontent_type | Contractapplication/json |
Lifecycle: exact retained published method
Method shapes and IDs are exact; supplier capability is separate from admission.
POST /v1/embeddingsAiloot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\n: "${AILOOT_TEXT_FILE:?Set your actual UTF-8 document}"\nexport AILOOT_TEXT_FILE\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model": "nemotron-embed-vl", "input": []}')\nf=Path(os.environ["AILOOT_TEXT_FILE"])\nif not f.is_file():raise SystemExit("Choose your actual UTF-8 source material")\ntext=f.read_text(encoding="utf-8")\nif not text.strip():raise SystemExit("Source material must contain text")\np["input"]=[text]\nPath("request.json").write_text(json.dumps(p,ensure_ascii=False))\nPY_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/embeddings \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \\\n --data-binary @request.json -D response-headers.txt -o result.json\npython3 - <<'PY_RESULT'\nimport json\nfrom pathlib import Path\nr=json.loads(Path("result.json").read_text())\nfor x in r.get("data",[]): print(json.dumps({"index":x.get("index"),"embedding":x.get("embedding")}))\nPY_RESULT\njson
Keep every complete returned response, file and content part.
Decode base64 vectors only when that format was requested; retain every vector and its original index.
No embedding job endpoint is introduced.
| Field | Type |
|---|---|
Fielddata[].index | Typeinteger |
Fielddata[].embedding | Typefinite float[]|base64 float32 vector |
Fieldmodel | Typestring |
Fieldusage | Typeobject |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionInvalid required fields, material representation or parameter value | Next stepCorrect the indicated fields before a new creation request. |
Ailoot nvidia/llama-nemotron-embed-vl-1b-v2, nemotron-embed-vl, nemotron-embed-vl-latest
Use the method paths shown above.
nvidia/llama-nemotron-embed-vl-1b-v2, nemotron-embed-vl, nemotron-embed-vl-latest
nemotron-embed-vl-latestLlama Nemotron Embed VL v2 creates vectors for semantic document search. Ailoot's current catalogue presents text input for queries and material.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Llama Nemotron Embed VL 1B v2 was developed by NVIDIA to retrieve material that helps answer a question. The query is text. Ailoot's current catalogue presents text input for the model: the question and material are encoded as text. The manufacturer also describes a document as a page image or an image together with extracted text. These visual variants require a separate method supporting images. With the appropriate input, the model represents tables, charts and infographics in a shared vector space.
This retrieval is suited to question-answering assistants over large document collections, internal reference material and research material. Encoding produces a numerical vector. The retrieved page or passage can then be supplied to a model that forms an answer: Embed VL itself performs the context-retrieval stage.
Query and document are encoded independently. A collection can therefore be processed in advance, with vectors stored alongside source pages, and only the query vector computed for a new question. Vector similarity helps select material by meaning. The text roles of queries and documents differ, which needs to be preserved when preparing data.
Each input becomes one vector of up to 2048 numbers. Text mode uses a budget of 8192 tokens. The manufacturer's visual method combining an image and text describes a shared budget of 10,240 tokens including both parts; this is not the text limit of Ailoot's method. The image is split into patches: the described configuration permits up to six patches and a downscaled copy of the whole page, totaling around 1792 visual tokens. Long text needs to be divided or shortened.
The language component is based on Llama 3.2 1B and the visual component on SigLIP2. They are combined into an encoder that averages input-token representations and returns a single vector. Contrastive training moves a question closer to the appropriate page and farther from irrelevant material. Retrieval remains probabilistic: vector proximity does not guarantee that the selected passage contains the correct answer.
POST /v1/embeddingstextembedding| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits nemotron-embed-vl-latest | Default— |
Fieldinput | Typestring|string[] | RequirementRequired | Values & limits Minimum items: 1 Conditions: Every string nonempty after trim; data URIs are rejected; accepted budget measures relevant UTF8 bytes. | Default— |
Fieldinput_type | Typestring | RequirementOptional | Values & limits query, passage | Defaultquery |
Fieldencoding_format | Typestring | RequirementOptional | Values & limits float, base64 | Defaultfloat |
Fieldtruncate | Typestring | RequirementOptional | Values & limits NONE, START, END | Default— |
Fielddimensions | Typeinteger | RequirementOptional | Values & limits Minimum: 1 | Default— |
Fielduser | Typestring | RequirementOptional | Values & limits | Default— |
text_to_embedding
| Parameter | Contract |
|---|---|
Parametermaterial_types | Contracttext |
Parameterrequired | Contractmodel, input |
Parameteroptional_controls | Contractnull |
Parametercontent_type | Contractapplication/json |
Lifecycle: exact retained published method
Method shapes and IDs are exact; supplier capability is separate from admission.
POST /v1/embeddingsAiloot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\n: "${AILOOT_TEXT_FILE:?Set your actual UTF-8 document}"\nexport AILOOT_TEXT_FILE\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model": "nemotron-embed-vl-latest", "input": []}')\nf=Path(os.environ["AILOOT_TEXT_FILE"])\nif not f.is_file():raise SystemExit("Choose your actual UTF-8 source material")\ntext=f.read_text(encoding="utf-8")\nif not text.strip():raise SystemExit("Source material must contain text")\np["input"]=[text]\nPath("request.json").write_text(json.dumps(p,ensure_ascii=False))\nPY_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/embeddings \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \\\n --data-binary @request.json -D response-headers.txt -o result.json\npython3 - <<'PY_RESULT'\nimport json\nfrom pathlib import Path\nr=json.loads(Path("result.json").read_text())\nfor x in r.get("data",[]): print(json.dumps({"index":x.get("index"),"embedding":x.get("embedding")}))\nPY_RESULT\njson
Keep every complete returned response, file and content part.
Decode base64 vectors only when that format was requested; retain every vector and its original index.
No embedding job endpoint is introduced.
| Field | Type |
|---|---|
Fielddata[].index | Typeinteger |
Fielddata[].embedding | Typefinite float[]|base64 float32 vector |
Fieldmodel | Typestring |
Fieldusage | Typeobject |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionInvalid required fields, material representation or parameter value | Next stepCorrect the indicated fields before a new creation request. |
Ailoot nvidia/llama-nemotron-embed-vl-1b-v2, nemotron-embed-vl, nemotron-embed-vl-latest
Use the method paths shown above.
nvidia/llama-nemotron-embed-vl-1b-v2, nemotron-embed-vl, nemotron-embed-vl-latest