Amazon: Titan Multimodal Embeddings G1

{"about":"### Text input into a multimodal space\n\nAmazon Titan Multimodal Embeddings G1 is an Amazon embedding model with a shared representation space for text and images. This Ailoot API listing provides a text-based workflow: submit a short description or search phrase and receive a numerical vector. It represents the text’s content in a form your application can compare with other compatible vectors.\n\n### What you can use it for\n\nThe model can form part of catalog search, recommendations and personalization. For example, the description “red travel bag” becomes a search vector that your application uses to find similar entries in a prepared index. Search results depend on both the catalog’s contents and how it is indexed: the embedding represents the query, while the search system assembles the actual selection.\n\nThis version of Titan is designed around a shared space for textual and visual data. A text vector can be useful in a system where compatible object representations have already been prepared. For conventional search across text documents, chunk length and language suitability also matter: the developer describes G1 as an English-language model with a short text input.\n\n### How to organize the workflow\n\n1. Express a single object, attribute or search intent in a short phrase.\n2. Obtain an embedding for that phrase.\n3. Compare it with the vectors in your index and select suitable entries.\n4. Pass the retrieved material to the next stage, such as a catalog interface or a separate model that writes an answer.\n\nThe index and query must use compatible data representations. Before populating a large catalog, test real search phrases for semantic matches, uncommon attributes and cases where similar words describe different objects.\n\n### Input and output in Ailoot API\n\nThe method available here accepts **text** and returns an **embedding**; this listing has an input limit of **128 tokens**. Use brief descriptions for this operation and split long documents into separate, meaningful chunks in advance. Image uploads, image generation and video processing are outside this method’s scope. “Multimodal” describes the underlying model; the workflow available through this listing creates a vector from text.","alias_of":null,"aliases":[],"author":"Amazon","canonical_model":"amazon/titan-embed-image-v1","ctx":128,"formats":["embeddings"],"id":"amazon/titan-embed-image-v1","input":["text"],"kind":"embedding","max_out":0,"methods":[{"cache_read":null,"cache_write":null,"cache_write_1h":null,"cache_write_5m":null,"conditional_meter":null,"ctx":128,"family":"openai","fast_multiplier":1,"flex_multiplier":1,"fmt":"embeddings","http":"POST","img_in":null,"img_out":null,"in":1.6,"input":["text"],"label":"Эмбеддинги","max_out":0,"max_tool_calls":null,"out":0,"output":["embedding"],"path":"/v1/embeddings","request_model_id":"amazon/titan-embed-image-v1","route_label":"Ultra","status_reads":[],"tiers":[],"tool_prices":null,"unit":"token","units":[]}],"model_aliases":[{"canonical_model_id":"amazon/titan-embed-image-v1","request_model_id":"amazon/titan-embed-image-v1","route_label":"Ultra"}],"name":"Amazon: Titan Multimodal Embeddings G1","price":{"in":1.6,"in_label":null,"out":null,"out_label":null,"sort_in":1.6,"sort_out":null,"varies":false},"public_id":"amazon/titan-embed-image-v1","short":"Creates vectors from short text for semantic search and recommendations in Titan’s multimodal embedding space."}