Nomic Embed v2
Free.ai (self-hosted)
·
embeddings
·
~100 Token call
Nomic Embed v2 bụ an embedding model e mepụtara site na Nomic AI. Ọrụ na Retrieval augmented generation with flexible vector sizes.. Self-hosted na Free.ai GPUs - na-agba ọsọ n'efu megide ụbọchị gị token pool (100 tokens n'ime oku). E wepụtara ya n'okpuru Apache 2.0 - iji ahịa ekwenyela na Free.ai.
Ajụjụ ndị a jụrụkarị
Nomic Embed v2 na-atụgharị ngwe n'ime vektor dị n'ime (ụdị ndesịta nke floats) nke na-echekwa ihe n'ime ya. Jiri ya maka ọchụchọ semantik, n'ịrụnye, n'enyemaka, n'ịrụnye-amụbaworị mmegharị (RAG), nakwa ọrụ ọbụla ebe "ihe a dị ka ngwe a n'ihe ahụ" dị mkpa.
Tipu Dimensions bụ 384, 768, 1024, ma ọ bụ 1536 dabere na model. BGE-M3 emits 1024-dim; OpenAI Ada emits 1536. The API response includes the dimension so your vector DB picks the right index.
Modern embedding models (na-agụnye ọtụtụ nhọrọ na Free.ai) bụ ndị na-enyocha na 100 + asụsụ. Cross-asụsụ na-eweghachi ọrụ - n'ịhụnanya na English, match faịlụ na Spanish.
512 ruo 8,192 tokens n'ihe nlereanya. Nhazi nke ogologo bụ nke a na-ewepụ - wepụ ogologo faịlụ ndị ahụ n'ime paragraf tupú ịnyepụta.
Nomic Embed v2 na-agba ọsọ na GPUs anyị na-enye anyị na-eme ka ọ bụrụ ihe dị ọnụ ala - banyere ~ 100 tokens kwa oku na-abịa site na ụbọchị gị n'efu pool. $5 = 200K tokens.
Ya - POST ndesịta strings na /v1/embeddings/ na Nomic Embed v2 na-eziga ndesịta nke vektor na usoro iheomume ahụ. Báà nha ruo 2,048 n'ihi nkwenye.
L2-normalized site na difọ́ọ̀ltụ̀ - cosine similitudes = dot product. Pas̃ `normalize=false` ma ọ bụrụ na ịchọrọ vektor n'ime n'ime maka n'ebe dị iche iche.
Ọbụla - Pinecone, Weaviate, Qdrant, Chroma, pgvector, FAISS, LanceDB. Nomic Embed v2 na-ezigaghachi JSON floats; DB anaghị ahụ model.
Ya - POST ka /v1/embeddings/ na model="Nomic Embed v2". OpenAI-compatible response shape, ka ndịna-enyemaka laibịretị nọ ugbua na-arụ ọrụ na-enweghị mgbanwe. /api/ nwere nlekọta zuru ezu.
Self-hosted models na-echekwa ngwe gị na GPUs anyị nakwa wepụ ya mgbe ọbịa ahụ na-abịa. Premium na-aga site na DPA. Anyị anaghị arụ ọrụ na init gị.
Sub-100ms maka ngwe n'ime onwe-n'ụlọ, 100-500ms na premium. Batị ajụjụ na-ebuli elu n'ụzọ linear - 1,000 chunks zuru ezu na 2-10 sekọnd.
Ee - Free.ai na-enye ọrụ n'ime ụlọ nke embeddings. Bipụta nchọgharị mmepe, RAG pipelines, usoroiheomume n'ihe nlereanya na-enweghị per-vector royalty.