Nomic Embed v2
Free.ai (self-hosted)
·
embeddings
·
~100 Ii-token call
Nomic Embed v2 yi an Imodeli eyenziwe ngu Nomic AI. Enamandla kakhulu kwi Retrieval augmented generation with flexible vector sizes.. I-self-hosted kwi-Free.ai GPUs — iqhuba ngokukhululekileyo kwi-token pool yakho yosuku (100 tokens 100% yexabiso). Ikhutshwe phantsi kwe Apache 2.0 — ukusetyenziswa korhwebo kuvunyelwe kwi Free.ai.
Sebenzisa nge-API
I-REST API ehambelana ne-OpenAI. Yenza iqhosha kwaye unxulumane nale modeli kwimizuzu.
curl -X POST https://api.free.ai/v1/image/generate/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"nomic-embed-v2","prompt":"your prompt here"}'Imibuzo ebuzwa rhoqo
Nomic Embed v2 iguqula umbhalo ube yi-vector eqinileyo (uluhlu lwee-floats) ethatha uphawu. Sebenzisa ukuphendla kwe-semantic, ukudwelisa, ukucebisa, ukudala okuphuculweyo kokufumana (RAG), nakuphi na umsebenzi apho "ukuba lo mbhalo ufana nalo mbhalo" ubalulekile.
Ii-dimensi eziqhelekileyo ziyi-384, 768, 1024, okanye 1536 kuxhomekeke kwimodeli. BGE-M3 ikhupha i-1024-dim; OpenAI Ada ikhupha i-1536. Impendulo ye-API iquka ubungakanani ukuze i-vector DB yakho ikhethe isalathisi esifanelekileyo.
Iimodeli zokudibanisa zexesha elidlulileyo (eziquka ezininzi iinketho kwi Free.ai) ziqeqeshwe kwiilwimi ezingaphezu kwe 100. Ukufumana kwakhona kweelwimi eziphesheya kusebenza — khangela ngesiNgesi, dibanisa amaxwebhu ngesiSpanish.
512 ukuya kwi 8, 192 iimpawu ngokuxhomekeke kwimodeli. Iindawo ezingenisweyo ezide zicuthiweyo - inxalenye yamaxwebhu ade kumacandelo phambi kokufaka.
Nomic Embed v2 iqhuba kwiGPU zethu zethu kwaye iphakathi kwezixhobo ezibiza kakhulu - malunga ne-100 ye-token nganye ebizwayo evela kwi-pool yakho yasimahla yosuku ngalunye. $5 = 200K ye-token.
Ewe — thumela uluhlu lwamagama ku /v1/embeddings/ kwaye Nomic Embed v2 ibuyisela uluhlu lwee-vectors ngokulandelana okufanayo. Ubungakanani beqela ukuya kuthi ga kwi-2,048 ngesicelo ngasinye.
L2-iqhelekileyo ngokumiselweyo — uthelekiso lwe-cosin = umkhiqizo we-dot. Phatha `i-normalize=false` ukuba ufuna i-vector emnyama ye-metric yokuhamba ehlukileyo.
Nayiphi na — Pinecone, Weaviate, Qdrant, Chroma, pgvector, FAISS, LanceDB. Nomic Embed v2 ibuyisela i-JSON eqhelekileyo ejikelezayo; i-DB ayizukubona imodeli.
Ewe — UTHENGA ku /v1/embeddings/ ngemodeli="Nomic Embed v2". Uhlobo lwempendulo oluhambelana ne-OpenAI, ngoko ke iilayibrari zomxhasi ezikhoyo zisebenza zitshintshile. /api/ inesiqendu esipheleleyo.
Iimodeli ezimkelweyo zigcina umbhalo wakho kwi-GPU yethu kwaye ziyisuse emva kokuba unxulumano lubuyiselwe. I-Premium idlula nge-DPA. Asiqeqeshi kwingeniso yakho.
I-Sub-100ms yombhalo omfutshane kwi-self-hosted, 100-500ms kwi-premium. I-batch calls scale roughly linearly — 1,000 chunks complete in 2-10 seconds.
Ewe - Free.ai inikezela ngenkonzo yorhwebo yokufaka. Yenza ukukhangela kokwenziwa, i-RAG imibhobho, inkqubo yokucebisa ngaphandle kwe-royalty nganye ye-vector.