BGE-M3
Free.ai (self-hosted)
·
embeddings
·
~100 token kasta call
BGE-M3 waa an embedding model dhisay BAAI. Ugu xoog badan ee Text embeddings, semantic search. Is-hoosaysiinta Free.ai GPUs - bilaash ayaa ka socda ishaada maalinlaha ah (100 tokens wicitaan kasta). Soo baxay hoos MIT — isticmaalka ganacsi ee la oggol yahay on Free.ai.
isticmaalka API
OpenAI-ku habboon REST API. abuuro fure iyo wicitaan noocan ah daqiiqado.
curl -X POST https://api.free.ai/v1/image/generate/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"bge-m3","prompt":"your prompt here"}'Su'aalaha badanaa la isweydiiyo
BGE-M3 beddela qoraalka in vector cufan (liiska flots) oo ka dhigan tahay. U isticmaal si loo raadiyo semantic, clustering, talooyin, soo kabashada-ku kordhay dhalasho (RAG), iyo wax kasta oo shaqada halkaas oo "waa qoraalkan la mid ah in qoraalka" arrimaha.
Miisaanka caadiga ah waa 384, 768, 1024, ama 1536 ku xiran tahay qaabka. BGE-M3 soo saartaa 1024-dim; OpenAI Ada soo saartaa 1536. jawaabta API waxaa ka mid ah miisaanka si aad vector DB doorato liiska midig.
Modern embedding models (oo ay ku jiraan doorashooyinka ugu badan ee Free.ai) waxaa tababaray 100 + luqadood. Cross-luqadeed soo kabashada shaqooyinka — raadinta Ingiriisi, la mid ah dukumiintiga in Spanish.
512 ilaa 8,192 calaamadaha ku xiran tahay qaabka. Inputs dheeri ah waa la gooyo — chunk dheer dukumiintiyo in qodobbada ka hor embedding.
BGE-M3 ku socda GPUs our gaar ah oo ka mid ah qalabka ugu jaban — ku saabsan ~ 100 calaamadaha per wicitaan ka soo baxaya aad maalin kasta free pool. $ 5 = 200K calaamadaha.
Haa — POST liiska strings in / v1 / embeddings / iyo BGE-M3 soo celin doonaa liiska vector in order isku mid ah. Batch size ilaa 2,048 weydiinta kasta.
L2-normalized by default — cosine isku mid ah = dot alaab. Pass `normalize = false` haddii aad rabto vector raw u ah metric fog oo kala duwan.
Wax kasta - Pinecone, Weaviate, Qdrant, Chroma, pgvector, FAISS, LanceDB. BGE-M3 ku soo laabtaan caadiga ah JSON daboolka; DB marnaba ma arko qaabka.
Haa — POST in /v1/embeddings/ la model="BGE-M3". OpenAI-ku habboon qaabka jawaabta, sidaas darteed buugaagta macaamiisha hadda jira shaqada aan la beddelin. /api/ waxaa ku qoran soo jeedinta buuxda.
Self-hosted qaabab ku hayn qoraalka ku GPUs our iyo ka saaro ka dib markii la soo celiyo. Premium ka gudbaan DPA. Aan tababarka ku saabsan aad inputs.
Sub-100ms qoraal gaaban oo ku saabsan is-hoosaysiinta, 100-500ms on premium. Codsiyada tirada badan waxay si toos ah u kala saaraan - 1,000 chunks oo ku dhammaada 2-10 ilbiriqsi.
Haa — Free.ai deeqaha isticmaalka ganacsi ee embeddings. dhiso raadinta wax soo saarka, RAG pipelines, nidaamka talooyinka aan per-vector royalty.