{"id":64546,"date":"2024-05-31T17:51:00","date_gmt":"2024-05-31T14:21:00","guid":{"rendered":"https:\/\/nabfollower.com\/blog\/rag-%d8%a8%d8%a7-llama-cpp-%d9%88-%d8%ae%d8%af%d9%85%d8%a7%d8%aa-api-%d8%ae%d8%a7%d8%b1%d8%ac%db%8c\/"},"modified":"2024-05-31T17:51:00","modified_gmt":"2024-05-31T14:21:00","slug":"rag-%d8%a8%d8%a7-llama-cpp-%d9%88-%d8%ae%d8%af%d9%85%d8%a7%d8%aa-api-%d8%ae%d8%a7%d8%b1%d8%ac%db%8c","status":"publish","type":"post","link":"https:\/\/nabfollower.com\/blog\/rag-%d8%a8%d8%a7-llama-cpp-%d9%88-%d8%ae%d8%af%d9%85%d8%a7%d8%aa-api-%d8%ae%d8%a7%d8%b1%d8%ac%db%8c\/","title":{"rendered":"RAG \u0628\u0627 llama.cpp \u0648 \u062e\u062f\u0645\u0627\u062a API \u062e\u0627\u0631\u062c\u06cc"},"content":{"rendered":"

<\/p>\n

\n

txtai \u06cc\u06a9 \u067e\u0627\u06cc\u06af\u0627\u0647 \u062f\u0627\u062f\u0647 \u062a\u0639\u0628\u06cc\u0647\u200c\u0634\u062f\u0647 \u06cc\u06a9\u067e\u0627\u0631\u0686\u0647 \u0628\u0631\u0627\u06cc \u062c\u0633\u062a\u062c\u0648\u06cc \u0645\u0639\u0646\u0627\u06cc\u06cc\u060c \u0627\u0631\u06a9\u0633\u062a\u0631\u0627\u0633\u06cc\u0648\u0646 LLM \u0648 \u06af\u0631\u062f\u0634\u200c\u0647\u0627\u06cc \u06a9\u0627\u0631\u06cc \u0645\u062f\u0644 \u0632\u0628\u0627\u0646 \u0627\u0633\u062a.<\/p>\n

txtai \u0647\u0645\u06cc\u0634\u0647 \u06cc\u06a9 \u0686\u0627\u0631\u0686\u0648\u0628 \u0645\u062d\u0644\u06cc \u0628\u0648\u062f\u0647 \u0648 \u062e\u0648\u0627\u0647\u062f \u0628\u0648\u062f. \u062f\u0631 \u0627\u0635\u0644 \u0628\u0631\u0627\u06cc \u0627\u062c\u0631\u0627\u06cc \u0645\u062f\u0644\u200c\u0647\u0627 \u0628\u0631 \u0631\u0648\u06cc \u0633\u062e\u062a\u200c\u0627\u0641\u0632\u0627\u0631 \u0645\u062d\u0644\u06cc \u0628\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 \u062a\u0631\u0627\u0646\u0633\u0641\u0648\u0631\u0645\u0627\u062a\u0648\u0631 Hugging Face \u0637\u0631\u0627\u062d\u06cc \u0634\u062f\u0647 \u0628\u0648\u062f. \u0647\u0645\u0627\u0646\u0637\u0648\u0631 \u06a9\u0647 \u0641\u0636\u0627\u06cc \u0647\u0648\u0634 \u0645\u0635\u0646\u0648\u0639\u06cc \u0637\u06cc \u0633\u0627\u0644 \u06af\u0630\u0634\u062a\u0647 \u062a\u06a9\u0627\u0645\u0644 \u06cc\u0627\u0641\u062a\u0647 \u0627\u0633\u062a\u060c txtai \u0646\u06cc\u0632 \u062a\u06a9\u0627\u0645\u0644 \u06cc\u0627\u0641\u062a\u0647 \u0627\u0633\u062a. \u0686\u0627\u0631\u0686\u0648\u0628\u200c\u0647\u0627\u06cc \u0627\u0633\u062a\u0646\u062a\u0627\u062c LLM \u0627\u0636\u0627\u0641\u06cc \u0628\u0631\u0627\u06cc \u0645\u062f\u062a\u06cc \u0628\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 llama.cpp \u0648 \u062e\u062f\u0645\u0627\u062a API \u062e\u0627\u0631\u062c\u06cc (\u0627\u0632 \u0637\u0631\u06cc\u0642 LiteLLM) \u062f\u0631 \u062f\u0633\u062a\u0631\u0633 \u0628\u0648\u062f\u0647\u200c\u0627\u0646\u062f. \u062a\u063a\u06cc\u06cc\u0631\u0627\u062a \u0627\u062e\u06cc\u0631 \u062a\u0648\u0627\u0646\u0627\u06cc\u06cc \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 \u0627\u06cc\u0646 \u0686\u0627\u0631\u0686\u0648\u0628 \u0647\u0627 \u0631\u0627 \u0628\u0631\u0627\u06cc \u0628\u0631\u062f\u0627\u0631\u06cc \u0627\u0636\u0627\u0641\u0647 \u06a9\u0631\u062f\u0647 \u0648 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 \u0622\u0646 \u0631\u0627 \u0628\u0631\u0627\u06cc \u0627\u0633\u062a\u0646\u062a\u0627\u062c LLM \u0622\u0633\u0627\u0646\u062a\u0631 \u06a9\u0631\u062f\u0647 \u0627\u0633\u062a.<\/p>\n

\u0627\u06cc\u0646 \u0645\u0642\u0627\u0644\u0647 \u0646\u062d\u0648\u0647 \u0627\u062c\u0631\u0627\u06cc \u0641\u0631\u0622\u06cc\u0646\u062f\u0647\u0627\u06cc \u0628\u0627\u0632\u06cc\u0627\u0628\u06cc \u0646\u0633\u0644 \u0627\u0641\u0632\u0648\u062f\u0647 (RAG) (\u0628\u0631\u062f\u0627\u0631\u0633\u0627\u0632\u06cc \u0648 \u0627\u0633\u062a\u0646\u062a\u0627\u062c LLM) \u0628\u0627 llama.cpp \u0648 \u0633\u0631\u0648\u06cc\u0633\u200c\u0647\u0627\u06cc API \u062e\u0627\u0631\u062c\u06cc \u0631\u0627 \u0646\u0634\u0627\u0646 \u0645\u06cc\u200c\u062f\u0647\u062f.<\/p>\n

\u0646\u0635\u0628 txtai<\/code> \u0648 \u0647\u0645\u0647 \u0648\u0627\u0628\u0633\u062a\u06af\u06cc \u0647\u0627<\/p>\n

\n
# Install txtai and dependencies\npip install llama-cpp-python[server] --extra-index-url https:\/\/abetlen.github.io\/llama-cpp-python\/whl\/cu121\npip install txtai[pipeline-llm]\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u0645\u062b\u0627\u0644 \u0627\u0648\u0644 \u06cc\u06a9 \u067e\u0627\u06cc\u06af\u0627\u0647 \u062f\u0627\u062f\u0647 Embeddings \u0628\u0627 \u067e\u0634\u062a\u0648\u0627\u0646\u0647 \u0628\u0631\u062f\u0627\u0631\u06cc llama.cpp \u0645\u06cc \u0633\u0627\u0632\u062f.<\/p>\n
\u067e\u0631\u0648\u0698\u0647 llama.cpp \u0628\u06cc\u0627\u0646 \u0645\u06cc \u06a9\u0646\u062f: \u0647\u062f\u0641 \u0627\u0635\u0644\u06cc llama.cpp \u0641\u0639\u0627\u0644 \u06a9\u0631\u062f\u0646 \u0627\u0633\u062a\u0646\u0628\u0627\u0637 LLM \u0628\u0627 \u062d\u062f\u0627\u0642\u0644 \u0631\u0627\u0647 \u0627\u0646\u062f\u0627\u0632\u06cc \u0648 \u0639\u0645\u0644\u06a9\u0631\u062f \u067e\u06cc\u0634\u0631\u0641\u062a\u0647 \u062f\u0631 \u0637\u06cc\u0641 \u06af\u0633\u062a\u0631\u062f\u0647 \u0627\u06cc \u0627\u0632 \u0633\u062e\u062a \u0627\u0641\u0632\u0627\u0631 – \u0628\u0647 \u0635\u0648\u0631\u062a \u0645\u062d\u0644\u06cc \u0648 \u062f\u0631 \u0641\u0636\u0627\u06cc \u0627\u0628\u0631\u06cc \u0627\u0633\u062a.<\/em>.<\/p>\n
\u0628\u06cc\u0627\u06cc\u06cc\u062f \u0622\u0646 \u0631\u0627 \u0627\u0645\u062a\u062d\u0627\u0646 \u06a9\u0646\u06cc\u0645.<\/p>\n
\n
from<\/span> txtai<\/span> import<\/span> Embeddings<\/span>\n\n# Create Embeddings with llama.cpp GGUF model\n<\/span>embeddings<\/span> =<\/span> Embeddings<\/span>(<\/span>\n    path<\/span>=<\/span>\"<\/span>second-state\/All-MiniLM-L6-v2-Embedding-GGUF\/all-MiniLM-L6-v2-Q4_K_M.gguf<\/span>\"<\/span>,<\/span>\n    content<\/span>=<\/span>True<\/span>\n)<\/span>\n\n# Load dataset\n<\/span>wikipedia<\/span> =<\/span> Embeddings<\/span>()<\/span>\nwikipedia<\/span>.<\/span>load<\/span>(<\/span>provider<\/span>=<\/span>\"<\/span>huggingface-hub<\/span>\"<\/span>,<\/span> container<\/span>=<\/span>\"<\/span>neuml\/txtai-wikipedia<\/span>\"<\/span>)<\/span>\n\nquery<\/span> =<\/span> \"\"\"<\/span>\nSELECT id, text FROM txtai\norder by percentile desc\nLIMIT 10000\n<\/span>\"\"\"<\/span>\n\n# Index dataset\n<\/span>embeddings<\/span>.<\/span>index<\/span>(<\/span>wikipedia<\/span>.<\/span>search<\/span>(<\/span>query<\/span>))<\/span>\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u0627\u06a9\u0646\u0648\u0646 \u06a9\u0647 \u067e\u0627\u06cc\u06af\u0627\u0647 \u062f\u0627\u062f\u0647 Embeddings \u0622\u0645\u0627\u062f\u0647 \u0627\u0633\u062a\u060c \u0628\u06cc\u0627\u06cc\u06cc\u062f \u06cc\u06a9 \u0639\u0628\u0627\u0631\u062a \u062c\u0633\u062a\u062c\u0648 \u0631\u0627 \u0627\u062c\u0631\u0627 \u06a9\u0646\u06cc\u0645.<\/p>\n
\n
embeddings<\/span>.<\/span>search<\/span>(<\/span>\"<\/span>Inventors of electric-powered devices<\/span>\"<\/span>)<\/span>\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\n
[{'id': 'Thomas Edison',\n  'text': 'Thomas Alva Edison (February 11, 1847October 18, 1931) was an American inventor and businessman. He developed many devices in fields such as electric power generation, mass communication, sound recording, and motion pictures. These inventions, which include the phonograph, the motion picture camera, and early versions of the electric light bulb, have had a widespread impact on the modern industrialized world. He was one of the first inventors to apply the principles of organized science and teamwork to the process of invention, working with many researchers and employees. He established the first industrial research laboratory.',\n  'score': 0.6758285164833069},\n {'id': 'Nikola Tesla',\n  'text': 'Nikola Tesla (; , ;  1856\\xa0\u2013 7 January 1943) was a Serbian-American inventor, electrical engineer, mechanical engineer, and futurist. He is best-known for his contributions to the design of the modern alternating current (AC) electricity supply system.',\n  'score': 0.6077840328216553},\n {'id': 'Alexander Graham Bell',\n  'text': 'Alexander Graham Bell (, born Alexander Bell; March 3, 1847 \u2013 August 2, 1922) was a  Scottish-born Canadian-American inventor, scientist and engineer who is credited with patenting the first practical telephone. He also co-founded the American Telephone and Telegraph Company (AT&T) in 1885.',\n  'score': 0.4573010802268982}]\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u0647\u0645\u0627\u0646\u0637\u0648\u0631 \u06a9\u0647 \u0645\u06cc \u0628\u06cc\u0646\u06cc\u0645\u060c \u0627\u06cc\u0646 \u067e\u0627\u06cc\u06af\u0627\u0647 \u062f\u0627\u062f\u0647 Embeddings \u062f\u0631\u0633\u062a \u0645\u0627\u0646\u0646\u062f \u0633\u0627\u06cc\u0631 \u067e\u0627\u06cc\u06af\u0627\u0647 \u062f\u0627\u062f\u0647 \u0647\u0627\u06cc Embeddings \u06a9\u0627\u0631 \u0645\u06cc \u06a9\u0646\u062f.  \u062a\u0641\u0627\u0648\u062a \u0627\u06cc\u0646 \u0627\u0633\u062a \u06a9\u0647 \u0628\u0647 \u062c\u0627\u06cc PyTorch \u0627\u0632 \u06cc\u06a9 \u0645\u062f\u0644 llama.cpp \u0628\u0631\u0627\u06cc \u0628\u0631\u062f\u0627\u0631\u06cc \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0645\u06cc \u06a9\u0646\u062f.<\/p>\n
\u0627\u0633\u062a\u0646\u062a\u0627\u062c LLM \u0628\u0627 llama.cpp \u06cc\u06a9 \u0648\u06cc\u0698\u06af\u06cc txtai \u062c\u062f\u06cc\u062f \u0646\u06cc\u0633\u062a.  \u062a\u063a\u06cc\u06cc\u0631\u0627\u062a \u0627\u062e\u06cc\u0631 \u0639\u0644\u0627\u0648\u0647 \u0628\u0631 \u062f\u0631\u062e\u0648\u0627\u0633\u062a\u200c\u0647\u0627\u06cc \u0627\u0633\u062a\u0627\u0646\u062f\u0627\u0631\u062f\u060c \u0627\u0632 \u067e\u06cc\u0627\u0645\u200c\u0647\u0627\u06cc \u0645\u06a9\u0627\u0644\u0645\u0647 \u067e\u0634\u062a\u06cc\u0628\u0627\u0646\u06cc \u0645\u06cc\u200c\u06a9\u0646\u062f.  \u0627\u06cc\u0646 \u0627\u0645\u0631 \u0646\u06cc\u0627\u0632 \u0628\u0647 \u062f\u0631\u06a9 \u0641\u0631\u0645\u062a \u0647\u0627\u06cc \u062f\u0631\u062e\u0648\u0627\u0633\u062a\u06cc \u0631\u0627 \u0627\u0632 \u0628\u06cc\u0646 \u0645\u06cc \u0628\u0631\u062f.<\/p>\n
\u0628\u06cc\u0627\u06cc\u06cc\u062f \u06cc\u06a9 \u0641\u0631\u0622\u06cc\u0646\u062f \u0628\u0627\u0632\u06cc\u0627\u0628\u06cc \u0646\u0633\u0644 \u0627\u0641\u0632\u0648\u062f\u0647 (RAG) \u0631\u0627 \u0627\u062c\u0631\u0627 \u06a9\u0646\u06cc\u0645 \u06a9\u0647 \u0628\u0647 \u0637\u0648\u0631 \u06a9\u0627\u0645\u0644 \u062a\u0648\u0633\u0637 \u0645\u062f\u0644\u200c\u0647\u0627\u06cc llama.cpp \u067e\u0634\u062a\u06cc\u0628\u0627\u0646\u06cc \u0645\u06cc\u200c\u0634\u0648\u062f.<\/p>\n
\u062a\u0648\u062c\u0647 \u0628\u0647 \u0627\u06cc\u0646 \u0646\u06a9\u062a\u0647 \u0645\u0647\u0645 \u0627\u0633\u062a \u06a9\u0647 \u067e\u06cc\u0627\u0645\u200c\u0647\u0627\u06cc \u0645\u06a9\u0627\u0644\u0645\u0647 \u0628\u0627 \u062a\u0645\u0627\u0645 \u067e\u0634\u062a\u06cc\u0628\u0627\u0646\u200c\u0647\u0627\u06cc LLM \u06a9\u0647 \u062a\u0648\u0633\u0637 txtai \u067e\u0634\u062a\u06cc\u0628\u0627\u0646\u06cc \u0645\u06cc\u200c\u0634\u0648\u0646\u062f \u06a9\u0627\u0631 \u0645\u06cc\u200c\u06a9\u0646\u0646\u062f (\u062a\u0631\u0627\u0646\u0633\u0641\u0648\u0631\u0645\u0627\u062a\u0648\u0631\u0647\u0627\u060c llama.cpp\u060c litellm).<\/em><\/p>\n
\n
from<\/span> txtai<\/span> import<\/span> LLM<\/span>\n\n# LLM instance\n<\/span>llm<\/span> =<\/span> LLM<\/span>(<\/span>path<\/span>=<\/span>\"<\/span>TheBloke\/Mistral-7B-OpenOrca-GGUF\/mistral-7b-openorca.Q4_K_M.gguf<\/span>\"<\/span>)<\/span>\n\n# Question and context\n<\/span>question<\/span> =<\/span> \"<\/span>Write a list of invented electric-powered devices<\/span>\"<\/span>\ncontext<\/span> =<\/span> \"<\/span>\\n<\/span>\"<\/span>.<\/span>join<\/span>(<\/span>x<\/span>[<\/span>\"<\/span>text<\/span>\"<\/span>]<\/span> for<\/span> x<\/span> in<\/span> embeddings<\/span>.<\/span>search<\/span>(<\/span>question<\/span>))<\/span>\n\n# Pass messages to LLM\n<\/span>response<\/span> =<\/span> llm<\/span>([<\/span>\n    {<\/span>\"<\/span>role<\/span>\"<\/span>:<\/span> \"<\/span>system<\/span>\"<\/span>,<\/span> \"<\/span>content<\/span>\"<\/span>:<\/span> \"<\/span>You are a friendly assistant. You answer questions from users.<\/span>\"<\/span>},<\/span>\n    {<\/span>\"<\/span>role<\/span>\"<\/span>:<\/span> \"<\/span>user<\/span>\"<\/span>,<\/span> \"<\/span>content<\/span>\"<\/span>:<\/span> f<\/span>\"\"\"<\/span>\nAnswer the following question using only the context below. Only include information specifically discussed.\n\nquestion: <\/span>{<\/span>question<\/span>}<\/span>\ncontext: <\/span>{<\/span>context<\/span>}<\/span>\n<\/span>\"\"\"<\/span>}<\/span>\n])<\/span>\nprint<\/span>(<\/span>response<\/span>)<\/span>\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\n
Based on the given context, here's a list of invented electric-powered devices:\n\n1. Electric light bulb by Thomas Edison\n2. Phonograph by Thomas Edison\n3. Motion picture camera by Thomas Edison\n4. Alternating current (AC) electricity supply system by Nikola Tesla\n5. Telephone by Alexander Graham Bell\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u0648 \u0647\u0645\u06cc\u0646\u0637\u0648\u0631\u060c RAG \u0628\u0627 llama.cpp\ud83e\udd99!<\/p>\n
\u062f\u0631 \u0645\u0631\u062d\u0644\u0647 \u0628\u0639\u062f\u060c \u0646\u0634\u0627\u0646 \u062e\u0648\u0627\u0647\u06cc\u0645 \u062f\u0627\u062f \u06a9\u0647 \u0686\u06af\u0648\u0646\u0647 \u06cc\u06a9 \u067e\u0627\u06cc\u06af\u0627\u0647 \u062f\u0627\u062f\u0647 Embeddings \u0645\u06cc \u062a\u0648\u0627\u0646\u062f \u0628\u0627 \u0633\u0631\u0648\u06cc\u0633 \u0647\u0627\u06cc API \u062e\u0627\u0631\u062c\u06cc \u0627\u0632 \u0637\u0631\u06cc\u0642 LiteLLM \u0627\u062f\u063a\u0627\u0645 \u0634\u0648\u062f.<\/p>\n
\u0628\u0647 \u0642\u0648\u0644 \u062e\u0648\u062f \u067e\u0631\u0648\u0698\u0647 LiteLLM: LiteLLM \u062a\u0639\u0627\u062f\u0644 \u0628\u0627\u0631\u060c \u0628\u0627\u0632\u06af\u0634\u062a \u0628\u0647 \u0639\u0642\u0628 \u0648 \u0631\u062f\u06cc\u0627\u0628\u06cc \u0647\u0632\u06cc\u0646\u0647 \u0631\u0627 \u062f\u0631 \u0628\u06cc\u0634 \u0627\u0632 100 LLM \u0627\u0646\u062c\u0627\u0645 \u0645\u06cc \u062f\u0647\u062f.  \u0647\u0645\u0647 \u062f\u0631 \u0642\u0627\u0644\u0628 OpenAI.<\/em><\/p>\n
\u0628\u06cc\u0627\u06cc\u06cc\u062f \u0627\u0628\u062a\u062f\u0627 \u06cc\u06a9 \u0633\u0631\u0648\u06cc\u0633 API \u0645\u062d\u0644\u06cc \u0631\u0627 \u0631\u0627\u0647 \u0627\u0646\u062f\u0627\u0632\u06cc \u06a9\u0646\u06cc\u0645 \u062a\u0627 \u0628\u0631\u0627\u06cc \u0627\u06cc\u0646 \u0646\u0633\u062e\u0647 \u0646\u0645\u0627\u06cc\u0634\u06cc \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u06a9\u0646\u06cc\u0645.<\/p>\n
\n
# Download models\nwget https:\/\/huggingface.co\/second-state\/All-MiniLM-L6-v2-Embedding-GGUF\/resolve\/main\/all-MiniLM-L6-v2-Q4_K_M.gguf\nwget https:\/\/huggingface.co\/TheBloke\/Mistral-7B-OpenOrca-GGUF\/resolve\/main\/mistral-7b-openorca.Q4_K_M.gguf\n\n# Start local API services\nnohup python -m llama_cpp.server --n_gpu_layers -1 --model all-MiniLM-L6-v2-Q4_K_M.gguf --host 127.0.0.1 --port 8000 &> vector.log &\nnohup python -m llama_cpp.server --n_gpu_layers -1 --model mistral-7b-openorca.Q4_K_M.gguf --chat_format chatml --host 127.0.0.1 --port 8001 &> llm.log &\nsleep 30\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u062d\u0627\u0644\u0627 \u0628\u06cc\u0627\u06cc\u06cc\u062f \u0628\u0647 \u0647\u0645 \u0645\u062a\u0635\u0644 \u0634\u0648\u06cc\u0645 \u0648 \u0627\u0632 \u0627\u06cc\u0646 \u0633\u0631\u0648\u06cc\u0633 \u0645\u062d\u0644\u06cc \u0628\u0631\u0627\u06cc \u062a\u0648\u0644\u06cc\u062f \u0628\u0631\u062f\u0627\u0631\u0647\u0627 \u0628\u0631\u0627\u06cc \u067e\u0627\u06cc\u06af\u0627\u0647 \u062f\u0627\u062f\u0647 Embeddings \u062c\u062f\u06cc\u062f \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u06a9\u0646\u06cc\u0645.  \u062a\u0648\u062c\u0647 \u062f\u0627\u0634\u062a\u0647 \u0628\u0627\u0634\u06cc\u062f \u06a9\u0647 \u0633\u0631\u0648\u06cc\u0633 \u0645\u062d\u0644\u06cc \u062f\u0631 \u0642\u0627\u0644\u0628 \u067e\u0627\u0633\u062e OpenAI \u067e\u0627\u0633\u062e \u0645\u06cc \u062f\u0647\u062f\u060c \u0627\u0632 \u0627\u06cc\u0646 \u0631\u0648 path<\/code> \u062a\u0646\u0638\u06cc\u0645 \u0632\u06cc\u0631<\/p>\n
\n
from<\/span> txtai<\/span> import<\/span> Embeddings<\/span>\n\n# Create Embeddings instance with external vectorization\n<\/span>embeddings<\/span> =<\/span> Embeddings<\/span>(<\/span>\n    path<\/span>=<\/span>\"<\/span>openai\/gpt-4-turbo<\/span>\"<\/span>,<\/span>\n    content<\/span>=<\/span>True<\/span>,<\/span>\n    vectors<\/span>=<\/span>{<\/span>\n        \"<\/span>api_base<\/span>\"<\/span>:<\/span> \"<\/span>http:\/\/localhost:8000\/v1<\/span>\"<\/span>,<\/span>\n        \"<\/span>api_key<\/span>\"<\/span>:<\/span> \"<\/span>sk-1234<\/span>\"<\/span>\n    }<\/span>\n)<\/span>\n\n# Load dataset\n<\/span>wikipedia<\/span> =<\/span> Embeddings<\/span>()<\/span>\nwikipedia<\/span>.<\/span>load<\/span>(<\/span>provider<\/span>=<\/span>\"<\/span>huggingface-hub<\/span>\"<\/span>,<\/span> container<\/span>=<\/span>\"<\/span>neuml\/txtai-wikipedia<\/span>\"<\/span>)<\/span>\n\nquery<\/span> =<\/span> \"\"\"<\/span>\nSELECT id, text FROM txtai\norder by percentile desc\nLIMIT 10000\n<\/span>\"\"\"<\/span>\n\n# Index dataset\n<\/span>embeddings<\/span>.<\/span>index<\/span>(<\/span>wikipedia<\/span>.<\/span>search<\/span>(<\/span>query<\/span>))<\/span>\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\n
embeddings<\/span>.<\/span>search<\/span>(<\/span>\"<\/span>Inventors of electric-powered devices<\/span>\"<\/span>)<\/span>\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\n
[{'id': 'Thomas Edison',\n  'text': 'Thomas Alva Edison (February 11, 1847October 18, 1931) was an American inventor and businessman. He developed many devices in fields such as electric power generation, mass communication, sound recording, and motion pictures. These inventions, which include the phonograph, the motion picture camera, and early versions of the electric light bulb, have had a widespread impact on the modern industrialized world. He was one of the first inventors to apply the principles of organized science and teamwork to the process of invention, working with many researchers and employees. He established the first industrial research laboratory.',\n  'score': 0.6758285164833069},\n {'id': 'Nikola Tesla',\n  'text': 'Nikola Tesla (; , ;  1856\\xa0\u2013 7 January 1943) was a Serbian-American inventor, electrical engineer, mechanical engineer, and futurist. He is best-known for his contributions to the design of the modern alternating current (AC) electricity supply system.',\n  'score': 0.6077840328216553},\n {'id': 'Alexander Graham Bell',\n  'text': 'Alexander Graham Bell (, born Alexander Bell; March 3, 1847 \u2013 August 2, 1922) was a  Scottish-born Canadian-American inventor, scientist and engineer who is credited with patenting the first practical telephone. He also co-founded the American Telephone and Telegraph Company (AT&T) in 1885.',\n  'score': 0.4573010802268982}]\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u0645\u0627\u0646\u0646\u062f \u0645\u062b\u0627\u0644 \u0642\u0628\u0644\u06cc \u0628\u0627 llama.cpp\u060c \u0627\u06cc\u0646 \u067e\u0627\u06cc\u06af\u0627\u0647 \u062f\u0627\u062f\u0647 Embeddings \u062f\u0642\u06cc\u0642\u0627\u064b \u06cc\u06a9\u0633\u0627\u0646 \u0631\u0641\u062a\u0627\u0631 \u0645\u06cc \u06a9\u0646\u062f.  \u062a\u0641\u0627\u0648\u062a \u0627\u0635\u0644\u06cc \u0627\u06cc\u0646 \u0627\u0633\u062a \u06a9\u0647 \u0645\u062d\u062a\u0648\u0627 \u0628\u0631\u0627\u06cc \u0628\u0631\u062f\u0627\u0631\u06cc \u0628\u0647 \u06cc\u06a9 \u0633\u0631\u0648\u06cc\u0633 \u062e\u0627\u0631\u062c\u06cc \u0627\u0631\u0633\u0627\u0644 \u0645\u06cc \u0634\u0648\u062f.<\/p>\n
\u0628\u0631\u0627\u06cc \u0622\u062e\u0631\u06cc\u0646 \u06a9\u0627\u0631\u0645\u0627\u0646\u060c \u06cc\u06a9 \u0641\u0631\u0622\u06cc\u0646\u062f \u0628\u0627\u0632\u06cc\u0627\u0628\u06cc \u0646\u0633\u0644 \u0627\u0641\u0632\u0648\u062f\u0647 (RAG) \u0631\u0627 \u0627\u062c\u0631\u0627 \u0645\u06cc \u06a9\u0646\u06cc\u0645 \u06a9\u0647 \u0628\u0647 \u0637\u0648\u0631 \u06a9\u0627\u0645\u0644 \u062a\u0648\u0633\u0637 \u06cc\u06a9 \u0633\u0631\u0648\u06cc\u0633 API \u062e\u0627\u0631\u062c\u06cc \u067e\u0634\u062a\u06cc\u0628\u0627\u0646\u06cc \u0645\u06cc \u0634\u0648\u062f.<\/p>\n
\n
from<\/span> txtai<\/span> import<\/span> LLM<\/span>\n\n# LLM instance\n<\/span>llm<\/span> =<\/span> LLM<\/span>(<\/span>path<\/span>=<\/span>\"<\/span>openai\/gpt-4-turbo<\/span>\"<\/span>,<\/span> api_base<\/span>=<\/span>\"<\/span>http:\/\/localhost:8001\/v1<\/span>\"<\/span>,<\/span> api_key<\/span>=<\/span>\"<\/span>sk-1234<\/span>\"<\/span>)<\/span>\n\n# Question and context\n<\/span>question<\/span> =<\/span> \"<\/span>Write a list of invented electric-powered devices<\/span>\"<\/span>\ncontext<\/span> =<\/span> \"<\/span>\\n<\/span>\"<\/span>.<\/span>join<\/span>(<\/span>x<\/span>[<\/span>\"<\/span>text<\/span>\"<\/span>]<\/span> for<\/span> x<\/span> in<\/span> embeddings<\/span>.<\/span>search<\/span>(<\/span>question<\/span>))<\/span>\n\n# Pass messages to LLM\n<\/span>response<\/span> =<\/span> llm<\/span>([<\/span>\n    {<\/span>\"<\/span>role<\/span>\"<\/span>:<\/span> \"<\/span>system<\/span>\"<\/span>,<\/span> \"<\/span>content<\/span>\"<\/span>:<\/span> \"<\/span>You are a friendly assistant. You answer questions from users.<\/span>\"<\/span>},<\/span>\n    {<\/span>\"<\/span>role<\/span>\"<\/span>:<\/span> \"<\/span>user<\/span>\"<\/span>,<\/span> \"<\/span>content<\/span>\"<\/span>:<\/span> f<\/span>\"\"\"<\/span>\nAnswer the following question using only the context below. Only include information specifically discussed.\n\nquestion: <\/span>{<\/span>question<\/span>}<\/span>\ncontext: <\/span>{<\/span>context<\/span>}<\/span>\n<\/span>\"\"\"<\/span>}<\/span>\n])<\/span>\nprint<\/span>(<\/span>response<\/span>)<\/span>\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\n
Based on the given context, a list of invented electric-powered devices includes:\n\n1. Phonograph by Thomas Edison\n2. Motion Picture Camera by Thomas Edison\n3. Early versions of the Electric Light Bulb by Thomas Edison\n4. AC (Alternating Current) Electricity Supply System by Nikola Tesla\n5. Telephone by Alexander Graham Bell\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
txtai \u0627\u0632 \u062a\u0639\u062f\u0627\u062f\u06cc \u0628\u0627\u0637\u0646 \u0645\u062e\u062a\u0644\u0641 \u0648\u06a9\u062a\u0648\u0631 \u0648 LLM \u067e\u0634\u062a\u06cc\u0628\u0627\u0646\u06cc \u0645\u06cc \u06a9\u0646\u062f.  \u0631\u0648\u0634 \u067e\u06cc\u0634\u200c\u0641\u0631\u0636 \u0627\u0632 \u0645\u062f\u0644\u200c\u0647\u0627\u06cc PyTorch \u0627\u0632 \u0637\u0631\u06cc\u0642 \u06a9\u062a\u0627\u0628\u062e\u0627\u0646\u0647 Hugging Face Transformers \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0645\u06cc\u200c\u06a9\u0646\u062f.  \u0627\u06cc\u0646 \u0645\u0642\u0627\u0644\u0647 \u0646\u0634\u0627\u0646 \u062f\u0627\u062f \u06a9\u0647 \u0686\u06af\u0648\u0646\u0647 \u0645\u06cc\u200c\u062a\u0648\u0627\u0646 \u0627\u0632 llama.cpp \u0648 \u0633\u0631\u0648\u06cc\u0633\u200c\u0647\u0627\u06cc API \u062e\u0627\u0631\u062c\u06cc \u0646\u06cc\u0632 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u06a9\u0631\u062f.<\/p>\n
\u0627\u06cc\u0646 \u067e\u0634\u062a\u06cc\u0628\u0627\u0646\u200c\u0647\u0627\u06cc \u0628\u0631\u062f\u0627\u0631\u06cc \u0627\u0636\u0627\u0641\u06cc \u0648 LLM \u062d\u062f\u0627\u06a9\u062b\u0631 \u0627\u0646\u0639\u0637\u0627\u0641\u200c\u067e\u0630\u06cc\u0631\u06cc \u0648 \u0645\u0642\u06cc\u0627\u0633\u200c\u067e\u0630\u06cc\u0631\u06cc \u0631\u0627 \u0645\u0645\u06a9\u0646 \u0645\u06cc\u200c\u0633\u0627\u0632\u0646\u062f.  \u0628\u0647 \u0639\u0646\u0648\u0627\u0646 \u0645\u062b\u0627\u0644\u060c \u0628\u0631\u062f\u0627\u0631\u0633\u0627\u0632\u06cc \u0631\u0627 \u0645\u06cc \u062a\u0648\u0627\u0646 \u0628\u0647 \u0637\u0648\u0631 \u06a9\u0627\u0645\u0644 \u062f\u0631 \u06cc\u06a9 \u0633\u0631\u0648\u06cc\u0633 API \u062e\u0627\u0631\u062c\u06cc \u06cc\u0627 \u06cc\u06a9 \u0633\u0631\u0648\u06cc\u0633 \u0645\u062d\u0644\u06cc \u062f\u06cc\u06af\u0631 \u0628\u0627\u0631\u06af\u0630\u0627\u0631\u06cc \u06a9\u0631\u062f.  llama.cpp \u0627\u0632 \u062f\u0633\u062a\u06af\u0627\u0647\u200c\u0647\u0627\u06cc macOS\u060c \u0634\u062a\u0627\u0628\u200c\u062f\u0647\u0646\u062f\u0647\u200c\u0647\u0627\u06cc \u062c\u0627\u06cc\u06af\u0632\u06cc\u0646 \u0645\u0627\u0646\u0646\u062f \u067e\u0631\u062f\u0627\u0632\u0646\u062f\u0647\u200c\u0647\u0627\u06cc \u06af\u0631\u0627\u0641\u06cc\u06a9\u06cc AMD ROCm \/ Intel \u067e\u0634\u062a\u06cc\u0628\u0627\u0646\u06cc \u0645\u06cc\u200c\u06a9\u0646\u062f \u0648 \u0628\u0631 \u0631\u0648\u06cc \u062f\u0633\u062a\u06af\u0627\u0647\u200c\u0647\u0627\u06cc Raspberry Pi \u0627\u062c\u0631\u0627 \u0645\u06cc\u200c\u0634\u0648\u062f.<\/p>\n
\u062f\u06cc\u062f\u0646 \u062a\u0644\u0627\u0642\u06cc \u0647\u0645\u0647 \u0627\u06cc\u0646 \u067e\u06cc\u0634\u0631\u0641\u062a \u0647\u0627\u06cc \u062c\u062f\u06cc\u062f \u062f\u0631 \u06a9\u0646\u0627\u0631 \u0647\u0645 \u0647\u06cc\u062c\u0627\u0646 \u0627\u0646\u06af\u06cc\u0632 \u0627\u0633\u062a.  \u062f\u0631 \u0627\u062f\u0627\u0645\u0647 \u0628\u0627 \u0645\u0627 \u0647\u0645\u0631\u0627\u0647 \u0628\u0627\u0634\u06cc\u062f!<\/p>\n<\/p><\/div>\n","protected":false},"excerpt":{"rendered":"
txtai \u06cc\u06a9 \u067e\u0627\u06cc\u06af\u0627\u0647 \u062f\u0627\u062f\u0647 \u062a\u0639\u0628\u06cc\u0647\u200c\u0634\u062f\u0647 \u06cc\u06a9\u067e\u0627\u0631\u0686\u0647 \u0628\u0631\u0627\u06cc \u062c\u0633\u062a\u062c\u0648\u06cc \u0645\u0639\u0646\u0627\u06cc\u06cc\u060c \u0627\u0631\u06a9\u0633\u062a\u0631\u0627\u0633\u06cc\u0648\u0646 LLM \u0648 \u06af\u0631\u062f\u0634\u200c\u0647\u0627\u06cc \u06a9\u0627\u0631\u06cc \u0645\u062f\u0644 \u0632\u0628\u0627\u0646 \u0627\u0633\u062a. txtai \u0647\u0645\u06cc\u0634\u0647 \u06cc\u06a9 \u0686\u0627\u0631\u0686\u0648\u0628 \u0645\u062d\u0644\u06cc \u0628\u0648\u062f\u0647 \u0648 \u062e\u0648\u0627\u0647\u062f \u0628\u0648\u062f. \u062f\u0631 \u0627\u0635\u0644 \u0628\u0631\u0627\u06cc \u0627\u062c\u0631\u0627\u06cc \u0645\u062f\u0644\u200c\u0647\u0627 \u0628\u0631 \u0631\u0648\u06cc \u0633\u062e\u062a\u200c\u0627\u0641\u0632\u0627\u0631 \u0645\u062d\u0644\u06cc \u0628\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 \u062a\u0631\u0627\u0646\u0633\u0641\u0648\u0631\u0645\u0627\u062a\u0648\u0631 Hugging Face \u0637\u0631\u0627\u062d\u06cc \u0634\u062f\u0647 \u0628\u0648\u062f. \u0647\u0645\u0627\u0646\u0637\u0648\u0631 \u06a9\u0647 \u0641\u0636\u0627\u06cc \u0647\u0648\u0634 \u0645\u0635\u0646\u0648\u0639\u06cc \u0637\u06cc \u0633\u0627\u0644 \u06af\u0630\u0634\u062a\u0647 \u062a\u06a9\u0627\u0645\u0644 \u06cc\u0627\u0641\u062a\u0647 \u0627\u0633\u062a\u060c …<\/p>\n","protected":false},"author":2,"featured_media":64547,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"https:\/\/media.dev.to\/cdn-cgi\/image\/width=1000,height=500,fit=cover,gravity=auto,format=auto\/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2F73nkcxci43cuujei0lnu.png","fifu_image_alt":"","footnotes":""},"categories":[339],"tags":[],"class_list":["post-64546","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-dev"],"_links":{"self":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/posts\/64546","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/comments?post=64546"}],"version-history":[{"count":0,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/posts\/64546\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/media\/64547"}],"wp:attachment":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/media?parent=64546"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/categories?post=64546"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/tags?post=64546"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}