From 1ec6abfaee2fdc189b348b7e3a8145bd9a84da74 Mon Sep 17 00:00:00 2001 From: sigoden Date: Wed, 5 Jun 2024 09:02:23 +0800 Subject: feat: support RAG (#560) * feat: support RAG * support more embeddings models and implement concurrent embedding api * show the progress of addings paths * ignore embedding context when saving message * embedding model max_chunk_size => default_chunk_size * support pdf and pandoc formats (docx, epub, ipynb) --- models.yaml | 43 +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 43 insertions(+) (limited to 'models.yaml') diff --git a/models.yaml b/models.yaml index c306f1b..f4c980b 100644 --- a/models.yaml +++ b/models.yaml @@ -65,6 +65,16 @@ max_output_tokens: 4096 input_price: 60 output_price: 120 + - name: text-embedding-3-large + mode: embedding + max_input_tokens: 8191 + default_chunk_size: 8000 + max_concurrent_chunks: 100 + - name: text-embedding-3-small + mode: embedding + max_input_tokens: 8191 + default_chunk_size: 8000 + max_concurrent_chunks: 100 - platform: gemini # docs: @@ -100,6 +110,10 @@ output_price: 10.5 supports_vision: true supports_function_calling: true + - name: text-embedding-004 + mode: embedding + max_input_tokens: 2048 + default_chunk_size: 2000 - platform: claude # docs: @@ -162,6 +176,10 @@ input_price: 8 output_price: 24 supports_function_calling: true + - name: mistral-embed + mode: embedding + max_input_tokens: 8092 + default_chunk_size: 8000 - platform: cohere # docs: @@ -183,6 +201,16 @@ input_price: 3 output_price: 15 supports_function_calling: true + - name: embed-english-v3.0 + mode: embedding + max_input_tokens: 512 + default_chunk_size: 1000 + max_concurrent_chunks: 96 + - name: embed-multilingual-v3.0 + mode: embedding + max_input_tokens: 512 + default_chunk_size: 1000 + max_concurrent_chunks: 96 - platform: perplexity # docs: @@ -281,6 +309,16 @@ input_price: 1.25 output_price: 3.75 supports_vision: true + - name: text-embedding-004 + mode: embedding + max_input_tokens: 3072 + default_chunk_size: 3000 + max_concurrent_chunks: 5 + - name: text-multilingual-embedding-002 + mode: embedding + max_input_tokens: 3072 + default_chunk_size: 3000 + max_concurrent_chunks: 5 - platform: vertexai-claude # docs: @@ -509,6 +547,11 @@ input_price: 2.8 output_price: 2.8 supports_vision: true + - name: text-embedding-v2 + mode: embedding + max_input_tokens: 2048 + default_chunk_size: 2000 + max_concurrent_chunks: 5 - platform: moonshot # docs: -- cgit v1.2.3