feat: add huggingface client

Argcfile.sh

@@ -85,7 +85,10 @@ OPENAI_COMPATIBLE_PLATFORMS=( \
   deepinfra,meta-llama/Meta-Llama-3.1-8B-Instruct,https://api.deepinfra.com/v1/openai \
   deepseek,deepseek-chat,https://api.deepseek.com \
   fireworks,accounts/fireworks/models/llama-v3p1-8b-instruct,https://api.fireworks.ai/inference/v1 \
+  github,gpt-4o-mini,https://models.inference.ai.azure.com \
   groq,llama3-8b-8192,https://api.groq.com/openai/v1 \
+  huggingface,meta-llama/Meta-Llama-3-8B-Instruct,https://api-inference.huggingface.co/v1 \
+  lingyiwanwu,yi-large,https://api.lingyiwanwu.com/v1 \
   mistral,open-mistral-nemo,https://api.mistral.ai/v1 \
   moonshot,moonshot-v1-8k,https://api.moonshot.cn/v1 \
   openrouter,openai/gpt-4o-mini,https://openrouter.ai/api/v1 \
@@ -95,8 +98,6 @@ OPENAI_COMPATIBLE_PLATFORMS=( \
   qianwen,qwen-turbo,https://dashscope.aliyuncs.com/compatible-mode/v1 \
   together,meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo,https://api.together.xyz/v1 \
   zhipuai,glm-4-0520,https://open.bigmodel.cn/api/paas/v4 \
-  lingyiwanwu,yi-large,https://api.lingyiwanwu.com/v1 \
-  github,gpt-4o-mini,https://models.inference.ai.azure.com \
 )
 
 # @cmd Chat with any LLM api 

config.example.yaml

@@ -228,6 +228,12 @@ clients:
     api_base: https://api.cloudflare.com/client/v4/accounts/{ACCOUNT_ID}/ai/v1
     api_key: xxx
 
+  # See https://huggingface.co/inference-api/serverless
+  - type: openai-compatible
+    name: huggingface
+    api_base: https://api-inference.huggingface.co/v1
+    api_key: hf_xxx
+
   # See https://replicate.com/docs
   - type: replicate
     api_key: xxx
@@ -303,6 +309,8 @@ clients:
     api_base: https://api.together.xyz/v1
     api_key: xxx
 
+  # ----- RAG dedicated -----
+
   # See https://jina.ai
   - type: openai-compatible
     name: jina

models.yaml

@@ -542,6 +542,24 @@
       default_chunk_size: 1000
       max_batch_size: 100
 
+# Links:
+#  - https://huggingface.co/models?inference=warm&pipeline_tag=text-generation&other=text-generation-inference&sort=trending
+#  - https://huggingface.co/docs/text-generation-inference/en/reference/api_reference
+- platform: huggingface
+  models:
+    - name: meta-llama/Meta-Llama-3-8B-Instruct
+      max_input_tokens: 8192
+      max_output_tokens: 4096
+      require_max_tokens: true
+      input_price: 0
+      output_price: 0
+    - name: mistralai/Mistral-Nemo-Instruct-2407
+      max_input_tokens: 128000
+      max_output_tokens: 4096
+      require_max_tokens: true
+      input_price: 0
+      output_price: 0
+
 # Links:
 #  - https://replicate.com/explore
 #  - https://replicate.com/pricing

src/client/mod.rs

@@ -37,15 +37,15 @@ register_client!(
     (ernie, "ernie", ErnieConfig, ErnieClient),
 );
 
-pub const OPENAI_COMPATIBLE_PLATFORMS: [(&str, &str); 19] = [
+pub const OPENAI_COMPATIBLE_PLATFORMS: [(&str, &str); 20] = [
     ("ai21", "https://api.ai21.com/studio/v1"),
     ("cloudflare", ""),
     ("deepinfra", "https://api.deepinfra.com/v1/openai"),
     ("deepseek", "https://api.deepseek.com"),
     ("fireworks", "https://api.fireworks.ai/inference/v1"),
     ("github", "https://models.inference.ai.azure.com"),
     ("groq", "https://api.groq.com/openai/v1"),
-    ("jina", "https://api.jina.ai/v1"),
+    ("huggingface", "https://api-inference.huggingface.co/v1"),
     ("lingyiwanwu", "https://api.lingyiwanwu.com/v1"),
     ("mistral", "https://api.mistral.ai/v1"),
     ("moonshot", "https://api.moonshot.cn/v1"),
@@ -59,5 +59,7 @@ pub const OPENAI_COMPATIBLE_PLATFORMS: [(&str, &str); 19] = [
     ),
     ("together", "https://api.together.xyz/v1"),
     ("zhipuai", "https://open.bigmodel.cn/api/paas/v4"),
+    // RAG-dedicated
+    ("jina", "https://api.jina.ai/v1"),
     ("voyageai", "https://api.voyageai.com/v1"),
 ];