Class: HuggingFaceInferenceAPI
Wrapper on the Hugging Face's Inference API. API Docs: https://huggingface.co/docs/huggingface.js/inference/README List of tasks with models: huggingface.co/api/tasks
Note that Conversational API is not yet supported by the Inference API. They recommend using the text generation API instead. See: https://github.com/huggingface/huggingface.js/issues/586#issuecomment-2024059308
Extends
BaseLLM
Constructors
new HuggingFaceInferenceAPI()
new HuggingFaceInferenceAPI(
init
):HuggingFaceInferenceAPI
Parameters
• init: HFConfig
Returns
Overrides
BaseLLM.constructor
Source
packages/core/src/llm/huggingface.ts:87
Properties
contextWindow
contextWindow:
number
Source
packages/core/src/llm/huggingface.ts:84
hf
hf:
HfInference
Source
packages/core/src/llm/huggingface.ts:85
maxTokens?
optional
maxTokens:number
Source
packages/core/src/llm/huggingface.ts:83
model
model:
string
Source
packages/core/src/llm/huggingface.ts:80
temperature
temperature:
number
Source
packages/core/src/llm/huggingface.ts:81
topP
topP:
number
Source
packages/core/src/llm/huggingface.ts:82
Accessors
metadata
get
metadata():LLMMetadata
Returns
Source
packages/core/src/llm/huggingface.ts:108
Methods
chat()
chat(params)
chat(
params
):Promise
<AsyncIterable
<ChatResponseChunk
>>
Parameters
• params: LLMChatParamsStreaming
<object
, object
>
Returns
Promise
<AsyncIterable
<ChatResponseChunk
>>
Overrides
BaseLLM.chat
Source
packages/core/src/llm/huggingface.ts:119
chat(params)
chat(
params
):Promise
<ChatResponse
<object
>>
Parameters
• params: LLMChatParamsNonStreaming
<object
, object
>
Returns
Promise
<ChatResponse
<object
>>
Overrides
BaseLLM.chat
Source
packages/core/src/llm/huggingface.ts:122
complete()
complete(params)
complete(
params
):Promise
<AsyncIterable
<CompletionResponse
>>
Parameters
• params: LLMCompletionParamsStreaming
Returns
Promise
<AsyncIterable
<CompletionResponse
>>
Inherited from
BaseLLM.complete
Source
packages/core/src/llm/base.ts:22
complete(params)
complete(
params
):Promise
<CompletionResponse
>
Parameters
• params: LLMCompletionParamsNonStreaming