HuggingFace Inference API — ML models for text, vision and NLP. Operations: generate (text generation), classify (text classification), ner (named entity recognition), summarize, translate, embed (feature extraction), qa (question answering), zero_shot (zero-shot classification), image_classify, search_models. API key optional for some models.
huggingfacenlptransformersinferenceai
Input Parameters
Parameter
Type
Description
operationrequired
string
Operation to perform. One of: generate, classify, ner, summarize, translate, embed, qa, zero_shot, image_classify, search_models.
apiKeyoptional
string
HuggingFace API token (required for gated models).
modeloptional
string
Model ID (e.g. gpt2, facebook/bart-large-cnn). Each operation has a default.
textoptional
string
Input text for generate, classify, ner, summarize, qa, zero_shot.
textsoptional
string
JSON array of strings for batch embed or classify.
contextoptional
string
Context text for qa (question answering) operation.
questionoptional
string
Question for qa operation.
labelsoptional
string
Comma-separated candidate labels for zero_shot classification.
srcLangoptional
string
Source language code for translate (e.g. en).
tgtLangoptional
string
Target language code for translate (e.g. fr, de, es).
maxTokensoptional
integer
Max new tokens to generate. Default: 200.
minTokensoptional
integer
Min new tokens to generate. Default: 0.
temperatureoptional
number
Sampling temperature (default: 1.0). Default: 1.0.
searchoptional
string
Search query for search_models.
taskoptional
string
Task filter for search_models (e.g. text-generation, translation).
limitoptional
integer
Max models to return for search_models (default: 10). Default: 10.
imageUrloptional
string
Image URL for image_classify operation.
Output Fields
Field
Type
Description
generated_text
string
results
array[object]
answer
string
embeddings
array[any]
error
string
Examples
Text generation
// Input
{
"operation": "generate",
"model": "gpt2",
"text": "The future of AI is",
"maxTokens": 50
}
// Output
{
"generated_text": "The future of AI is bright and..."
}
Install & Discovery
Install
ppm install mcp-huggingface
Get JSON Schema
GET /v1/packages/mcp-huggingface/1.0.0/schema
Discover by keyword
GET /v1/mcp/discover?q=huggingface
Discovery hint: Install with ppm install mcp-huggingface or invoke remotely via POST /v1/invoke/mcp-huggingface on the MCP Service.
PascalAI Usage
uses toolslib;
var Tool := LoadTool('mcp-huggingface');
var R := Tool.Call(JsonObj(['operation','generate','model','gpt2','text','Hello world','maxTokens',50]));
Writeln(R.ToJSON);