tessl/maven-dev-langchain4j--langchain4j-hugging-face
LangChain4j integration library for Hugging Face inference capabilities including chat, language, and embedding models
language-model.mddocs/
Language Model API
Complete API reference for HuggingFaceLanguageModel (deprecated).
Deprecation Notice
This class is deprecated since version 1.7.0-beta13 and scheduled for removal.
Migration: Use OpenAiChatModel from langchain4j-open-ai module with Hugging Face's OpenAI-compatible endpoint.
See Migration Guide for complete details.
Overview
Provides simple text generation from prompts using Hugging Face language models.
Package: dev.langchain4j.model.huggingface
Status: ⚠️ Deprecated (scheduled for removal)
Interfaces: LanguageModel
Class Signature
package dev.langchain4j.model.huggingface;
@Deprecated(forRemoval = true, since = "1.7.0-beta13")
public class HuggingFaceLanguageModel implements dev.langchain4j.model.language.LanguageModel {
// Construction
public static Builder builder();
public static HuggingFaceLanguageModel withAccessToken(String accessToken);
// Generation operations
public dev.langchain4j.model.output.Response<String> generate(String prompt);
public dev.langchain4j.model.output.Response<String> generate(dev.langchain4j.model.input.Prompt prompt);
}Builder API
public static final class Builder {
public Builder baseUrl(String baseUrl);
public Builder accessToken(String accessToken);
public Builder modelId(String modelId);
public Builder timeout(java.time.Duration timeout);
public Builder temperature(Double temperature);
public Builder maxNewTokens(Integer maxNewTokens);
public Builder returnFullText(Boolean returnFullText);
public Builder waitForModel(Boolean waitForModel);
public HuggingFaceLanguageModel build();
}Construction
builder()
public static Builder builder()Example:
HuggingFaceLanguageModel model = HuggingFaceLanguageModel.builder()
.accessToken(System.getenv("HF_API_KEY"))
.modelId("microsoft/Phi-3-mini-4k-instruct")
.temperature(0.8)
.maxNewTokens(150)
.build();withAccessToken()
public static HuggingFaceLanguageModel withAccessToken(String accessToken)Quick construction with defaults.
Generation Operations
generate(String)
public dev.langchain4j.model.output.Response<String> generate(String prompt)Parameters:
prompt- Input prompt for generation
Returns: Response containing generated text
Throws: RuntimeException on API errors
Example:
Response<String> response = model.generate("Write a haiku:");
String text = response.content();generate(Prompt)
public dev.langchain4j.model.output.Response<String> generate(dev.langchain4j.model.input.Prompt prompt)Parameters:
prompt- Prompt object
Returns: Response containing generated text
Configuration
See Configuration Guide for all options.
Key Parameters:
accessToken(required) - API keymodelId(recommended) - Model identifiertemperature(optional) - Sampling temperature (0.0-2.0)maxNewTokens(optional) - Max tokens to generatereturnFullText(optional) - Include prompt (default: false)waitForModel(optional) - Wait if loading (default: true)timeout(optional) - Request timeout (default: 15s)baseUrl(optional) - Custom endpoint
Prompt Type
package dev.langchain4j.model.input;
public class Prompt {
public String text();
public static Prompt from(String text);
}Response Type
package dev.langchain4j.model.output;
public class Response<T> {
public T content();
public static <T> Response<T> from(T content);
}Usage Examples
Basic Generation
HuggingFaceLanguageModel model = HuggingFaceLanguageModel.builder()
.accessToken(System.getenv("HF_API_KEY"))
.modelId("tiiuae/falcon-7b-instruct")
.build();
Response<String> response = model.generate("Explain AI:");
String text = response.content();Code Generation
HuggingFaceLanguageModel model = HuggingFaceLanguageModel.builder()
.accessToken(System.getenv("HF_API_KEY"))
.modelId("microsoft/Phi-3-mini-4k-instruct")
.temperature(0.2) // Lower for deterministic code
.maxNewTokens(200)
.build();
String prompt = "Write a Java function to check if prime:\n";
String code = model.generate(prompt).content();Creative Writing
HuggingFaceLanguageModel model = HuggingFaceLanguageModel.builder()
.accessToken(System.getenv("HF_API_KEY"))
.modelId("tiiuae/falcon-7b-instruct")
.temperature(1.0) // Higher for creativity
.maxNewTokens(300)
.build();
String story = model.generate("Write a story about AI:").content();Using Prompt Template
import dev.langchain4j.model.input.*;
import java.util.Map;
PromptTemplate template = PromptTemplate.from(
"Write a {{length}} story about {{topic}}:"
);
Prompt prompt = template.apply(Map.of(
"length", "short",
"topic", "robots"
));
Response<String> response = model.generate(prompt);Error Handling
try {
Response<String> response = model.generate(prompt);
} catch (RuntimeException e) {
// Error format: "status code: <code>; body: <body>"
System.err.println("Generation failed: " + e.getMessage());
}Common Errors:
- 401: Invalid token
- 404: Model not found
- 429: Rate limiting
- 503: Model loading
Comparison with ChatModel
LanguageModel (this class):
- Simple prompt-to-completion
- No conversation history
- Single-turn text generation
ChatModel (HuggingFaceChatModel):
- Multi-turn conversation support
- Structured message types
- Conversation history
Both deprecated. Use OpenAiChatModel for new code.
Recommended Models
| Model | Size | Use Case |
|---|---|---|
tiiuae/falcon-7b-instruct | 7B | General purpose |
microsoft/Phi-3-mini-4k-instruct | Small | Efficient, 4K context |
mistralai/Mistral-7B-Instruct-v0.2 | 7B | High quality |
HuggingFaceH4/zephyr-7b-beta | 7B | Conversational |
Related Documentation
- Migration Guide - Migrate to OpenAI module
- Configuration Guide - Configuration options
- Error Handling - Error scenarios
- Common Tasks - Usage examples
- Chat Model API - Alternative (also deprecated)
Install with Tessl CLI
npx tessl i tessl/maven-dev-langchain4j--langchain4j-hugging-face