Package com.tnsai.llm

package com.tnsai.llm

LLM (Large Language Model) client interfaces and abstractions.

This package provides a unified interface for interacting with various LLM providers. The LLMClient interface defines the contract that all provider implementations must follow.

Architecture

 ┌─────────────────────────────────────────────────────────────┐
 │                      LLMClient                               │
 │              (Interface - this package)                      │
 └──────────────────────────┬──────────────────────────────────┘
                            │
        ┌───────────────────┼───────────────────┐
        │                   │                   │
        ▼                   ▼                   ▼
 ┌─────────────┐    ┌─────────────┐    ┌─────────────┐
 │OpenAIClient │    │AnthropicCl  │    │GeminiClient │
 └─────────────┘    └─────────────┘    └─────────────┘
        │                   │                   │
        └───────────────────┴───────────────────┘
                            │
                            ▼
              ┌─────────────────────────┐
              │  com.tnsai.llm.providers │
              │  (Implementation package) │
              └─────────────────────────┘
 

Supported Providers

Provider	Class	Models
OpenAI	OpenAIClient	GPT-4, GPT-3.5-turbo
Azure OpenAI	AzureOpenAIClient	GPT-4, GPT-3.5
Anthropic	AnthropicClient	Claude 3, Claude 2
Google	GeminiClient	Gemini Pro, Gemini Ultra
Ollama	OllamaClient	Llama 2, Mistral, etc.
Groq	GroqClient	Llama, Mixtral
OpenRouter	OpenRouterClient	Multiple providers
Mistral	MistralClient	Mistral models
AWS Bedrock	BedrockClient	Claude, Llama

Key Features

• Unified API: Same interface for all providers
• Tool Support: Function/tool calling across providers
• Retry Logic: Built-in exponential backoff for resilience
• Streaming: Support for streaming responses (provider-dependent)
• Multimodal: Image support for capable models (Gemini, GPT-4V)

Usage Example

 // Create a client using factory method
 LLMClient client = LLMClient.create("openai");

 // Or create directly
 LLMClient client = new OpenAIClient("gpt-4");

 // Simple chat
 String response = client.chat("What is the capital of France?");

 // Chat with messages
 List<ChatMessage> messages = List.of(
     ChatMessage.system("You are a helpful assistant"),
     ChatMessage.user("Hello!")
 );
 ChatMessage response = client.chat(messages);

 // Chat with tools
 List<Tool> tools = List.of(new Tool("calculator", "Perform math"));
 ChatMessage response = client.chat(messages, tools);
 

Configuration

API keys can be provided via:

• Environment variables (e.g., OPENAI_API_KEY)
• .env file in project root
• Constructor parameters

See Also:

• LLMClient
• invalid reference

  com.tnsai.llm.providers

• ChatResponse

Related Packages

Package	Description
com.tnsai	TnsAI Java SDK - A multi-agent framework with BDI architecture and Gaia methodology.

Class	Description
ChatChunk	Represents a single chunk in a streaming LLM response.
ChatChunk.ChunkType	The type of streaming chunk.
ChatChunk.FinishReason	The reason why generation stopped.
ChatRequest	Unified request model for LLM chat operations.
ChatRequest.Builder	Fluent builder for ChatRequest.
ChatResponse	Represents a chat response from an LLM, potentially including tool calls.
ChatResponse.ToolCall	Represents a tool/function call from the LLM.
FallbackLLMClient	LLM client with automatic fallback support.
FallbackLLMClient.Builder	Builder for FallbackLLMClient.
LLMCapabilities	Describes the capabilities of an LLM provider/model.
LLMCapabilities.Modality	Input modalities the model can process.
LLMClient	Base interface for LLM (Large Language Model) client implementations.
LLMClientProvider	SPI for creating LLM clients from provider configuration.
LLMResolver	Resolves the appropriate LLM client using a fallback chain.
LLMResolver.Builder	Builder for LLM resolution with full fallback chain support.
LLMResolver.LLMSpec	LLM configuration record for passing model settings.
LLMResolver.ResolutionResult	Result of LLM resolution with source tracking.
LLMResolver.ResolutionSource	Source of the resolved LLM in the fallback chain.
LLMSpecSource	Common interface for LLM configuration sources.
ModelCapabilities	Default implementation of LLMCapabilities with builder pattern.
ModelInfo	Represents information about an LLM model from a provider's API.
ModelInfo.Builder
ParallelLLM	Parallel LLM execution utilities.
ParallelLLM.LLMResult	Result of a parallel LLM call.
ParallelLLM.ParallelResults	Aggregated results from multiple LLMs.