Doc Revamping - Table of content proposal

[cescoffier]

Hello,

The current Quarkus langchain4j documentation is a bit outdated and needs some restructuring. Here is a proposal for a new structure. Some of the new documents have been written already in #1549.

Getting Started

• Quick start - Simple CLI asking a question to openai - new but written
• Implementing a summarization service - new but written
• Extracting text from an image - new but written
• Implementing a simple RAG application - new but written
• Using function calling - new but written

Guides (How to, beyond getting started, but single task)

• Prompt engineering patterns - revamped, written
• Returning structured data from a model - new, todo
• Streaming responses - new, todo
• Using function calling - new, todo
• Logging Model Interactions - new, todo
• Implementing Agentic patterns - new, todo, was planning to reuse @mariofusco' content
• Tracking token usages - new, todo
• Using few-shots in prompts - new, todo
• Using local chat models - new, todo
• Using in-process chat models - new, todo
• Passing Images to models - new, todo
• Generating Images from Prompts - new, todo
• Testing AI-Infused applications - new, todo
• Evaluating AI-Infused applications - new, todo
• Passing knowledge to AI models - new, todo (RAG, function calling, prompt template...)
• Handling error and timeouts - new, todo

Reference

• AI Services - new but written

  • Prompt and template - todo
  • Messages, Memory and Context Windows - todo
  • Function Calling - todo
    • WebSearch Tools - Reuse
  • Guardrails - Reuse
  • Enabling and Disabling Integrations - Reuse
  • Model selection - Reuse
  • Response Augmenter - Reuse
  • Fault-Tolerance - Reuse

• Models - new but written

  • Chat Models - basics + for each: config, example... (the list should be sorted alphabetically)
    • Amazon Bedrock (must be implemented)
    • Anthropic (Claude) - must be revamped
    • Azure OpenAI - must be revamped
    • HuggingFace - must be revamped
    • IBM watsonx.ai - must be revamped
    • Jlama - must be revamped
    • Llama3.java - must be revamped
    • Mistral AI - must be revamped
    • Ollama - must be revamped
    • OpenAI - done
    • Podman AI Lab - must be revamped
  • Image Models
  • OpenAI - must be revamped
  • to be determined and tested
  • Embedding Models
    • ONXX (In-Process Embeddings) - must be revamped
    • OpenAI - Done
    • to be determined and tested
  • Moderation Models
    • OpenAI - todo
    • to be determined and tested

• Retrieved Augmented Generation (RAG)

  • RAG Principles - todo
  • Ingestion - todo
  • Augmentation - todo
  • EasyRAG - reuse
  • Contextual RAG - todo
  • Evaluating RAG Quality - todo
  • Reranking - Reuse
  • Document and Vector Stores (reuse)
    • Chroma Store
    • CSV files - Maybe a guide
    • Infinispan Store
    • Milvus Store
    • Neo4j Store
    • PgVector (PostgreSQL) Store
    • Pinecone Store
    • Qdrant Store
    • Redis Store
    • Weaviate

• Testing and Evaluation - Reuse, to be extended

• Observability - Reuse

• WebSocket integration - Reuse

• Model Context Protocol (MCP) - To be revamped

• Security & Privacy Considerations - Reuse

• Prompt injection protection - Todo

• Dev UI - Reuse

• AI Services patterns - new, todo

• Semantic Compression - new, todo

Development

• [https://github.com/quarkiverse/quarkus-langchain4jSource Code^,role=github]
• https://github.com/quarkiverse/quarkus-langchain4j/issues[Issues^,role=github]