docker/genai-stack
Genai Stack
Ce projet est une stack de développement conteneurisée et un framework d'application pour construire des systèmes de génération augmentée par récupération (RAG). Il fournit un sandbox IA dockerisé qui intègre des runtimes de modèles locaux, des graphes de connaissances et des vector stores pour permettre la création de chatbots contextuels.
La stack se distingue par son vector store basé sur les graphes, qui combine des graphes de connaissances structurés avec des index vectoriels pour la récupération de données sémantiques et structurelles. Elle permet l'hébergement de modèles locaux avec accélération CPU ou GPU, permettant des tâches génératives sans dépendance aux API cloud externes.
Le framework couvre un large éventail de capacités, notamment le traitement et l'indexation de documents PDF, l'orchestration de services IA basés sur des conteneurs et l'implémentation d'une génération de réponses ancrées (grounded). Il inclut une interface de chat web avec streaming de réponse incrémentiel et une interface standardisée pour basculer entre différents fournisseurs de modèles de langage.
L'environnement est amorcé via l'orchestration de conteneurs pour déployer rapidement une stack préconfigurée de modèles et de bases de données.
Features
- LLM Application Frameworks - Provides a comprehensive toolkit for integrating large language models with vector and graph databases to build contextual chatbots.
- Hybrid Vector-Graph Databases - Combines knowledge graphs with vector indices in a single database for semantic and structural data retrieval.
- Container Orchestration Environments - Orchestrates a containerized stack of models, databases, and interfaces for rapid AI environment bootstrapping.
- Vector Knowledge Bases - Uses Neo4j to create a data layer combining structured knowledge graphs and vector embeddings.
- Standard RAG Development - Builds systems that retrieve context from vector and graph databases to provide grounded answers.
- Graph Retrieval Augmented Generation - Combines graph-based knowledge structures with retrieval-augmented generation to provide contextually grounded AI responses.
- LLM Provider Integrations - Implements a standardized interface for connecting to various local and cloud large language model providers.
- LLM Provider Interfaces - Uses a common abstraction layer to seamlessly switch between local model runtimes and cloud AI services.
- Local Model Runtimes - Includes a deployment setup for running open-source language models on local hardware with CPU or GPU acceleration.
- Model Provider Configurations - Provides standardized connection settings and credential management for integrating diverse AI model providers.
- RAG Frameworks - Offers a containerized development environment combining LangChain, Neo4j, and Ollama for building RAG applications.
- Retrieval Augmented Generation Pipelines - Implements a pipeline that fetches relevant document chunks and graph nodes to ground AI responses.
- Text Embedding Generators - Converts text data into numerical vector representations to enable semantic search and similarity analysis.
- Vector Embeddings - Implements a workflow that processes raw PDF text into vector embeddings for semantic retrieval.
- Document Chunking and Embedding Pipelines - Provides an automated pipeline to chunk, embed, and load external documents into a vector-graph database.
- External Data Ingestion - Ingests questions and answers from external APIs, generating embeddings for storage in a graph database.
- Graph Databases - Provides connectivity to a graph database for the storage and retrieval of structured knowledge.
- Graph RAG Frameworks - Implements retrieval-augmented generation using a hybrid of vector and graph-based data structures for grounded answering.
- AI Infrastructure Stacks - Sets up a pre-configured stack of model runtimes and databases using Docker for rapid AI prototyping.
- Pre-Configured Sandbox Deployments - Provides a pre-configured set of containers for deploying local model runtimes and knowledge graphs.
- LLM Deployments - Runs language models and embedding providers in containers to enable offline generative tasks.
- Full-Stack Orchestration - Orchestrates a full stack of AI models and databases using containers for rapid development.
- Local Language Model Hosting - Executes language models in containers with CPU or GPU acceleration for local generative tasks.
- AI Project Bootstrapping - Provides tools for initializing a pre-configured environment with a model runner and graph database.
- Chatbot Containerization - Packages the AI application and its web interface into containers for consistent execution.
- Response Grounding - Produces grounded responses with source citations by constraining model outputs to verified facts within a knowledge base.
- Local Document Indexing - Processes local PDF files into searchable vector stores to enable similarity-based questioning within RAG pipelines.
- Chat Interfaces - Renders a web-based conversational interface for interacting with the integrated AI models.
- LLM Response Streaming - Delivers generated text tokens incrementally from the model to the UI to minimize perceived latency.
- PDF Chatbots - Processes PDF files into text chunks and embeddings for natural language querying.
- PDF Content Querying - Extracts text from PDF documents and indexes it in a vector database to enable natural language querying.
- PDF Document Analyzers - Combines text extraction and semantic search to analyze and query information within local PDF documents.
- Knowledge Graph Builders - Extracts external data and processes it into embedded vectors and graph nodes for retrieval.
- Vector Indexing - Creates specialized vector indexes within a graph database to enable high-dimensional similarity searches for questions and answers.
- AI Interface Deployments - Bootstraps a containerized user interface to interact with integrated AI services.
- PDF Chatbot Deployments - Containerizes a generative AI application to provide an interface for interacting with PDF documents.
- AI Stack Deployments - Wraps AI functionality with a server to serve generative capabilities over HTTP.
- Incremental Response Streaming - Streams tokens incrementally to the user interface to reduce perceived latency during response generation.
