20 Repos
Workflows for preparing and generating data for retrieval-augmented generation.
Distinguishing note: Focuses on the pipeline aspect of RAG data preparation.
Explore 20 awesome GitHub repositories matching artificial intelligence & ml · RAG Data Pipelines. Refine with filters or upvote what's useful.
This project is a comprehensive educational resource and technical guide focused on the development, optimization, and application of large language models. It provides a structured curriculum for mastering prompt engineering, ranging from foundational principles of instruction design to advanced techniques for improving model reasoning, accuracy, and reliability. The guide distinguishes itself by offering deep technical insights into agentic workflows and autonomous system design. It covers the implementation of multi-step reasoning chains, tool integration through function calling, and stat
Addresses the common challenge of data scarcity in RAG projects through synthetic generation.
Docling is a multimodal content converter and document parser designed to transform PDFs, Office files, and HTML into structured Markdown or JSON for generative AI applications. It functions as an OCR document processor and a PDF layout analyzer that extracts tables, charts, and hierarchical structures while preserving the original page layout. The system operates as a local-first inference engine, allowing for the processing of sensitive data in air-gapped environments without external network connectivity. It can also be deployed as an API or a Model Context Protocol server to provide parsi
Converts diverse document formats into machine-readable text to power RAG pipelines.
Quiver is a framework for integrating retrieval augmented generation into applications. It provides a generative AI integration layer that connects large language models with vector stores to produce context-aware responses based on custom data. The project features a knowledge base pipeline that parses diverse file types into searchable embeddings and a vector database orchestrator to manage data across different storage implementations. It utilizes a provider-agnostic model interface, allowing users to switch between various external AI providers or local models through a single unified sys
Orchestrates the sequential data flow from document ingestion and embedding through to final response generation.
Marker is an LLM-powered document parser and OCR pipeline designed to convert PDFs and unstructured files into structured markdown, JSON, and HTML. It functions as a data preprocessor that transforms complex documents into machine-readable formats while preserving tables, equations, and layout structures. The system utilizes large language models to refine OCR accuracy, clean mathematical notation, and merge fragmented tables across multiple pages. It employs model-based layout analysis to predict block types and bounding boxes, ensuring a more precise conversion of document elements. Capabi
Prepares complex documents by converting them into structured chunks and HTML for use in RAG systems.
Bisheng is an enterprise AI framework and LLM DevOps platform designed to manage the full lifecycle of large language models. It provides a unified system for dataset curation, supervised fine-tuning, model versioning, and performance evaluation. The platform features a visual workflow orchestrator for building retrieval-augmented generation pipelines and complex task sequences using flowcharts with conditional logic and human intervention points. It also includes an AI agent framework that uses a specialized guidance language to embed domain expertise and professional business logic into aut
Combines document parsing, dataset management, and retrieval mechanisms to provide context for generative AI tasks.
A fast, helpful, and open-source document parser
Reconstructs headings, tables, lists, images, and links from spatial layout for direct use in LLMs and RAG pipelines.
Skill Seekers is a toolset for generating large language model knowledge bases, featuring a multi-source content scraper and a dedicated RAG data pipeline. It extracts technical data from documentation, code, and video to create structured assets and configuration files for AI-powered IDE extensions. The project distinguishes itself through the ability to transform raw data into polished tutorials and specialized skills for AI plugin marketplaces. It utilizes abstract syntax tree parsing and optical character recognition to analyze GitHub repositories, PDFs, and video frames, converting these
Implements workflows for preparing and chunking technical data to optimize retrieval-augmented generation accuracy.
Megaparse is a document parsing tool and RAG data preprocessor designed to convert PDFs, Word documents, and presentations into clean text formats. It functions as a vision-based document extractor that recovers high-fidelity information from images and complex layouts to optimize data for large language model ingestion. The system employs multimodal AI and vision models to perform schema-preserving parsing, which maintains structural hierarchies such as tables and headers. It utilizes lossless structural transformation to turn layout-heavy binary files into text sequences while preserving th
Implements a preprocessing pipeline that cleans and formats unstructured files to optimize retrieval-augmented generation.
This project is an educational implementation guide and framework for building Retrieval Augmented Generation systems. It provides a workflow for constructing a knowledge base pipeline that partitions documents, indexes them as vectors, and provides external context for language model prompts. The system features a document chunking framework that uses recursive character splitting to fit text into model context windows. It includes an in-memory vector store and a similarity search system that retrieves relevant text segments by calculating the mathematical distance between dense embedding ve
Designs the end-to-end workflow from document partitioning and embedding to context retrieval.
This project is an educational course and technical blueprint for building production-ready retrieval-augmented generation systems. It provides a curriculum and implementation strategies for designing agentic workflows, containerized AI infrastructure, and retrieval pipelines using large language models. The materials focus on agentic design patterns, utilizing state-based decision nodes to rewrite queries and grade retrieved documents. It differentiates its approach by providing a deployment framework for managing databases, search engines, and API services through container orchestration.
Provides a practical blueprint for constructing data pipelines that process PDFs and store embeddings for semantic search.
Data-Juicer is an open-source framework for cleaning, filtering, deduplicating, and transforming multimodal datasets to prepare them for training large language and vision models. It functions as a distributed data pipeline engine that runs processing jobs across Ray clusters, handling billions of samples with automatic operator fusion and adaptive parallelism. The framework provides a library of operators that leverage large language models for semantic extraction, filtering, and data synthesis within processing pipelines. The project distinguishes itself through a YAML-based data recipe sys
Ships operators for extracting, normalizing, chunking, and deduplicating content for RAG indexes.
Connects data ingestion and embedding pipelines to search or vector services for retrieval-augmented generation.
Kiln ist eine LLM-Entwicklungs-Workbench und ein Evaluierungs-Framework, das für das Design, das Testen und die Optimierung von Prompts und KI-Agenten entwickelt wurde. Es fungiert als Multi-Agenten-Orchestrator und RAG-Optimierungstool und bietet eine visuelle Schnittstelle für die iterative Entwicklung von KI-Systemen. Das Projekt zeichnet sich durch eine umfassende Fine-Tuning-Pipeline aus, die Zero-Code-Modelltraining und Reasoning-Destillation unterstützt. Es ermöglicht die Erstellung hierarchischer Multi-Agenten-Systeme, in denen spezialisierte Akteure über Tool-Calling koordinieren, und implementiert einen Model Context Protocol-Server, um diese Agenten und Suchfunktionen als standardisierte Tools für externe Clients bereitzustellen. Die Plattform deckt ein breites Spektrum an Fähigkeiten ab, einschließlich automatisierter KI-Jury-Bewertung zur Qualitätssicherung, synthetischer Datengenerierung für Training und Evaluierung sowie hybrider Vektor-Keyword-Suche zur Erdung von Modellantworten. Sie bietet zudem Tools für Prompt-Evolution, Trace-Auditing und die Verwaltung kollaborativer Datensätze via Git-Integration. Die Workbench ist über eine selbst-hostbare REST-API und eine dedizierte Python-Bibliothek für die programmatische Workflow-Ausführung zugänglich.
Provides a visual workbench for building and refining RAG retrieval pipelines to improve model grounding.
Rudder Server ist eine Customer Data Platform (CDP) und Event-Routing-Pipeline, die darauf ausgelegt ist, Kundendaten zu sammeln, zu transformieren und von verschiedenen Quellen an Data Warehouses und Business-Tools weiterzuleiten. Es fungiert als Customer-Identity-Resolver, der Identifikatoren aus mehreren Quellen verknüpft, um einen einheitlichen Identitätsgraphen und umfassende verhaltensbasierte Kundenprofile zu erstellen. Das System zeichnet sich durch Reverse-ETL-Funktionen aus, die verarbeitete Kundensegmente und Zielgruppen aus Data Warehouses zurück in operative Drittanbieteranwendungen pushen. Es bietet zudem eine containerisierte Datenebene für Kubernetes-Deployments, was die Verwaltung der Dateninfrastruktur als Code ermöglicht. Die Plattform deckt eine breite Palette von Datenmanagement-Funktionen ab, einschließlich Echtzeit-Event-Transformation, Schema-Validierung via Datenkatalogen und Privacy-Governance. Dazu gehören Tools zur Verwaltung der Benutzereinwilligung, zur Durchsetzung der Datenresidenz innerhalb spezifischer geografischer Regionen und zur Maskierung personenbezogener Daten während der Übertragung. Installation und Deployment der Datenebenen-Komponenten werden mittels Helm-Charts verwaltet.
Sends notifications when critical failures or anomalies occur within active data pipelines.
Chonkie ist eine Text-Chunking-Bibliothek, die für Retrieval-Augmented-Generation-Pipelines (RAG) konzipiert wurde. Sie fungiert als semantischer Text-Splitter und RAG-Ingestion-Pipeline und transformiert Rohtext in eingebettete Segmente für die Speicherung in Vektordatenbanken. Das Projekt zeichnet sich durch spezialisierte Splitting-Strategien aus, einschließlich eines AST-basierten Code-Splitters zur Bewahrung logischer Grenzen im Quellcode und eines semantischen Text-Splitters, der Embedding-Modelle verwendet, um Grenzen basierend auf der Bedeutung zu bestimmen. Es bietet zudem einen Vektordatenbank-Ingestor, um die Generierung von Embeddings und deren Export in verschiedene Speicher zu automatisieren. Die Bibliothek deckt ein breites Spektrum an Funktionen ab, einschließlich Dokumenten-Parsing via OCR und Markdown-Extraktion, einer Vielzahl von Splitting-Methoden wie Token-Count und hierarchische Segmentierung sowie Workflow-Orchestrierung durch wiederverwendbare Pipelines. Sie unterstützt eine breite Palette an Vektorspeicher-Integrationen, einschließlich Qdrant, Milvus, Weaviate und Elasticsearch, sowie den Datenexport in JSON- und Hugging-Face-Datensätze. Nutzer können diese Operationen über eine Kommandozeilenschnittstelle ausführen oder das System als containerisierten API-Dienst bereitstellen.
Implements workflows for preparing, cleaning, splitting, and embedding unstructured text for retrieval-augmented generation.
Uptrace is an OpenTelemetry-based observability platform designed to collect, store, and analyze distributed traces, metrics, and logs. It functions as a centralized logging backend, a distributed tracing system, and a metrics engine to monitor application performance and system health. The platform is distinguished by AI-powered operational capabilities, allowing users to query telemetry data and manage monitoring dashboards using natural language. It specifically includes specialized monitoring for generative AI pipelines, tracking token usage and response quality for LLM interactions and r
Triggers notifications for empty retrieval results or latency spikes within AI pipelines.
Dieses Projekt bietet eine Reihe struktureller Templates und Frameworks zum Bootstrapping von Produktionsservern, hochperformanten Backends, Kubernetes-Microservices und KI-Pipelines unter Verwendung der Programmiersprache Go. Es dient als grundlegende Architektur für den Aufbau von Infrastruktur mit hohem Durchsatz und skalierbaren Produktionsservern mit integriertem Routing und Middleware. Das Framework enthält eine spezialisierte Infrastruktur zur Entwicklung von Retrieval-Augmented-Generation-Systemen, wobei der Schwerpunkt auf lokaler Modell-Inferenz und sicherer Datensouveränität liegt. Es bietet zudem ein dediziertes Microservice-Template für containerisierte Deployments mit Fokus auf Ressourcenquoten und Service-Lifecycle-Management. Das Projekt deckt breite Funktionsbereiche ab, einschließlich Observability-Instrumentierung zur Überwachung der Systemgesundheit, Interface-basierter Dependency-Injection und Migrations-basierter Schemaverwaltung für relationale Daten. Es integriert zudem Channel-basierte Concurrency zur Verwaltung asynchroner Aufgaben und Ressourcenoptimierung durch CPU- und Memory-Profiling.
Implements RAG data pipelines that ground AI responses in verified internal data with secure sovereignty controls.
Paper2Slides is an AI-driven presentation generator and content extractor designed to transform academic papers and scientific documents into structured slides and posters. It utilizes retrieval-augmented generation to distill key data points and identify critical figures while maintaining direct traceability to the original source text. The system functions as an AI slide designer that applies professional themes or custom visual styles defined through natural language. It integrates with external image generation services to produce high-quality visuals and research visualizations for acade
Implements a pipeline that identifies critical data and figures from source documents with direct traceability for verification.
Data engineering practice repository providing tutorials, distributed processing engines, and Python data pipeline automation scripts. The system encompasses automated data validation, distributed compute aggregation, embedded columnar querying, lazy evaluation planning, partitioned storage export, and cloud storage retrieval. The capability surface covers cloud integration and storage, data engineering and pipelines, data processing and analytics, data quality and testing, database and storage, file management, and monitoring and observability.
Triggers notifications when incoming records violate expected business rules or exhibit extreme outliers.
Ramalama is a containerized runtime and management tool for large language models. It functions as an OCI AI model manager and registry client, allowing users to package, distribute, and execute AI models as standardized container images. The project differentiates itself by using OCI-compliant distribution for models and retrieval augmented generation assets, enabling the packaging of vector databases into immutable container images. It features hardware-aware image selection that automatically detects GPU or CPU capabilities to pull the most optimized image for the host environment. The sy
Implements pipelines for processing documents into vector databases and packaging them as container images.