Projects sharing features with Bert As Service

AdalFlow is an autonomous AI agent framework and LLM application library designed for building modular workflows. It serves as a model-agnostic interface and RAG pipeline orchestrator, allowing users to develop ReAct agents that utilize iterative reasoning and external tool execution to solve complex tasks. The project distinguishes itself through a prompt optimization system that uses textual gradient descent to automatically refine prompt templates and few-shot examples. It treats model feedback as a differentiable signal, enabling a form of LLM backpropagation to iteratively improve output

This project is a PyTorch model serving framework designed to deploy and scale machine learning models in production via scalable network endpoints. It functions as a high-performance inference server, optimizer, and model lifecycle manager that handles model loading, request batching, and hardware acceleration. The system distinguishes itself through advanced orchestration and optimization capabilities, such as chaining multiple models into sequential workflows using execution graphs and employing dynamic batching to improve throughput and latency. It provides specialized support for generat

Text Embeddings Inference is a high-performance inference server designed to host text embedding and sequence classification models as scalable API endpoints. It provides a vector embedding API to convert text into dense representations and a cross-encoder reranking server for scoring the relevance of document sequences against a query. The project features a GPU-accelerated inference engine that utilizes dynamic batching and specialized kernels to maximize throughput. It offers a high-performance binary interface via gRPC as an alternative to standard HTTP to reduce network latency and seria

This project is a reference implementation and application template for Retrieval-Augmented Generation (RAG). It integrates Azure OpenAI with Azure AI Search to enable conversational chat interfaces that provide grounded responses based on private enterprise data. The system is distinguished by its multimodal AI interface, allowing it to process and reason over combined text, image, and PDF content. It employs a hybrid search architecture that combines vector and keyword retrieval with semantic reranking to prioritize the most relevant documents for prompt augmentation. The project covers a

pgai is a PostgreSQL AI toolkit and framework designed to integrate large language models and vector embeddings directly into a database. It serves as a bridge for executing machine learning model requests and performing text-to-SQL translations within standard database queries. The project provides an automated vector embedding pipeline that handles the loading, parsing, and chunking of text from tables and unstructured documents. This system utilizes a background worker to synchronize embeddings automatically as source data changes and includes specialized tools for building retrieval-augme

Chonkie is a text chunking library designed for retrieval-augmented generation pipelines. It functions as a semantic text splitter and RAG ingestion pipeline, transforming raw text into embedded segments for storage in vector databases. The project distinguishes itself through specialized splitting strategies, including an AST-based code splitter for preserving logical boundaries in source code and a semantic text splitter that uses embedding models to determine boundaries based on meaning. It also provides a vector database ingestor to automate the generation of embeddings and their export t

Marqo is an ecommerce product discovery platform, multimodal vector database, and AI search merchandising tool. It provides infrastructure for implementing semantic search and recommendations, allowing shoppers to find products using natural language and images. The platform distinguishes itself through a hybrid ranking pipeline that combines neural semantic scores with business-defined boosting and pinning rules. It features a conversational commerce engine that uses large language models to process user intent and provides a search performance analytics suite for measuring conversion uplift

Starspace is a vector embedding framework designed for training high-dimensional representations of text and images. It functions as a machine learning system for neural ranking, text classification, and knowledge graph embedding, mapping different object types into a shared numerical space to facilitate retrieval and prediction tasks. The system includes specialized tools for knowledge graph completion and link prediction by representing entities and their relationships within a multi-relational vector space. It further provides capabilities for semantic content recommendation and large-scal

This repository is a collection of guides, notebooks, and recipes for implementing advanced prompting techniques and workflow patterns with large language models. It serves as a prompt engineering guide, an evaluation suite for scoring prompt quality, and a framework for orchestrating agents and integrating external tools. The project provides implementation patterns for building applications with Claude, specifically focusing on coordinating multiple models to split complex tasks between high-reasoning and high-efficiency agents. It includes technical demonstrations for multimodal data proce

Koog is an LLM agent framework used to build autonomous entities that execute tool-based workflows. It utilizes a graph-based workflow engine to define agent behaviors and decision paths as a directed graph of nodes and edges. The framework distinguishes itself through a model provider orchestrator that enables dynamic switching, load balancing, and automatic fallbacks between different AI backends. It implements the Model Context Protocol to connect agents to remote tool servers and features a RAG memory system using vector embeddings to maintain long-term conversation context. The project

jetson-inference is a set of libraries and tools for executing optimized deep learning models on embedded GPU hardware. Its primary purpose is to enable real-time computer vision and AI inference at the edge with low latency and high throughput. The project distinguishes itself through high-performance streaming analytics and the ability to execute concurrent AI pipelines on auto-grade silicon. It provides specialized support for multi-sensor stream processing, utilizing zero-copy data transport to load camera frames directly into GPU memory. The codebase covers a broad surface of capabiliti

This project is a comprehensive educational resource and curriculum focused on the design and implementation of the full machine learning software and hardware stack. It serves as a technical reference for architecting machine learning systems, spanning from low-level programming interfaces to large-scale deployment infrastructure. The project provides instructional guidance on several specialized domains, including the development of AI compilers through intermediate representations and graph optimizations. It covers the architectural patterns required for distributed training across GPU clu

Lorax is a GPU-accelerated inference server and multi-adapter engine designed for serving large language models. It functions as a high-throughput system capable of deploying models via Kubernetes and managing the dynamic swapping of Low-Rank Adaptation adapters per request. The server distinguishes itself through multi-adapter dynamic batching, which allows requests using different adapter weights to be processed in a single GPU forward pass. It employs just-in-time adapter loading and weighted adapter merging to maximize throughput and enable multi-tasking without sacrificing performance.

LanceDB is a vector database and columnar data store designed to function as a versioned dataset manager and vector search engine. It serves as a high-performance backend for indexing and retrieving high-dimensional embeddings, providing the foundation for machine learning data pipelines. The system distinguishes itself through a combination of cloud-native object storage and immutable version tracking, allowing for data time-travel and reproducible AI experiments. It integrates hybrid search capabilities, merging dense vector similarity with BM25 full-text search and SQL-like scalar filters

Reader is an AI data ingestion pipeline and web content parser designed to convert websites and documents into clean markdown for use with large language models. It functions as a headless browser content extractor and web-to-markdown converter, transforming URLs and PDF files into structured text formats while removing irrelevant web clutter. The system optimizes retrieval augmented generation by acting as a search optimizer that retrieves web results and applies re-ranking to improve context relevance. It further enhances content accessibility by using vision models to generate descriptive

This project is a containerized development stack and application framework for building retrieval-augmented generation systems. It provides a dockerized AI sandbox that integrates local model runtimes, knowledge graphs, and vector stores to enable the creation of contextual chatbots. The stack is distinguished by its graph-based vector store, which combines structured knowledge graphs with vector indices for both semantic and structural data retrieval. It allows for local model hosting with CPU or GPU acceleration, enabling generative tasks without reliance on external cloud APIs. The frame

LitServe is a Python AI inference server framework and LLM serving framework designed for high-concurrency inference. It functions as a distributed AI model server and dynamic batching inference engine, providing the tools to build and host custom servers that run AI models. The framework distinguishes itself through a dynamic-batching request queue that groups individual inference requests into single tensors to maximize GPU throughput. It supports distributed GPU scaling, allowing model workloads to be spread across multiple hardware accelerators to balance compute loads and increase total

ollama-python is a Python client for interacting with large language models. It provides an interface for sending prompts to receive text and chat completions, as well as a dedicated client for generating numerical vector embeddings from text. The project includes a wrapper that emulates the OpenAI API, allowing applications built for that standard to interact with local models. It also provides a non-blocking asynchronous client for executing concurrent requests. The library covers the full model lifecycle, including the ability to pull, create, list, and delete models within a local enviro

Casibase is an open-source platform that orchestrates multi-turn conversations with large language models and manages retrieval-augmented knowledge bases from a single interface. It provides a unified system for connecting to over 30 AI model providers, ingesting documents into vector embeddings for semantic search, and running autonomous agent loops that can drive a browser, search the web, execute commands, and integrate with external tools. The platform distinguishes itself by combining AI conversation management with infrastructure and application orchestration capabilities. It includes a

KServe is a Kubernetes-native platform for deploying and serving machine learning models as scalable inference services. It supports both generative AI models, including large language models, and traditional predictive models from frameworks such as TensorFlow, PyTorch, Scikit-Learn, XGBoost, and ONNX. The platform manages the full lifecycle of model deployments, including revision tracking, canary rollouts, A/B testing, and automatic rollbacks, and provides serverless scale-to-zero capabilities for cost-efficient resource management. KServe distinguishes itself through a standardized infere

KServe is an open platform for deploying and serving generative and predictive AI models on Kubernetes. It defines inference services as custom resources with declarative YAML specifications, enabling a Kubernetes-native approach to model deployment and lifecycle management. The platform leverages Knative-based serverless scaling for automatic scale-to-zero and revision management, and supports a pluggable serving runtime architecture that maps model formats to containerized execution environments. KServe distinguishes itself through model-aware autoscaling that scales replicas based on token

This project is a Java microservices reference implementation that demonstrates the application of structural design patterns for distributed systems. It consists of a suite of containerized services and infrastructure dependencies designed to showcase how to coordinate multiple independent services into a functional system. The implementation focuses on several advanced architectural patterns, including a centralized API gateway for routing and aggregating responses, Command Query Responsibility Segregation to separate read and write operations, and event sourcing to persist the state of bus

Seldon Core is a Kubernetes-based machine learning model server and MLOps inference framework. It functions as a multi-model serving engine and pipeline orchestrator, packaging models as scalable microservices that are exposed via standardized REST and gRPC APIs. The project distinguishes itself through graph-based inference pipelines that chain models and data transformers into sequential workflows. It optimizes hardware utilization via multi-model shared serving and dynamic memory overcommit strategies, while supporting production experimentation through weighted traffic routing, A/B testin

Superlinked is a development framework designed for building semantic search and retrieval pipelines. It functions as a machine learning data pipeline and semantic retrieval engine, providing the tools necessary to unify data schema definition, embedding generation, and vector database integration within a single application. The framework distinguishes itself by acting as a vector database orchestrator that manages the lifecycle of machine learning models alongside complex search logic. It enables developers to construct structured data models that map raw content and metadata into unified r

Nameko is a Python microservices framework designed for building distributed systems using a service-oriented architecture. It provides a system for implementing remote procedure calls and asynchronous event broadcasting between decoupled services by utilizing the AMQP protocol and message brokers. The framework includes a managed HTTP client pool to fetch remote content while preventing resource exhaustion. It also functions as a microservice web gateway, allowing internal service logic to be exposed as external web endpoints via standard HTTP protocols. The project covers broad capabilitie

FlexLLMGen is an inference engine and runtime designed to run large language models on a single GPU by combining weight compression with tensor offloading. It reduces model weight memory usage by approximately 70% through 4-bit quantization, and stores model parameters, attention cache, and hidden states across GPU, CPU, and disk to fit models larger than available GPU memory. The project distinguishes itself through a throughput-oriented batching approach that processes multiple generation requests together in large batches to maximize throughput on a single GPU. It also supports distributed

ollama-js is a JavaScript client library and API wrapper that provides a programmatic interface for interacting with and managing large language models. It enables the execution of models in both local and cloud-based environments, facilitating the generation of conversational text and the management of model lifecycles. The project distinguishes itself by offering specialized tools for model administration, including the ability to download, create, and remove models, as well as the capacity to define custom model blueprints and prompt templates. It also provides a vector embedding client fo