Packages

Distributed SQL query engine for querying data across multiple sources (S3/HDFS Parquet/ORC, Hive, Iceberg, Delta Lake, PostgreSQL, MySQL, Kafka, Redis, MongoDB, and more) using standard ANSI SQL. Trino (formerly PrestoSQL) runs federated queries spanning multiple data sources in a single SQL statement — JOIN a Hive table with PostgreSQL with a Kafka topic. Excels at interactive analytics on large datasets. Widely used as the query engine behind data platforms at Meta, LinkedIn, Netflix, and others.

An MCP server that enables AI agents to publish content, search posts, retrieve recommendations, manage comments, and access user profiles on Xiaohongshu (小红书/RedNote) via headless browser automation.

Zen is a self-hosted, single-binary notes app that stores markdown notes in SQLite with tag-based organization, BM25 full-text search, and an optional MCP server for AI agent integration with note operations.

Python bindings for llama.cpp — runs quantized LLMs locally on CPU/GPU without PyTorch or CUDA required. llama-cpp-python features: Llama class (load GGUF model, generate text), OpenAI-compatible API server (llama_cpp.server), llama(prompt) call interface, n_gpu_layers for partial/full GPU offload, chat completion format (llm.create_chat_completion), streaming responses, function calling support, embedding generation (llm.create_embedding), context window control, and mmap for large model loading. Runs Llama, Mistral, Phi, Gemma, and other GGUF-format models. Primary tool for CPU-only LLM inference in agent pipelines.

Python refactoring library — provides programmatic code refactoring operations on Python projects. rope features: rename (variables, functions, classes, modules), extract function/variable, move (class/function to different module), inline variable, introduce parameter, change function signature, auto-import suggestions, code completion via analysis, find-occurrences for symbol lookup, organize imports, undo/redo for refactoring operations, and Project abstraction for multi-file refactoring. Used by IDE plugins (Vim, Emacs, Sublime) for Python refactoring support.

Built-in WebSocket framework for Ruby on Rails — integrates WebSockets with Rails app for real-time features using channels. Action Cable features: Channel classes (ApplicationCable::Channel), subscriptions via JavaScript consumer, broadcasting with ActionCable.server.broadcast, stream_from / stream_for for pub/sub, authentication via connection.rb (current_user), Redis or async adapter for pub/sub backend, cable.yml configuration, Action Cable JavaScript consumer (consumer.subscriptions.create), and full Rails session/cookie access in connection. Replaced Need for Pusher in many Rails apps. ChatChannel, NotificationChannel, AgentStatusChannel patterns enable real-time agent updates without external WebSocket service.

Python-native stateful stream processing framework built on a Rust runtime. Bytewax brings the developer experience of Python to stream processing — write dataflows as Python functions without JVM, Java/Scala knowledge, or complex cluster configuration. Processes data from Kafka, Kinesis, files, or custom sources with stateful windowing, joining, and aggregation. Runs locally for development and scales to distributed execution. Positioned as the 'Python Flink' for teams that want streaming without the JVM ecosystem.

Kubernetes operator that syncs secrets from external secret management systems (AWS Secrets Manager, HashiCorp Vault, Azure Key Vault, GCP Secret Manager, Doppler, Infisical, etc.) into Kubernetes Secrets. Enables GitOps workflows where secret references (not values) are stored in Git, while actual values remain in secure external stores. ExternalSecret resources define what to fetch and how to map to Kubernetes Secrets — operators pull values periodically and keep Kubernetes Secrets in sync.

Finastra FusionFabric.cloud REST API marketplace for banking and financial services platform. Enables AI agents to manage corporate banking and trade finance workflow automation, handle retail banking account and transaction data access, access FusionBanker and FusionInvestor core banking integrations, retrieve payments processing and SWIFT messaging data, manage lending origination and servicing lifecycle automation, handle treasury and capital markets workflow data, access compliance and regulatory reporting data, retrieve open banking and PSD2-compliant data, manage FusionCredit and FusionLend lending automation, and integrate Finastra banking applications with fintech ecosystem and third-party services.

Framework-agnostic GraphQL server built on the Fetch API that runs on Node.js, Cloudflare Workers, Deno, and Bun with a plugin system powered by Envelop.

Python framework for building production-grade AI agents and pipelines. Griptape provides agents, tasks, pipelines, and workflows as composable primitives. Core abstractions: Agents (single-task LLM loops with tools), Pipelines (sequential task chains), and Workflows (parallel DAGs). Includes built-in memory, RAG toolchain, task memory (off-prompt data handling), and cloud platform (Griptape Cloud) for deployment. Enterprise-focused alternative to LangChain with emphasis on predictability and off-prompt data handling.

HackerOne bug bounty and vulnerability disclosure platform REST API for enterprises to automate security report management, researcher bounty payments, program configuration, and vulnerability data integration — enabling AI agents to retrieve reports, update triage states, manage bounty awards, and integrate HackerOne security findings with ITSM and DevSecOps platforms through HackerOne's market-leading crowdsourced security platform. Enables AI agents to manage report management for vulnerability report retrieval, filtering, and triage status automation, handle bounty management for researcher bounty award and payment processing automation, access program management for bug bounty program scope and configuration automation, retrieve hacker management for security researcher reputation and profile retrieval automation, manage activity management for report activity thread and analyst communication automation, handle asset management for in-scope target and asset inventory automation, access integration management for JIRA, ServiceNow, and GitHub vulnerability tracking automation, retrieve analytics for program performance and vulnerability trend analytics automation, manage inbox management for report triage workflow and assignment automation, and integrate HackerOne with DevSecOps pipelines, ITSM, and vulnerability management for crowdsourced security program automation.

Haystack is an open-source Python framework by deepset for building RAG and NLP pipelines; it offers a composable component model for connecting LLMs, document stores, retrievers, and preprocessors, with an optional Hayhooks REST server for HTTP-based pipeline invocation.

MCP server enabling AI agents to interact with LinkedIn — searching profiles, reading job postings, accessing company information, and retrieving professional network data through MCP tool calls. Supports recruitment workflows, job search automation, and competitive intelligence gathering from LinkedIn's professional network.

In-memory graph database optimized for real-time analytics and streaming graph workloads. Memgraph uses Cypher (same as Neo4j), supports the Bolt protocol, and integrates with Kafka/Pulsar for streaming graph updates. MAGE (Memgraph Advanced Graph Extensions) provides graph algorithms and machine learning on graphs. Memgraph Cloud is the managed option. Strong focus on real-time graph analytics.

Open standard for data lineage collection and propagation. OpenLineage defines a spec for how data pipelines emit lineage events (who ran what job, which datasets were inputs, which were outputs). Compatible backends: Marquez (open-source server), DataHub, Atlan. Integrations for Spark, Airflow, dbt, Flink, and more auto-emit lineage events without code changes.

Recruitee REST API for collaborative applicant tracking system. Enables AI agents to query job openings, retrieve candidate profiles and applications, access pipeline stage data, and integrate recruiting intelligence into HR workflows. Recruitee is known for its team-based collaborative hiring approach and modern API design.

In-memory vector and full-text search via the RediSearch module, enabling sub-millisecond semantic similarity search and real-time keyword search over Redis data.

Kafka-compatible streaming platform written in C++ that delivers 10x lower latency than Kafka with no JVM dependency, available as self-hosted or managed cloud.

MCP server for Toutiao — ByteDance's flagship Chinese content platform and news aggregator with over 300M active users. Enables AI agents to search Toutiao articles, retrieve trending content, access news feeds, and monitor Chinese digital media — supporting content research and social listening for the Chinese market.