Packages

Groq provides ultra-low-latency LLM inference using proprietary Language Processing Units (LPUs), delivering 200-500 tokens/second on models like Llama 3.1 70B and Mixtral 8x7B via an OpenAI-compatible REST API.

Android's recommended dependency injection library, built on top of Dagger 2. Hilt simplifies Dagger setup for Android by providing standard components for Activity, Fragment, ViewModel, Service, and WorkManager. Eliminates the boilerplate of manual Dagger component creation while retaining compile-time DI code generation. The Google-endorsed replacement for manual DI or Service Locator patterns in Android.

Kafka-native Java library for stateful stream processing that runs as a lightweight embedded library without a separate cluster.

Pentest MCP server enabling AI agents to perform penetration testing and security assessment tasks — running reconnaissance tools, network scanning with nmap, subdomain enumeration, web vulnerability scanning, and integrating common pentesting workflows into agent-driven authorized security assessment pipelines.

Planning Center REST API for church management platform (ChMS). Enables AI agents to manage church member and household records, track giving and pledge data, manage service and worship planning, handle volunteer scheduling, process event registrations, sync ministry data across Planning Center products (People, Giving, Services, Groups, Registrations), and integrate church data with external systems.

Fastest key-value storage for React Native, based on Tencent's MMKV (memory-mapped key-value) library. 10x faster than AsyncStorage for synchronous reads/writes. MMKV uses memory-mapped files — reads are direct memory access without I/O overhead. Supports string, number, boolean, and buffer storage with encryption. Used by Discord, Shopify, and others needing high-performance local storage.

Compile-time dependency injection framework for Go. Wire generates dependency injection code at build time using provider functions — you define what each component needs (database, config, logger) and Wire generates the wiring code. Unlike runtime DI containers (Java Spring, Python DI), Wire produces plain Go code with zero runtime overhead and full type safety. Standard Google approach for DI in large Go services.

Minimalist single-file Python web framework — provides routing, request/response handling, templating, and HTTP server in a single .py file with no dependencies. bottle features: @route/@get/@post/@put/@delete decorators, URL parameters with <name> syntax, request object (request.json, request.forms, request.params), response object (response.status, response.headers), built-in template engine (jinja2/mako/cheetah adapters), static file serving, cookies, JSON responses, file uploads, WSGI middleware support, built-in development server, and gunicorn/uwsgi deployment. Single-file — copy bottle.py into project for zero-dependency deployment.

Python bindings for Google's Brotli compression algorithm — optimized for web content compression with better ratios than gzip for text. brotli features: brotli.compress()/brotli.decompress() for one-shot, quality parameter (0-11, default 11), mode parameter (GENERIC/TEXT/FONT), lgwin for window size control, brotli.Compressor for incremental streaming, brotli.Decompressor for streaming decompression, process()/finish() for streaming API, and C extension binding to the official brotli library. Brotli is supported in all modern browsers as Content-Encoding: br — better ratio than gzip for HTML/CSS/JS.

Extended Python pickle for serializing complex Python objects — extends pickle to handle lambdas, closures, generators, classes, and functions defined in __main__. dill features: dill.dumps()/dill.loads() for bytes, dill.dump()/dill.load() for files, serialization of lambda functions, nested functions, closures with free variables, class definitions, interactive session objects, generators (as frozen state), interactively-defined functions, class instances with unpicklable attributes, and Pickle compatibility mode. Used in multiprocessing, Pathos parallel processing, and distributed computing where standard pickle fails.

Thread-safe dictionary with auto-expiring keys — provides a dict subclass where entries automatically expire after a TTL. expiringdict features: ExpiringDict(max_len, max_age_seconds) constructor, standard dict interface (get/set/del/in/items/values), thread-safe via RLock, max_len for LRU-style eviction, max_age_seconds for TTL expiry, expired keys invisible on access (lazy expiry), copy() creates snapshot, and len() excludes expired entries. Simpler than cachetools TTLCache with fewer features but easier API.

Memory profiling for Python — tracks memory usage line-by-line and over time. memory_profiler features: @profile decorator for line-by-line memory usage, mprof run/plot for time-based memory tracking, memory_usage() for programmatic profiling, --include-children for subprocess memory, matplotlib integration for memory graphs, interval parameter for sampling frequency, backend selection (psutil, tracemalloc, posix), timestamps for tracking memory over time, and streaming output for long-running processes. Identifies memory leaks, peak usage, and objects causing growth.

Object graph visualization and memory leak detection for Python — shows what objects are in memory and what references them. objgraph features: show_most_common_types() for top object counts by type, show_growth() for new objects since last check, find_backref_chain() for reference chains, show_backrefs()/show_refs() for object reference graphs (requires graphviz), get_leaking_objects() for objects with no referrers, count() for type count, at_addrs() for objects at specific memory addresses, and typestats()/get_new_ids() for tracking object creation. Essential for finding Python memory leaks.

Simple Python job queue backed by Redis — enqueues any Python function as a background job with minimal setup. RQ features: Queue for job dispatch, Worker for processing, job.enqueue() / q.enqueue(fn, args), result fetching via job.result, job status tracking (queued/started/finished/failed), FailedJobRegistry for failed jobs, job dependencies via depends_on=, scheduled jobs via rq-scheduler extension, multiple queues with priorities, job timeouts, retries, and rq dashboard for monitoring. Much simpler than Celery.

Build tool for compiling Rust to WebAssembly and publishing to npm. Handles wasm-bindgen glue code generation, TypeScript type definitions, and packaging for browser, Node.js, or bundler targets. The standard workflow for Rust → Wasm → npm: write Rust with #[wasm_bindgen] attributes, run wasm-pack build, publish the resulting npm package.

An open-source workflow automation platform (Zapier alternative) with 280+ integrations, a visual no-code builder, AI agent capabilities, and MCP server support that exposes all pieces as MCP-compatible tools.

Amazon Simple Queue Service (SQS) is a fully managed message queuing service offering standard (at-least-once) and FIFO (exactly-once, ordered) queues with dead-letter queue support for reliable async task processing.

MCP server for Apple App Store Connect API enabling AI agents to manage iOS/macOS app distribution, review app metadata, manage TestFlight beta testing, access sales reports, and interact with App Store Connect programmatically — automating routine app publishing and management tasks.

CyberArk Conjur REST API for DevOps-focused machine identity and secrets management platform. Enables AI agents to manage application and service identity authentication and authorization, handle secrets retrieval and injection for CI/CD pipelines and containers, access policy-based authorization with role and permission management, retrieve audit trail and secrets access logging data, manage host and machine identity lifecycle in Kubernetes environments, handle dynamic secrets generation and rotation automation, access Conjur ODP (On-premises) or Conjur Cloud policy management, retrieve secrets batch loading and synchronization data, manage AWS, Azure, and GCP authenticator integration, and integrate machine identity events with SIEM and DevSecOps platforms.

Redis and Memcached-compatible in-memory data store delivering up to 25x higher throughput than Redis on a single instance via a multi-threaded, shared-nothing architecture.