Дайджест awesome-списков korchasa (14.06.2026)

В сегодняшнем обновлении заметен один источник изменений: awesome-aiops добавил 15 проектов. Подборка смещена в сторону инфраструктуры для агентных систем, наблюдаемости, приватности и прикладных контуров вокруг LLM в продакшене. Это не просто витрина новых репозиториев: по набору видно, как быстро практики AIOps начинают включать агентные рантаймы, локальный контроль затрат, защиту данных и специализированные платформы для развертывания.

iflytek/astron-agent — 8554 звёзд. Astron Agent is an enterprise-grade platform for building, orchestrating, and deploying production-ready AI agentic workflows with integrated RPA, model management, and team collaboration.

Open-Source-Legal/OpenContracts — 1352 звёзд. OpenContracts is an open-source document intelligence platform that builds a programmable citation graph from documents, enabling AI agents, structured extraction, and a Model Context Protocol (MCP…

LiteLLM-Labs/litellm-agent-platform — 810 звёзд. LiteLLM Agent Platform provides a unified API and management console for deploying, managing, and operating various AI agent runtimes, offering session management, scheduling, and memory.

ZhangJinHaHaHa/AgentLens — 589 звёзд. AgentLens is a decentralized infrastructure and marketplace for AI Agents that provides verifiable proof of capabilities, security, and track record using on-chain audit scores, Intel SGX TEE attes…

cloudshipai/station — 425 звёзд. Station is an open-source, self-hosted platform for building, testing, and deploying intelligent multi-agent AI systems with Git-backed workflows and full observability.

openinfer-project/openinfer — 383 звезды. openinfer is a pure Rust + CUDA LLM inference engine designed for high performance and low latency serving, focusing on understanding every layer of the inference stack.

packyme/privacy-filter — 226 звёзд. Privacy Filter is a Go-based LLM gateway component for millisecond-latency PII and secret redaction from text, ensuring data privacy before interaction with large language models.

raketenkater/llm-server — 223 звезды. An intelligent launcher and OpenAI-compatible server for GGUF models on llama.cpp, featuring auto-tuned flag optimization, multi-GPU tensor-split, MoE expert placement, and hardware-matched downloads.

alphadl/AdaRubrics — 216 звёзд. AdaRubric is a pipeline for evaluating LLM agent trajectories using task-adaptive rubrics and generating dense reward signals for DPO and RLHF, enhancing agent performance.

Tejas-TA/predikit — 206 звёзд. Predikit bridges traditional ML models (scikit-learn, XGBoost) with AI agents by generating OpenAI function schemas and LangChain tools, enabling seamless integration and callable interfaces.

Javis603/token-monitor — 191 звезда. Token Monitor is a real-time, multi-device syncable desktop widget that tracks token usage, costs, and AI tool limits across various AI coding tools like Claude Code, Codex, and Cursor.

alibaba/UnifiedModel — 151 звезда. UModel is a vendor-neutral semantic runtime for enterprise AI, data governance, and operational intelligence, turning fragmented data into a unified object graph usable by AI agents.

caura-ai/caura-memclaw — 107 звёзд. MemClaw is an open-source, governed, shared memory system for multi-agent AI fleets, enabling agents to learn, recall, and compound knowledge, improving collective intelligence.

VasiHemanth/tokentelemetry — 105 звёзд. TokenTelemetry is a local, open-source observability dashboard for AI coding and autonomous agents, tracking token usage, costs, tool calls, session traces, and reasoning steps across various LLMs …

petrobras/3W — 0 звёзд. 3W is an open project from Petrobras providing a dataset and toolkit for developing machine learning models to detect and classify undesirable events in offshore oil wells.

Главный практический вывод: AIOps-поле продолжает расширяться от классической эксплуатации к управляемым агентным процессам. На одном конце находятся платформы оркестрации и развертывания, на другом — точечные компоненты для приватности, оценки траекторий, мониторинга токенов и доменных датасетов. Такая структура обновления полезна: она показывает не хайп вокруг одного подхода, а формирование полного стека вокруг AI-систем, которые нужно запускать, измерять и защищать.