2026年7月15日 技术热点总结

技术栈：Dockerfile

核心介绍：این ریپازیتوری دقیقاً به همان روش ریپازیتوری PasarGuard ساخته شده: به‌جای اینکه سورس Marzban را داخل ریپو کپی (vendor) کنیم، Dockerfile در لحظه‌ی build خودش سورس رسمی را از Gozargah/Marzban کلون می‌کند، داشبورد را می‌سازد، Xray را نصب می‌کند و ایمیج نهایی را می‌سازد. مزیت این روش نسبت به کپی‌کردن سورس داخل ریپو: 1. این ریپو را در گیت‌هاب خودتان Fork کنید. 2. در Railway: New Project → Deploy from GitHub Repo و ریپوی فورک‌شده را انتخاب کنید.

项目数据：⭐ 660 Stars，🍴 1,158 Forks

项目简介：AI SRE AgenticOps for Kubernetes and cloud infrastructure.

技术栈：Python、agenticops、ai、aiops、aisre、cloud、cloud-native、devops、kubernetes

核心介绍：它不是另一个只会给建议的运维聊天框。Flawless 将“发现问题、收集证据、生成预演、人工授权、执行变更、恢复验证、经验沉淀”连接成一个可审计闭环。 Created in Shanghai by 陆宣宇 (Xuanyu Lu). These are real console captures from the running platform, not conceptual mockups. Current release: 3.2.0. Release 3.2 adds persistent remediation lineage: every failed strategy, action, verification result, and replacement plan stays linked across operator-approved follow-up jobs. The effectiveness ledger is persi…

项目简介：Attributed, resolvable artifact references for agent handoffs — a ~30-byte token instead of pasted context. MCP-native; the reference layer for the agent-harness world.

技术栈：Rust、ai-agents、harness-engineering、llm-tools、loop-engineering

核心介绍：waggle Not a path. Not a URL. A handoff that answers back. Locations are dumb — no per-agent shaping, no receipts, no way to fix them once sent. waggle’s 30-byte token resolves into each agent’s own view, counts every read, and propagates a correction to every holder. A path can’t do that; a URL needs a server; The problem · How it work…

项目简介：A Codex skill that finds evidence-backed potential first customers from recent public signals.

技术栈：Python、codex、codex-skill、customer-discovery、early-adopters、prospecting、startup

核心介绍：A Codex skill that turns a startup URL or product idea into a qualified shortlist of potential first customers using recent public pain, demand, and timing signals. It defines the ideal customer profile, researches public sources, links the evidence behind every prospect, ranks fit and timing, drafts a source-based opener, and creates a polished HTML report. It never sends outreach automatica…

技术栈：Dockerfile

核心介绍：مثل PasarGuard-Node، این ریپو سورس Marzban-node را کپی نمی‌کند؛ Dockerfile در لحظه‌ی build از Gozargah/Marzban-node کلون می‌کند و سپس دو پچ کوچک (دقیقاً به سبک پچ‌های PasarGuard) رویش می‌زند تا با Railway سازگار شود: 1. پورت گوش‌دادن را از $PORT (متغیری که Railway خودش تزریق می‌کند) می‌خواند. 2. یک حالت REST بدون TLS اضافه می‌کند (SERVICE_TLS=false، پیش‌فرض) چون Railway خودش روی لبه‌ی شبکه TLS را ترمینیت می‌کند و نود از طریق شبکه‌ی خصوصی در دسترس پنل است؛ گواهی کلاینت هم در این حالت لازم نیست.

项目数据：⭐ 561 Stars，🍴 1,048 Forks

Exploration is essential for reliable autonomy in multi-agent systems, yet it remains unclear whether large language model (LLM) agents can explore effectively when interacting with one another. We show that modern LLM agents fail to do so, often exhibiting myopic and polarized interaction patterns that lead to suboptimal coordination and increased regret. We formalize this challenge as the Multi-Agent Exploration problem, modeling it as a partially observable stochastic game (POSG) problem in which agents must probe peers to infer their capabilities and identify effective interaction strat…

Current Video Large Language Models (Video LLMs) excel in question answering (QA) but largely operate as black boxes, providing textual answers without verifiable visual grounding. Existing explainability efforts rely on textual rationales or sparse bounding boxes, which struggle to capture complex video dynamics such as occlusions and non-rigid deformations. We propose Evidence-Backed Video Question Answering (E-VQA), a novel task requiring models to jointly output a semantic answer and precise spatio-temporal evidence: temporal segments and dense, tracked object segmentation masklets. To …

Language models are increasingly used for moral decision-making across diverse linguistic and cultural contexts, yet existing work overlooks multilinguality on three aspects: 1) multilingual evaluation benchmarks use direct translation, failing to adapt culture-specific items; 2) inference-time methods for moral reasoning rely on static, English-centric scaffolds and lack grounding in moral theory; 3) training methods for moral decision-making typically require expensive supervision from stronger models or human annotators. We address these gaps with three contributions. First, we introduce…

Why does contrastive learning with simple images and augmentations yield useful representations for downstream tasks? We address this question by analytically computing the optimal representation in terms of a contrastive loss for a range of basic augmentations and any image dataset with stationary statistics. We show that for certain augmentations the optimum can be attained by a CNN whose first layer filters are sinusoids, followed by a pointwise nonlinearity, global average pooling, and a final linear layer that performs partial whitening. We also show that the optimal weights in such CN…

Recent foundation image and video generation models offer strong generalization and controllability, but their direct application to embodied scenarios is limited by requirements for multi-view consistency, geometric coherence, and robot embodiment constraints. Existing methods typically adapt foundation models with limited robot data, often sacrificing visual knowledge acquired during large-scale pre-training. We present Xiaomi-Robotics-U0, a 38-billion-parameter multimodal autoregressive model for unified embodied synthesis. It treats embodied generation as an extension of foundation imag…

Steerability is a defining capability of generalist robot policies, yet remains largely absent in dexterous-hand systems for lack of large-scale, language-aligned, and action-accurate demonstration data. To address this bottleneck, we present a full-stack system that scales dexterous VLA pre-training from egocentric human videos and enables data-efficient real-robot post-training. It integrates EgoSmith, a data pipeline that curates in-the-wild egocentric videos into 9.6K hours of high-quality pre-training data with 9x higher throughput and better accuracy than prior SOTA; a unified robot s…