开源可观测性和日志，安装并运行

CNCF Jaeger，一个分布式追踪平台

测试发现：jaeger (v2) builds from source, passes 5244 unit tests with 0 lint issues, and the all-in-one binary starts and serves HTTP on port 16686. 17 分钟。

为什么它在运行？追踪任意进程、端口、容器或文件，追溯到启动它的源头：CLI + TUI。

测试发现：witr builds, passes all 8 test packages (including -race), and traces real processes in the container. 4 分钟。

高性能可观测性数据管道。

测试发现：A preserved Vector binary reports version 0.59.0-dev-custom-3ce1054, validates a YAML config, and successfully processed three sentinel log records through demo_logs, remap, and console with no stderr output. 78 分钟。

VictoriaMetrics：快速、经济高效的监控解决方案和时间序列数据库

测试发现：VictoriaMetrics single binary builds, runs as an HTTP server on port 8428 returning valid responses, and the full unit test suite (131 packages) plus integration tests pass cleanly. 17 分钟。

Zipkin 是一个分布式追踪系统。

测试发现：Zipkin server 3.6.2-SNAPSHOT builds and runs on JDK 21, serving HTTP on port 9411, accepting spans via API v2 with InMemoryStorage, and passing all 817 unit tests. 10 分钟。

具有长期存储能力的高可用 Prometheus 设置。一个 CNCF 孵化项目。

测试发现：The thanos binary builds and runs: its query component answered HTTP 200 on /-/, served thanos_build_info metrics, resolved a real gRPC store.LabelNames call, and the filesystem object store round-tripped uploaded data, while the unit-test suite passed in every package except pkg/reloader, which cannot run in this... 21 分钟。

Coroot 是一个开源的可观测性与 APM 工具，提供 AI 驱动的根因分析。它结合了指标、日志、追踪、持续性能剖析以及基于 SLO 的告警，并带有预定义的仪表板和检查。

测试发现：Coroot binary builds, frontend static assets compile, all Go tests pass (18 packages ok, 0 failures), and the server starts on port 8090 responding HTTP 200 on its root and API endpoints. 12 分钟。

OpenTelemetry 收集器

测试发现：OpenTelemetry Collector core library (v0.161.0-dev) builds from source, binary reports its version, validates example configs, and starts/runs/stops with OTLP receiver, memory limiter processor, debug exporter, and zpages extension. 10 分钟。

基于 OpenTelemetry 的开源可观测性，适用于 GenAI 或 LLM 应用

测试发现：The OpenLLMetry monorepo is fully operational: all 35 Python packages install, build, and pass their test suites (thousands of tests exercising real OpenTelemetry instrumentation patterns via VCR-cassette-recorded API interactions and in-memory span exporters). 12 分钟。

一个 AI 驱动的下一代开源实时可观测性系统。

测试发现：HertzBeat server builds, its full Maven test suite passes except one Docker-only integration test, and the assembled server runs with working JWT login and authenticated API on port 1157 in this container. 37 分钟。

开源可观测性数据库。一个用于指标、日志和追踪的列式引擎，基于对象存储。

测试发现：GreptimeDB 1.3.0-alpha.1 built from source under Rust 1.96.1 starts successfully in standalone mode, binds HTTP on :4000, MySQL on :4002, Postgres on :4003, and gRPC on :4001, and correctly handles CREATE TABLE, INSERT, and SELECT SQL queries via the HTTP API. 82 分钟。

面向 Java 应用的生产安全动态追踪与诊断，可附加到运行中的 JVM，无需重启或重新编译。

测试发现：BTrace 3.0.0-SNAPSHOT builds from source, passes all 1552 unit and integration tests, produces distribution archives (tar.gz, zip, deb, rpm, sdkman), and the btrace CLI prints v3.0.0-SNAPSHOT (dev) with full help output. 34 分钟。

查看 Claude Code 和 Codex 实际向 API 发送的内容，以及每个部分的成本。

测试发现：All 223 tests pass, both mitmproxy modes (reverse and forward) bind correctly and route traffic, the TUI renders to the display with captures shown, and the full import chain from capture through materialization to pricing works without errors. 5 分钟。

惊艳的状态页面，开箱即用。

测试发现：Kener v4.1.3 is built, serves a status page on port 3000 with SQLite and Redis, all 136 tests pass, health endpoint confirms database and Redis connectivity. 14 分钟。

Aya 是一个用于 Rust 编程语言的 eBPF 库，专注于开发者体验和可运维性。

测试发现：The aya library builds without errors; all workspace crates pass their unit tests and doctests (173 aya lib tests, 91 doc tests, plus all other crate tests); the xtask CLI binary runs and prints usage help. 6 分钟。

从应用日志中提取内部监控数据，以便收集到时序数据库中。

测试发现：mtail binary built, all 24 Go test packages pass, and the HTTP server serves JSON and Prometheus metrics on /json and /metrics endpoints with health check responding 200. 6 分钟。

一个自托管的 CPA / CLIProxyAPI 管理面板和 AI 网关可观测性仪表板，用于请求、用量、成本、配额、失败和账户健康状态的监控。

测试发现：CPA Manager Plus web app installs, builds, type-checks, and runs all 4156 web tests passing; repo-level tests pass 256/257. The dev server serves the management panel on port 5173 and returns HTTP 200. 12 分钟。

Kiali 项目，Istio 服务网格的可观测性。

测试发现：Kiali backend binary builds and runs, all 121 Go unit test packages pass, and the React frontend UI builds successfully, but a Kubernetes cluster with Istio is required for full integration testing. 32 分钟。

适用于 Node.js/Express.js 应用的实时监控解决方案，受 status.github.com 启发，由 vi 赞助

测试发现：express-status-monitor installs with npm, all 24 mocha tests pass, and the example Express server starts and serves the status dashboard at / returning HTTP 200. 5 分钟。

超低延迟异步 C++17 日志和指标库，适用于对性能要求苛刻的应用。

测试发现：442 unit and integration tests pass on Ubuntu with GCC 13.3.0, covering bounded/unbounded queues, backend worker, logging macros, std library type formatting, sinks, metrics, thread safety, timestamp ordering, and user-defined types. 7 分钟。

快速易用的日志数据库，可高效处理 TB 级日志。

测试发现：VictoriaLogs builds from source, starts as a daemon listening on port 9428, serves HTTP 200 on its root endpoint with correct metadata, and all 17 test suites pass cleanly with zero failures. 9 分钟。

本地优先的 AI token 用量与成本追踪器，支持 31 款编码工具，包括 Claude Code、Codex、Cursor、Gemini 和 DeepSeek Harness，并提供原生应用。绝不读取提示词。

测试发现：All 3052 CLI tests pass with 0 failures; the tokentracker serve command starts and returns HTTP 200 on port 7680. 26 分钟。

可扩展的 Java 库，用于 HTTP 请求和响应的日志记录

测试发现：Logbook v4.3.0-SNAPSHOT builds and all tests pass with JDK 17.0.8.1 (Adoptium Temurin) downloaded from GitHub releases due to no system JDK being available on the Ubuntu 24.04 container. 20 分钟。

翻查 logs 不是可观测性，而是希望：宽事件、结构化错误、TypeScript 优先、每个运行时。

测试发现：All 1905 tests pass across 90 files, lint and typecheck both clean, with the redact clone function fixed to preserve exotic objects (DOMException, Error) instead of stripping them via structuredClone. 10 分钟。

一个用于构建生产级 agent 系统的 Go 框架，支持图工作流、工具、记忆、A2A、AG-UI、MCP、评估和可观测性。

测试发现：The tRPC-Agent-Go root module (trpc.group/trpc-go/trpc-agent-go) builds and all 235 test suites pass, the test/ module passes with Go 1.24.4 auto-downloaded, and the examples/ module builds successfully, all from a fresh container with Go installed from the upstream tarball. 14 分钟。

Kubernetes 的实时日志仪表板。可在终端或浏览器中查看日志。可在任何地方运行：桌面、集群、docker。

测试发现：Frontend test suite (596/596 passed, 33/33 files), Go backend test suite (all packages passed, 0 failures), and dashboard-ui production build all succeed; the kubetail CLI binary at bin/kubetail (86 MB) responds to --help and --version correctly. 17 分钟。

OpenChoreo 是一个面向 Kubernetes 的内部开发者平台。

测试发现：OpenChoreo 1.3.0 compiles, all 9 binaries build, all Go unit tests pass, and all Python agent test suites pass inside the container. 44 分钟。

Nerdlog：快速、远程优先、多主机的 TUI 日志查看器，带有时间线直方图且无需中央服务器

测试发现：nerdlog builds from Go source, reports version 0ff0e8c, and all three test suites (TestNerdlogAgent, TestCoreScenarios, TestE2EScenarios) pass without failures. 7 分钟。

Qtap：一个 eBPF agent，用于捕获加密前的网络流量，提供关于出站连接及其发起进程的丰富上下文。

测试发现：Go 1.27.1 was downloaded and installed; go build produces a working qtap binary at /work/repo/bin/qtap that responds to --help and --version; the test suite passes on 49 of 51 packages (only pkg/egress IPv6 tests fail because this container has IPv6 disabled). 14 分钟。

首个开源数据发现与可观测性平台。我们让数据从业者的生活更轻松，让你专注于业务。

测试发现：ODD Platform backend compiles, starts, and serves HTTP 200 on /actuator/health with local PostgreSQL; 294 JOOQ classes and 367 API contract classes generated; frontend builds with 705 API client files; 645/962 backend tests pass (317 Docker-dependent skipped), 372/384 frontend tests pass. 24 分钟。

面向 AI agent 的基于 trace 的 CI/CD：对生产 trace 进行评分，聚类失败，将其冻结为可重放的回归测试，并阻止导致回归的 PR。可自托管。

测试发现：Python dependencies installed and all backend + SDK tests (1330 passed, 43 skipped) pass. The FastAPI backend starts on port 8000 and serves the OpenAPI docs at /docs. Docker is not available in the container so the full stack (ClickHouse, Postgres, Redis, MinIO) cannot be launched. The frontend cannot build because... 9 分钟。

在终端或浏览器中，以实时流程图的形式观看 Claude Code 或 Codex 会话。

测试发现：Zoetrope builds from source, passes all 216 tests, passes clippy and fmt checks, and the zoe binary reads and inspects both Claude and Codex session transcripts, outputting complete agent trees with tool counts and metadata. 4 分钟。

观察 FastAPI 应用的可观测性三支柱：通过 OpenTelemetry 在 Grafana 上实现 Traces (Tempo)、Metrics (Prometheus)、Logs (Loki)。

测试发现：The FastAPI app runs standalone in this container via python main.py with all core endpoints returning expected responses (200) and Prometheus metrics exposed on /metrics; the /chain endpoint requires the Docker-based multi-service stack that could not be stood up without Docker. 4 分钟。

This repo was archived on Oct 16, 2026. Python SDK moved to langchain-ai/langsmith-python and JS SDK to langchain-ai/langsmith-javascript.

测试发现：Python SDK installs in a venv and all 2553 unit tests pass; one source file fixed for httpx2/pytest-httpx compatibility. JS SDK installs but tests fail on Node 18 (crypto.randomUUID missing) and vitest shows 113 pre-existing failures. 45 分钟。

Prometheus 规则检查器/验证器

测试发现：pint binary builds from source and passes all 14 test packages (cmd/pint, internal/checks, internal/comments, internal/config, internal/config/options, internal/diags, internal/discovery, internal/git, internal/log, internal/output, internal/parser, internal/parser/source, internal/promapi, internal/reporter) with... 4 分钟。

一个帮助 Python 开发者轻松调试 Python 程序中 HTTP(S) 客户端和服务器请求的工具。

测试发现：Editable install of httpdbg 2.2.1 in /work/repo/venv works and all 189 non-UI tests pass; pyhttpdbg traced a real local HTTP GET, returned 200 on its /requests API recording status 200, and --version printed 2.2.1. 7 分钟。

适用于 Golang 的 OpenTelemetry 编译时插桩

测试发现：The otel binary builds from source and instruments Go projects. 22 non-Docker tests pass: version package (16), build (9), helloworld, httpclient, errors, flags (2), custom, dependencies, multi-exporter, temporality (4), sharedlib. Docker-dependent tests and CompileTheWorld are not runnable in this environment. 33 分钟。

简单、模块化且可观测的 Go 框架，用于后端应用。

测试发现：Go 1.26.4 installed; go.work created; 5 issues fixed; all 28 Yokai framework modules build and pass their test suites with 0 failures. 39 分钟。

用于 Nest 框架 (node.js) 的 OpenTelemetry (追踪 + 指标) 模块

测试发现：The nestjs-otel library installs and builds under Node 22, its 118-test suite passes, and a NestJS app built against the compiled lib runs and answers an HTTP endpoint with 200. 4 分钟。

查看你的 coding agent 做了什么以及花费了多少。将每个任务分解为工作步骤：使用的工具、更改的文件、运行的测试、花费的时间和 token。本地优先的仪表板，适用于 Claude Code、Codex、OpenCode 等。无需登录，无需遥测。

测试发现：agentacct 0.12.11 installed in ~/.agentacct venv, all 4888 tests pass, demo command produces run reports, local JSON API returns 200 on /. 50 分钟。

一种用于优化多次运行的数据库查询的算法

测试发现：event-reduce-js installs, builds dist (cjs+esm), and its mocha suite passes 51/51 under Node v24.21.0. 2 分钟。

面向现代工作负载的速率限制、缓存和请求优先级

测试发现：Aperture controller, agent, and CLI binaries build; controller serves API on :8080, agent on :18081, agent connects to controller via etcd, flow-control SDK demo app sends requests through agent producing real metrics in Prometheus. 35 分钟。

OpenClaw 的官方插件，用于将 agent 追踪导出到 Opik。查看并监控 agent 行为、成本、tokens、错误等。

测试发现：The @opik/opik-openclaw v0.2.17 package installs, builds, typechecks, lints, and passes all 102 unit/smoke tests under Node.js v22.14.0 via vitest. The only skipped test is the E2E test that requires a live Opik server. 5 分钟。

OpenTelemetry 命令行工具，用于从 shell 脚本及类似环境发送事件

测试发现：otel-cli builds from source with go build and passes all 4 test packages (58 fixtures across 23 suites) using real in-process gRPC and HTTP OTLP servers; the binary also prints help, version, and status correctly. 2 分钟。

用于 JavaScript 的 Microsoft Application Insights SDK

测试发现：Rush install, full build (all 31 packages), and all 18 test suites (thousands of tests in headless Chrome) passed with zero failures. 58 分钟。

一个简单的自托管服务，用于跟踪你的个人图书馆。

测试发现：BookLogr API (Python/Flask) installed, database migrated, and running at localhost:5150 with registration, login, and OpenLibrary book search all verified working. 11 分钟。

开源、自托管的日志和追踪，支持在兼容 S3 的对象存储上直接进行全文搜索。

测试发现：Install, build, type-check, lint, format check, and all 93 API test cases pass against real PostgreSQL 16 and Quickwit 0.9.1 instances running in the container. 15 分钟。

面向生产环境 LLM 和 agent 系统的 AI 可观测性平台。

测试发现：The logfire SDK version 5.0.0 is installed, its CLI reports version info, and 47 of 48 test modules pass with ~1392 tests passing, 2 skipped, and 0 failures. 45 分钟。

Prefect 是一个用于在 Python 中构建稳健数据管道的工作流编排框架。

测试发现：Prefect 3.6.24 is installed in a .venv, imports cleanly, the server starts and responds on /api/health with 200, and 840 core tests pass when run with uv run pytest. 22 分钟。

用于 CI/CD 可靠性的开源 agent。

测试发现：OpenSRE v0.1 is installed from source and builds successfully; the CLI starts and responds to all commands (help, health, doctor, version); all 274 tests and all quality gates (lint, format, typecheck, import checks) pass. 5 分钟。

基于开源 TypeScript AI Agent 框架构建的 AI Agent 工程平台

测试发现：Monorepo builds (29/29 packages), lints (0 errors, 5 warnings), and all 25 packages with runnable unit tests pass (~2776 tests). The only test failure is @voltagent/e2e which requires an unreachable PostgreSQL database. Two test files with a pre-existing ESM dependency conflict are excluded. 26 分钟。

用于 Kubernetes 的 eBPF 分布式网络可观测性工具

测试发现：Retina agent (135 MB) and kubectl-retina CLI (132 MB) binaries compile and execute; 50 of 52 unit-test packages pass, 2 fail due to missing host binaries (etcd and bpftool) that require Kubernetes/dev cluster infrastructure. 36 分钟。

简单、可扩展、多云监控。

测试发现：All three Sensu Go binaries built successfully; backend running on port 18080 with PostgreSQL store, agent connected via websocket and sending keepalives every 20s, sensuctl authenticated performing CRUD operations, checks created and executed E2E with status 0 results visible in event list. 18 分钟。

在几分钟内将 AI agents 部署到 Google Cloud，而不是几个月。提供生产就绪的模板，内置 CI/CD、评估和可观测性。

测试发现：The agent-starter-pack v0.41.3 CLI installs, builds, and passes all 498 unit tests with the agent-starter-pack list, --help, and --version commands working correctly; uv is available at /home/runner/.local/bin/uv. 8 分钟。

k8spacket - 使用 eBPF 收集 Kubernetes 集群中的 TCP 流量和 TLS 连接元数据，并在 Grafana 中可视化

测试发现：The k8spacket Go project builds successfully with CGO_ENABLED=0 into a static binary; all 31 test packages pass (26 testable packages across cmd, internal/broker, internal/ebpf/*, internal/modules/*, internal/thirdparty/*); the binary starts and serves HTTP /metrics on port 6676 but eBPF program loading is blocked due... 6 分钟。

免费、本地工具，用于跟踪 37 个工具和 agent 的 AI 编码 token 使用量与成本（包括 Claude Code、Cursor、Codex、Gemini 等），按模型、项目和任务统计。npx codeburn

测试发现：CodeBurn v0.9.23 builds, passes 3546 of 3567 tests, TypeScript compiles clean, and the CLI prints help and version on demand. 8 分钟。

快速解决生产问题。一个由 ClickHouse 和 OpenTelemetry 驱动的开源可观测性平台，统一了会话回放、logs、指标、追踪和错误。

测试发现：All unit test suites pass across 4 packages with Node 22.23.1 (7120 tests, 0 failures). Lint passes with warnings only. 21 分钟。

APM，应用性能监控系统

测试发现：Maven build succeeds, all 641 unit tests pass, and the OAP server starts and serves GraphQL and admin HTTP endpoints using H2 database as the storage backend. 37 分钟。

CNCF 沙箱项目，一个开源分布式时序数据库，具有高并发、高性能和高可扩展性。

测试发现：openGemini ts-server builds, starts, accepts InfluxDB line protocol writes on port 8086, and returns queried data through the HTTP API; the full app test suite (14 packages) passes. 48 分钟。

一个面向最流行可观测性工具的应用可观测性门面。可以把它想象成 SLF4J，不过是用于可观测性。

测试发现：Micrometer builds and all 4682 tests pass across 39 modules; 3 pre-existing test failures remain in Dynatrace LTT percentile export, Statsd intermittent-failure timing, and Jetty12 request counting race. 30 分钟。

用于可观测性可视化的 CNCF 沙箱。支持 Prometheus、Tempo、Loki、Pyroscope 等。

测试发现：The Perses API server (81M binary) launches on port 8080, health endpoint returns HTTP 200, native auth login returns a JWT token, API returns 3 projects and 17 dashboards from provisioning data, the Perses CLI (37M binary) responds to all commands, 24 Go test suites and 162 UI tests all pass. 20 分钟。

用于 Kubernetes 故障排查的 log 查看器

测试发现：make setup completed successfully: npm install, protobuf codegen, backend codegen, MSDF font atlas, Angular production build, and Go compiler all built without errors; all 1417 frontend tests pass; all backend Go tests pass (skipping Cloud Logging tests due to missing GCP credentials); the khi binary runs and serves... 14 分钟。

开源、自托管的指标、日志和数据库告警，包含内置 AI agent 和 MCP server。

测试发现：n9e server builds, starts on port 17000 with SQLite+embedded TSDB, serves SPA and REST API, authenticates root user login, responds to Prometheus queries, and exposes a built-in MCP endpoint with 30+ tools over Streamable HTTP transport. 13 分钟。

一个 AI Gateway、注册表和代理，位于任何 MCP、A2A 或 REST/gRPC API 之前，暴露统一端点，提供集中发现、护栏和管理。优化 Agent 和 Tool 调用，并支持插件。

测试发现：ContextForge gateway installed in .venv, server launches and responds on HTTP, 592+ unit tests pass against real SQLite and auth infrastructure. 32 分钟。

自托管的 TypeScript agent 运行时，具备持久审批和可验证的运行记录。拥有它，批准它，审计它。

测试发现：The monorepo installs, builds, passes 2350 unit tests across all 4 workspaces plus the benchmark harness, and the CLI prints help, all with Node.js 22, no API keys or network required. 4 分钟。

用于生成和暴露集群级指标的附加 agent。

测试发现：kube-state-metrics builds successfully with all 18 unit test packages passing; the binary starts and reports its version; e2e tests require a real Kubernetes cluster which is not available in this container. 9 分钟。

无需修改代码即可实现分布式追踪。使用 OpenTelemetry 和 eBPF 即时监控任何应用程序。

测试发现：The Odigos monorepo Go modules common, api, and status each pass their test suites. The CLI binary compiles and runs. The frontend webapp builds as a static Next.js site. The autoscaler, k8sutils, and collector configgrpc module tests cannot execute due to missing kubebuilder/etcd infrastructure or build toolchain... 28 分钟。

高性能 Java APM。由 ASM 驱动。试试。测试。如果你觉得它更好，就使用它。

测试发现：MyPerf4J v3.6.0 builds and passes all 62 unit tests with JDK 17 + Maven 3.9.9. Java agent JAR (MyPerf4J-ASM-3.6.0.jar, 498K) loads and runs on -version. InfluxDB v2 client validated via mock HTTP server. 21 分钟。

开源的、开发者优先的 LLMOps 平台，旨在简化提示词设计、版本管理、即时交付、协作、故障排查、可观测性等。

测试发现：Pezzo server built successfully and started on port 3000, returning HTTP 200 with status OK and version 0.8.0 from /api/healthz. All nx targets (build, test) pass. Mock infrastructure for ClickHouse, Redis, Supertokens, and KMS was stood up to allow the server to initialize without Docker. 20 分钟。

跨测试框架的、自我改进的 AI 代理记忆层。

测试发现：Go 1.24.13 (auto-proxied to Go 1.25.0), Node 22.14.0, and Bun 1.4.2 are installed as user tools. All Go modules build and all test suites pass. All plugin and SDK npm/bun tests pass. Browser extension tsc check and 177 unit tests pass. The beacon-sandbox end-to-end scenario remains ungated because it requires real... 9 分钟。

面向 Kubernetes 的更好的 Prometheus 告警：智能分组、AI 增强和自动修复

测试发现：All 422 unit tests pass on Python 3.12, the Flask web app loads and accepts connections on port 9876 returning 500 on /healthz (expected: healthz requires Web.init which needs a Kubernetes cluster), and 3 infrastructure-dependent tests (k8s_utils, kind_cluster) are skipped as expected without a Kubernetes cluster. 12 分钟。

后端网络检查器

测试发现：subtrace binary builds with Go! 1.24.2, version command prints full build metadata, and the fd reference-counting unit test passes. 10 分钟。

简单易用的 Prometheus SLO（服务级别目标）生成器

测试发现：Sloth CLI is fully functional: generate, validate, server (with fake Prometheus backend), and version commands all work. The test suite passes completely (32 of 32 test packages). 9 分钟。

开源、端到端平台，用于评估、观察和改进 LLM 和 AI agent 应用。追踪 · 评估 · 模拟 · 数据集 · 网关 · 护栏。可自托管。Apache 2.0。

测试发现：Frontend dependencies installed with yarn, unit and integration tests pass (11/11) using jsdom environment. Backend dependencies installed via uv sync (441 packages). Django 5.1.8 and Django REST framework import correctly. Backend cannot fully start or run tests without Docker for PostgreSQL, ClickHouse, Redis, and... 50 分钟。

Kubernetes 智能。全球最具潜力的 Kubernetes 可视化工具，面向开发者和平台工程团队。

测试发现：Karpor binary builds and runs; its API server responds OK on /healthz, /livez, and /readyz over HTTPS; 25 Go test packages all pass; a mock in-memory search storage backend replaces the Elasticsearch dependency for development. 21 分钟。

开源的可观测性工具，使用 AI agents 自动修复你的软件

测试发现：pnpm install succeeded (51s). Node v22.23.3 with pnpm 9.15.9. 143 proxy tests pass, 78 worker tests pass, ~380+ total tests pass across all packages with 0 failures. Full stack requires Docker for postgres/clickhouse/collector which is unavailable without root. 46 分钟。

用 Go 编写的 AI 网关 / AI 控制平面 / AI 代理。为 OpenAI、Anthropic、Gemini、Groq、xAI、Ollama、vLLM 等提供统一的兼容 OpenAI 和 Anthropic 的 API。LiteLLM...

测试发现：GoModel v0.0.0-test binary compiles, all unit and e2e tests pass, server starts and responds 200 on /health and /admin/dashboard endpoints with 601 models listed. 8 分钟。

基于 eBPF 的云原生监控工具

测试发现：Kindling collector binary builds and runs continuously in unprivileged container, initializing all analyzers and emitting self-observability JSON metrics to stdout; the eBPF probe cannot load without root/kernel-module permissions, HTTP pprof/metrics endpoints stop serving after ~60s, and pre-existing test failures in... 66 分钟。

使用 eBPF 的 OpenTelemetry 自动插桩

测试发现：OpenTelemetry Go auto-instrumentation builds, all 23 Go packages pass tests, and the CLI binary produces correct usage output. 7 分钟。

Apache SkyWalking 的 Java agent

测试发现：The SkyWalking Java agent 9.8.0-SNAPSHOT builds, its 1888 unit tests pass, and a demo JVM launched under the agent sent a real trace segment to a mock collector. 43 分钟。

生产就绪的 MCP 服务器框架 • 构建、部署并扩展安全的 AI agent 基础设施 • 包含认证、可观测性、调试器、遥测与运行时 • 运行驱动 AI agents 的真实 MCP

测试发现：golf-mcp 0.4.1 is installed in a venv, CLI works (init/build/run), all 305 tests pass, and the built example server starts on port 3000 with working streamable-http transport and static-token auth. 4 分钟。

基于 eBPF 的 Kubernetes 网络可观测性。索引 L4/L7 流量，并带有完整的 K8s 上下文，无需密钥即可解密 TLS。可通过 MCP 供 AI agents 查询，也可通过仪表盘供人类查询。

测试发现：Go toolchain installed, project builds, binary produces valid output for help and version commands, and all 4 test packages (cmd, config, kubernetes, semver) pass with every subtask succeeding. 10 分钟。

一个用于 Grafana 的 operator，通过 Kubernetes/OpenShift CRs 安装和管理 Grafana 实例、仪表盘和数据源。

测试发现：Grafana Operator binary builds and runs. All 59 unit tests pass across 5 packages. Integration tests (requiring Docker/Grafana TestContainers) skipped without Docker. 13 分钟。

云原生网络与网络安全

测试发现：Go 1.27.1 installed, 38 test packages pass, 6 Calico binaries (calico-node, calicoctl, calico-ipam, kube-controllers, confd, apiserver) build and execute, libcalico-go library compiles and passes vet; felix and cni-plugin install/integration tests fail without system-level C headers or a Kubernetes cluster. 30 分钟。

使用 eBPF 捕获 SSL/TLS 明文：无需 MITM 代理或自定义 CA 安装。支持 Linux 和 Android，适用于 x86_64\arm64 和 loong64。

测试发现：eCapture binary (bin/ecapture) built as a 41 MB statically-linked ELF on Linux x86_64, runs and prints help/version output, 21 test packages pass, but real probe capture requires root and was not verified. 26 分钟。

基于 eBPF 的 Web 应用和网络指标自动插桩

测试发现：Prebuilt Beyla v3.36.0 binary launches and prints version. 11 Go test packages pass. Example HTTP service compiles. Source build fails due to missing eBPF bytecode generation requiring Docker or clang >=19. 44 分钟。

为你的 AI agent 提供一个 Pup：一个拥有 200+ 条命令、覆盖 33+ 个 Datadog 产品的 CLI 伴侣。

测试发现：无法验证；log 显示了它停在哪里。 34 分钟。

面向 AI 原生公司的自托管 AI Agents 平台。Claude Code、Codex、Gemini、OpenCode agents。Apache 2.0。

测试发现：无法验证；log 显示了它停在哪里。 29 分钟。

下一代 AI Agent 优化平台：Cozeloop 通过提供从开发、调试、评估到监控的全生命周期管理能力，应对 AI Agent 开发中的挑战。

测试发现：无法验证；log 显示了它停在哪里。 87 分钟。