珊瑚AI 是 AI Skill Hub 本期精选Agent工作流之一。综合评分 7.5 分,整体质量较高。我们推荐使用将其纳入你的 AI 工具库,帮助提升工作效率。
珊瑚AI 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
珊瑚AI 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一:pip 安装(推荐)
pip install coral-ai
# 方式二:虚拟环境安装(推荐生产环境)
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install coral-ai
# 方式三:从源码安装(获取最新功能)
git clone https://github.com/Coral-Bricks-AI/coral-ai
cd coral-ai
pip install -e .
# 验证安装
python -c "import coral_ai; print('安装成功')"
# 命令行使用
coral-ai --help
# 基本用法
coral-ai input_file -o output_file
# Python 代码中调用
import coral_ai
# 示例
result = coral_ai.process("input")
print(result)
# coral-ai 配置文件示例(config.yml) app: name: "coral-ai" debug: false log_level: "INFO" # 运行时指定配置文件 coral-ai --config config.yml # 或通过环境变量配置 export CORAL_AI_API_KEY="your-key" export CORAL_AI_OUTPUT_DIR="./output"
High-throughput inference for your agents — run many of them in parallel over your own private data, so you pay for your context once, not on every turn. Token economics and the swarm layer behind investment_analyst.
⭐ Featured —claude-code-token-xray: I broke a month of my own Claude Code logs into tokens, time, and cost. The surprise — you don't pay to generate, you pay to re-read: ~29M unique tokens get billed as 4.35B (~150×), and 84% of the bill is input. Runs on your own~/.claudelogs; nothing leaves your machine → the breakdown · full write-up.
Each subdirectory is an independently-installable package or example. They share a coralbricks.* PEP 420 namespace but have no hard runtime coupling — pick the pieces you need.
| Package | PyPI | What it is |
|---|---|---|
[context_prep/](context_prep/) | [coralbricks-context-prep](https://pypi.org/project/coralbricks-context-prep/) | Build-time context prep: clean → chunk → embed → enrich → hydrate. Plain functions over list[dict] records — no loaders, no orchestrator. |
[integrations/airbyte/](integrations/airbyte/) | [coralbricks-airbyte](https://pypi.org/project/coralbricks-airbyte/) | Ingestion bridge: reads Airbyte destination output (600+ connectors) into list[dict] records that feed context_prep. |
[py-gpu-inference/](py-gpu-inference/) | [coralbricks-gpu-inference](https://pypi.org/project/coralbricks-gpu-inference/) | Production gRPC GPU embedding server. Token-bucket batching, dual backpressure, torch.compile + CUDA graphs — pure Python/PyTorch, no ONNX/TensorRT. |
| Path | What it shows |
|---|---|
[event_scout/](event_scout/) | A small agent that scrapes upcoming AI/tech events (Luma + Eventbrite) via TinyFish and dedups against CoralBricks memory across runs. |
| Package | PyPI | What it is |
|---|---|---|
[integrations/crewai/](integrations/crewai/) | [coralbricks-crewai](https://pypi.org/project/coralbricks-crewai/) | CrewAI memory backend — CoralBricksMemory + SearchCoralBricksMemoryTool. |
[integrations/langchain/](integrations/langchain/) | [coralbricks-langchain](https://pypi.org/project/coralbricks-langchain/) | LangChain memory backend — CoralBricksMemory, CoralBricksRetriever, agent tools (store / search / forget). |
[integrations/openclaw/](integrations/openclaw/skills/persistent-agent-memory/) | — | OpenClaw skill persistent-agent-memory: bash-based coral_store / coral_retrieve / coral_delete_matching. |
高质量的AI工作流项目,提供GPU原生支持
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ Apache 2.0 — 宽松开源协议,可商用,需保留版权声明和 NOTICE 文件,含专利授权条款。
经综合评估,珊瑚AI 在Agent工作流赛道中表现稳健,质量良好。如果你已有明确的使用需求,可以直接上手体验;如果还在评估阶段,建议对比同类工具后再做决策。
| 原始名称 | coral-ai |
| 原始描述 | 开源AI工作流: The memory layer for agentic AI. GPU-native embedding training, inference, and 。⭐27 · Python |
| Topics | AI工作流Python |
| GitHub | https://github.com/Coral-Bricks-AI/coral-ai |
| License | Apache-2.0 |
| 语言 | Python |
收录时间:2026-05-25 · 更新时间:2026-05-30 · License:Apache-2.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端