惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

月光博客
月光博客
D
Docker
腾讯CDC
J
Java Code Geeks
大猫的无限游戏
大猫的无限游戏
The Cloudflare Blog
Martin Fowler
Martin Fowler
MongoDB | Blog
MongoDB | Blog
博客园 - Franky
博客园 - 三生石上(FineUI控件)
Recent Announcements
Recent Announcements
F
Fortinet All Blogs
IT之家
IT之家
WordPress大学
WordPress大学
M
MIT News - Artificial intelligence
爱范儿
爱范儿
Microsoft Azure Blog
Microsoft Azure Blog
Vercel News
Vercel News
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
小众软件
小众软件
N
Netflix TechBlog - Medium
T
Tailwind CSS Blog
Engineering at Meta
Engineering at Meta
博客园 - 【当耐特】

Hacker News - Newest: "AI"

AI can't read an investor deck AI as an attorney? Student uses ChatGPT, Gemini to sue UW over alleged racial discrimination Hacking MCP Servers in AI Systems – The Rug Pull: Tool Changes After Approval GitHub - MeepCastana/KubeezCut: Free Web based video editor Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers Coming soon: 10 Things That Matter in AI Right Now DARPA built an AI to fact-check enemy weapons claims What explains heterogeneity in AI adoption? When AI Meets Muscle: Context-Aware Electrical Stimulation Promises a New Way to Guide Human Movements - Department of Computer Science AI Changed How We Build. It Did Not Change What Matters. Linux rules on using AI-generated code - Copilot is OK, but humans must take 'full responsibility for the… Meta spins up AI version of Mark Zuckerberg to engage with employees Code Mode: Let Your AI Write Programs, Not Just Call Tools | TanStack Blog GitHub - Delavalom/graft: Go framework for building AI agents. Type-safe tools, multi-provider (OpenAI, Anthropic, Gemini, Bedrock), zero vendor SDKs. India's TCS tops estimates, says new AI models did not dent services demand Gen Z's fading AI hype Strong feeling: we are in a folded AI reality GitHub - machinarii/total-recall-catalog: A reference catalog of latest knowledge retrieval, memory & RAG systems GitHub - mensfeld/code-on-incus: Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically.. Quantization, LoRA, and the 8% Problem: Benchmarking Local LLMs for Production AI Iran war: We spoke to the man making Lego-style AI videos that experts say are powerful propaganda Powell, Bessent discussed Anthropic's Mythos AI cyber threat with major U.S. banks GitHub - immartian/bellamem: Persistent belief-graph memory for AI agents. Retrieves decisive context by importance — not recency, not RAG, not /compact. recursive-mode: The Repo-Native Operating System for AI Engineering After the attack on Sam Altman's home, will AI CEO's go on the offensive? The biggest advance in AI since the LLM Opus 4.6 vs GPT 5.4 One Prompt Unity World Generation Test “AI polls” are fake polls Client Challenge Can AI be a 'child of God'? Inside Anthropic's meeting with Christian leaders
GitHub - agentvoy/agentvoy: The universal AI agent platfo...
cthecm · 2026-05-19 · via Hacker News - Newest: "AI"

The universal AI agent platform.
Scaffold, configure, guard, and deploy AI agents across any framework.

Quick Start | Two Paths | DevTools | Frameworks | Deploy | Guard Config | Commands | Contributing

GitHub Stars CI License npm version PyPI version npm downloads


AgentVoy is a CLI tool and SDK that lets you scaffold and deploy production-ready AI agent projects in seconds — with built-in guardrails, security defaults, and support for every major agent framework.

One command. Any framework. Any model. Deploy anywhere.

AgentVoy CLI — create a project in seconds

Two Paths

AgentVoy asks upfront what you want to build:

$ npx agentvoy create my-project

  ? What do you want to build?
  > Agent — Local agent for development & experimentation
    App   — Deployable agentic app (API + UI + Docker + cloud)

Path A — Agent

Fast local development. Flat project structure, interactive REPL, zero infra.

npx agentvoy create my-project --yes
# Creates: my-project-agent/
my-project-agent/
├── agent.py           # Agent logic
├── tools.py           # Custom tools
├── run.py             # Interactive REPL
├── agent.guard.yml    # Guardrails & permissions
├── requirements.txt
└── .env.example

Path B — App

Deployable agentic app with a FastAPI server, Streamlit chat UI, real-time DevTools, and cloud configs.

npx agentvoy create my-project --build-mode app --deploy-target docker --yes
# Creates: my-project-app/
my-project-app/
├── src/
│   ├── agents/
│   │   └── agent.py          # Agent logic
│   ├── tools/
│   │   └── tools.py          # Custom tools
│   ├── trace/
│   │   └── tracer.py         # Execution tracing (auto-generated)
│   └── config/
│       └── settings.py
├── server.py                  # FastAPI — /run, /health, /dev, /ws/trace
├── streamlit_app.py           # Chat UI with model picker & glass theme
├── devtools.html              # Real-time agent DevTools dashboard
├── Dockerfile
├── docker-compose.yml
├── agent.guard.yml
├── requirements.txt
└── .env.example

Multi-agent pipelines are supported — choose sequential pipeline and name your agents:

npx agentvoy create my-project --build-mode app --agent-mode multi --yes
# Creates: researcher → writer → reviewer pipeline
src/
├── agents/
│   ├── researcher.py
│   ├── writer.py
│   └── reviewer.py
└── pipeline.py          # Sequential orchestration

Quick Start

AgentVoy — running an agent app

# Interactive — guided prompts for framework, model, and build mode
npx agentvoy create my-project

# Agent mode with defaults (OpenAI + GPT-4o)
npx agentvoy create my-project --yes

# App with Docker, fully non-interactive
npx agentvoy create my-project \
  --framework openai \
  --provider anthropic \
  --model claude-sonnet-4-20250514 \
  --build-mode app \
  --deploy-target docker \
  --yes

# Add guardrails to an existing project
npx agentvoy init

# Validate your config
npx agentvoy validate

DevTools

App-mode projects include a built-in DevTools dashboard for real-time agent observability.

agentvoy dev — Live development server

cd my-project-app
agentvoy dev

Starts your agent server with hot-reload and opens the DevTools dashboard at http://localhost:8080/dev.

What you get

  • Real-time trace streaming — WebSocket-powered event feed showing every agent action as it happens
  • Event timeline — agent_start, llm_call, tool_call, guard_check, pipeline_stage, agent_complete
  • Pipeline visualization — see multi-agent stages progress in real time
  • Detail inspector — click any event to see full payload (model, tokens, latency, tool I/O)
  • Dark-themed dashboard — single-page HTML, no extra dependencies

Endpoints

Every app-mode project exposes these DevTools endpoints:

Endpoint Description
GET /dev DevTools dashboard UI
WS /ws/trace Real-time trace event stream
GET /dev/events All events as JSON
GET /health Health check

Trace instrumentation

All 7 framework adapters are instrumented out of the box. The tracer collects:

Event Data
agent_start agent name, prompt, model
llm_call model, latency, tokens in/out
tool_call tool name, input, output, latency
guard_check check type (input/output), pass/fail
pipeline_stage stage name, index, status
agent_complete agent name, result preview

Streamlit Chat UI

App-mode projects include a production-ready chat interface:

  • Glassmorphism dark theme — styled with backdrop blur and gradient accents
  • Dynamic model switching — auto-detects API keys from .env and shows available models (GPT-4o, Claude Sonnet, Gemini Flash, etc.)
  • Guard summary — sidebar displays guardrail results for each response
  • Powered by AgentVoy footer on every generated app

Frameworks

Framework Language Status
OpenAI Agents SDK Python Available
Google ADK Python Available
CrewAI Python Available
LangGraph Python Available
Anthropic SDK Python Available
LlamaIndex Python Available
AutoGen Python Available

Model Providers

Provider Models API Key Env
OpenAI gpt-4o, o1, gpt-4-turbo OPENAI_API_KEY
Anthropic claude-opus-4, claude-sonnet-4 ANTHROPIC_API_KEY
Google gemini-2.0-flash, gemini-2.5-pro GOOGLE_API_KEY
Ollama llama3, mistral, codellama Local — no key needed
Groq llama-3.3-70b-versatile GROQ_API_KEY
Mistral mistral-large-latest MISTRAL_API_KEY

Deploy

agentvoy deploy — One-command deployment

cd my-project-app
agentvoy deploy --target docker

Docker — builds the image and runs the container:

agentvoy deploy --target docker
# Builds: docker build -t my-project .
# Runs:   docker run -p 8080:8080 --env-file .env my-project
# Agent:  http://localhost:8080
# DevTools: http://localhost:8080/dev

Fly.io — deploys to the cloud in one step:

agentvoy deploy --target fly-io
# Checks flyctl auth, sets secrets from .env, deploys
# Live URL: https://my-project.fly.dev
# DevTools: https://my-project.fly.dev/dev

Dry run — generate deployment files without deploying:

agentvoy deploy --target docker --dry-run

Deploy during creation

Pick a deployment target when creating an app project:

? Deployment target:
> Docker
  Fly.io
  Railway
  GCP Cloud Run
  AWS Lambda

Deploy an existing agent project

cd my-project-agent
npx agentvoy deploy --target docker

This generates server.py, streamlit_app.py, devtools.html, and all deployment files — without touching your existing agent code.

Deployment targets

Target Files generated CLI required
Docker Dockerfile, .dockerignore, docker-compose.yml docker
Fly.io deploy/fly.toml flyctl
Railway deploy/railway.json railway
GCP Cloud Run deploy/cloud-run.yaml gcloud
AWS Lambda deploy/template.yaml, deploy/lambda_handler.py aws, sam

Guard-to-infrastructure mapping

agent.guard.yml settings flow directly into deployment configuration:

guardrails:
  behavior:
    timeout: 5m       → Docker HEALTHCHECK interval, Cloud Run timeout
    cost_limit: $1.00 → Container memory limit (512Mi)
permissions:
  execution:
    allow_shell: false → non-root Docker user

Agent Guard Config

Every AgentVoy project includes agent.guard.yml — a universal declarative config for security and behavior.

version: "1.0"

identity:
  name: my-agent
  description: Research assistant agent
  version: 0.1.0

model:
  provider: anthropic
  model: claude-sonnet-4-20250514
  api_key_env: ANTHROPIC_API_KEY

permissions:
  network:
    mode: restricted
    allow: ["*.github.com", "*.stackoverflow.com"]
  filesystem:
    read: ["./**"]
    write: ["./output/**"]
  tools:
    require_approval: ["delete_*", "send_*", "deploy_*"]
  execution:
    allow_shell: false
    allow_subprocess: false

guardrails:
  input:
    block_prompt_injection: true
    max_tokens: 4096
    pii_detection: warn
    content_filter: moderate
  output:
    block_harmful_content: true
    max_output_tokens: 8192
  behavior:
    max_iterations: 20
    timeout: 5m
    cost_limit: "$1.00"

observability:
  tracing: true
  log_level: info
  cost_tracking: true

Runtime enforcement — agentvoy-guard

pip install agentvoy-guard
from agentvoy_guard import Guard

guard = Guard.from_config()  # reads agent.guard.yml

with guard.session() as session:
    session.check_input(user_prompt)
    result = my_agent.run(user_prompt)
    session.check_output(result)

Commands

agentvoy create [name]     # Create a new agent or app project
agentvoy dev               # Start agent server with DevTools dashboard (app mode)
agentvoy deploy            # Deploy to Docker, Fly.io, or other targets
agentvoy init              # Add agent.guard.yml to an existing project
agentvoy validate          # Validate your agent.guard.yml config
agentvoy list              # List supported frameworks, models, and targets

Architecture

agentvoy/                        # TypeScript monorepo
  packages/
    core/                        # Types, config, adapters, deployers
      src/
        adapters/                # Framework adapters (openai, crewai, …)
        deployers/               # Deployment adapters (docker, fly-io, …)
          tracer.ts              # Agent execution tracer generator
          devtools-dashboard.ts  # DevTools HTML dashboard generator
          api-wrapper.ts         # server.py generator (with /dev endpoints)
          streamlit-app.ts       # Streamlit chat UI generator
          pipeline.ts            # Multi-agent pipeline generator
        types.ts                 # Universal type system
        config.ts                # agent.guard.yml parser
    cli/                         # agentvoy CLI (create, dev, deploy, init, validate, list)
    create-agentvoy/             # npx create-agentvoy shorthand

agentvoy-guard/                  # Python runtime enforcement package

Contributing

Adding a new framework adapter

  1. Create packages/core/src/adapters/my-framework.ts
  2. Implement FrameworkAdapterscaffold() and validateConfig()
  3. Register in packages/core/src/adapters/index.ts
  4. Submit a PR

Adding a new deployment target

  1. Create packages/core/src/deployers/my-target.ts
  2. Implement DeploymentAdaptergenerateFiles() and validate()
  3. Register in packages/core/src/deployers/index.ts
  4. Submit a PR

Development

git clone https://github.com/agentvoy/agentvoy.git
cd agentvoy
npm install
npm run build

# Smoke test agent mode
node packages/cli/dist/index.js create test-project --yes

# Smoke test app mode
node packages/cli/dist/index.js create test-project --build-mode app --deploy-target docker --yes

# Test DevTools
cd test-project-app
pip install -r requirements.txt
agentvoy dev  # opens http://localhost:8080/dev

License

Apache 2.0 — see LICENSE for details.


Built by Chinmay Murugkar