Offline & CI Install

Install PraisonAI in air-gapped, sandboxed, or CI environments — no network side-effects at install time.

pip install praisonai performs no network step at install time — browser binaries are provisioned on demand, so installs succeed in offline, air-gapped, sandboxed, and CI environments.

Quick Start

1. Run on the slim install

   The default install needs no browser and no network step — agents run offline.

   from praisonaiagents import Agent
   
   # Runs on the slim install — no browser needed.
   agent = Agent(instructions="Summarise today's AI news headlines")
   agent.start("Give me 5 bullets")

2. Add a browser only when you need one

   Provision the browser engine on demand the first time an agent drives a browser.

   playwright install chromium
   # Docker / CI (adds OS-level deps):
   playwright install chromium --with-deps

   from praisonaiagents import Agent
   
   agent = Agent(name="browser-agent", instructions="Browse the web and complete tasks in Chrome.")
   agent.start("Go to google and search praisonai, then summarise the top result.")

What “No Network Step” Means

pip install praisonai installs the wrapper and SDK without pulling any browser engines.

• setup.py no longer runs python -m playwright install after install.
• praisonai[browser] and praisonai[all] install the Python packages but not the browser binaries.
• First browser-tool use surfaces Playwright <browser> binary not found. Install it on demand: playwright install <browser> if binaries are missing.

Recipes

Pick the recipe that matches your environment.

Docker image with browsers pre-baked

FROM python:3.11-slim
RUN pip install praisonai[browser] && \
    playwright install chromium --with-deps

CI job that only needs the SDK (no browsers, fastest)

- run: pip install praisonai
- run: python your_agent_script.py

CI job that DOES use the browser

- run: pip install "praisonai[browser]"
- run: playwright install chromium --with-deps
- run: python your_browser_agent.py

Air-gapped install

Point pip at a private PyPI mirror and pre-stage a Playwright browser bundle, then set PLAYWRIGHT_BROWSERS_PATH so Playwright finds the pre-downloaded engines without reaching the network.

export PIP_INDEX_URL=https://your-mirror.internal/simple
export PLAYWRIGHT_BROWSERS_PATH=/opt/playwright-browsers
pip install "praisonai[browser]"

Common Patterns

Choose between pre-baking browsers and installing on demand based on how often your pipeline touches a browser.

Best Practices

Related