praisonai-code - PraisonAI

praisonai-code is the standalone code-execution agent runtime that sits between the core SDK (praisonaiagents) and the full wrapper (praisonai). It is not a GUI or a gateway — it is the engine that powers the praisonai CLI, packaged so you can use it without pulling in gateway, bots, or other wrapper integrations.

Package Boundaries

praisonai-code sits in the four-tier package model between praisonaiagents (SDK) and praisonai (wrapper). It never imports the wrapper at module load, and reaches wrapper-only features through a single bridge (praisonai_code._wrapper_bridge). Its sibling praisonai-bot owns bots, gateway, channel CLI, and the gateway scheduler tick — see the praisonai-bot SDK page. You can build an Agent and call agent.start() standalone; only gateway/bot serving needs the wrapper.

Four-tier ownership

Package	Owns	Must not depend on
`praisonaiagents`	Agent, tools, memory, hooks, framework protocols, bot/gateway protocols	`praisonai`, `praisonai-code`, `praisonai-bot`
`praisonai-code`	Terminal CLI: `run`/`chat`/`code`, Typer, runtime, llm, tool resolution	PyPI cycle on `praisonai` (lazy `_wrapper_bridge` only)
`praisonai-bot`	Bots, gateway server, channel CLI, OS daemon, gateway scheduler tick	PyPI cycle on `praisonai` (lazy `_wrapper_bridge` + `_code_bridge` only)
`praisonai` wrapper	Framework adapters (CrewAI/AutoGen), train, serve, dashboard, async jobs API, `RunPolicy` safety gate	—

Publish order: praisonaiagents → praisonai-code + praisonai-bot → praisonai

Wrapper bridge

The only sanctioned way for praisonai-code code to reach wrapper modules is through praisonai_code._wrapper_bridge. This makes standalone installs work — every wrapper call is guarded, so a missing praisonai package downgrades cleanly instead of raising ImportError at import time.

Helper	Purpose
`wrapper_available()`	Probe if `praisonai` is importable — no side effects
`import_wrapper_module(name)`	Import a wrapper module (e.g. `"praisonai.framework_adapters.registry"`) with an install-hint error if missing
`get_wrapper_attr(module, attr)`	Fetch a wrapper attribute; raise with install hint if missing
`optional_wrapper_attr(module, attr, default)`	Fetch an attribute or return the supplied default — used by e.g. `recipe_creator.py` for graceful fallback

Quick Start

Install

pip install praisonai-code

For the full four-package overview and a “which package?” decision guide, see Installation.

Set your API key

export OPENAI_API_KEY=your_openai_api_key

Any supported provider key works (ANTHROPIC_API_KEY, GEMINI_API_KEY, GROQ_API_KEY, OLLAMA_HOST, …). See Provider Auto-Detection.

Run an agent

from praisonaiagents import Agent

agent = Agent(name="researcher", instructions="You are a helpful research assistant")
response = agent.start("What are the latest advances in multimodal AI?")
print(response)

What ships in this package

The praisonai_code Python package contains the terminal-native CLI and agent runtime. Current published versions are shown on the installation page. It contains the full CLI implementation used by the praisonai wrapper command, plus the agent runtime modules.

Module	Description
`praisonai_code.cli`	Full CLI command tree (`run`, `chat`, `code`, `agent`, `agents`, `memory`, `tools`, `mcp`, `hooks`, and more)
`praisonai_code.cli.execution`	Core agent execution engine
`praisonai_code.cli.configuration`	Configuration loader and schema
`praisonai_code.cli.commands.*`	Individual CLI sub-commands

praisonai-code installs a praisonai-code console script and a python -m praisonai_code module entrypoint. On a standalone install (no wrapper), praisonai-code run --output actions "prompt", daemon start, config, and doctor work directly. Default run "prompt", chat, and code require the wrapper — install with pip install praisonai to unlock them (a clear install hint is printed if you invoke them standalone). See the Standalone limits table below.

Standalone limits (`pip install praisonai-code` only)

Command	Works standalone?	Notes
`run --help`, `config`, `doctor`	Yes
`run --output actions "…"`	Yes	In-process `Agent`
`run "…"` (default)	No	Requires `pip install praisonai`
`chat`, `code`	No	TUI / interactive legacy live in wrapper
`daemon start` (foreground)	Yes
`daemon start --background`	Yes	Spawns `python -m praisonai_code.runtime`

For the full terminal UX (chat, code, default run), install the wrapper: pip install praisonai.

What is not in this package

praisonai-code deliberately excludes the features that live in the praisonai wrapper:

Gateway — messaging relay and session management. Install praisonai-bot[gateway] for the standalone gateway, or install the full praisonai wrapper. See Standalone Bot Gateway.
Bots — Telegram, Discord, Slack, WhatsApp integrations. Install praisonai-bot[bot] for standalone channel bots.
Gateway scheduler tick — scheduled agent execution. Lives in praisonai-bot; the RunPolicy safety gate stays wrapper-only.
YAML-driven multi-bot orchestration — praisonai onboard, praisonai setup (wrapper).
Framework adapters — CrewAI, AutoGen (wrapper).

When to pick this vs the wrapper vs the SDK

Use case	Install
Embed agents in your own Python application	`pip install praisonaiagents`
Run agents from a CLI with `run --output actions` / `daemon` only	`pip install praisonai-code`
Deploy a Telegram/Slack/Discord bot + gateway without the full wrapper	`pip install "praisonai-bot[gateway,bot]"`
Full terminal UX (`chat`, `code`, default `run`) + gateway + framework adapters	`pip install praisonai`

Best Practices

Use simple imports

Always import from praisonaiagents, not from praisonai_code internals. The core SDK API is stable; internal CLI modules are implementation details.

from praisonaiagents import Agent, Task, PraisonAIAgents

Environment variables over code

Set API keys as environment variables rather than hardcoding them. praisonai-code reads the same environment variables as praisonaiagents.

export OPENAI_API_KEY=your_openai_api_key
python my_agent.py

Multi-agent patterns

praisonai-code inherits the full praisonaiagents multi-agent API:

from praisonaiagents import Agent, Task, PraisonAIAgents

researcher = Agent(name="researcher", instructions="Research the topic")
writer = Agent(name="writer", instructions="Write a summary")

task1 = Task(description="Research AI trends", agent=researcher)
task2 = Task(description="Summarise findings", agent=writer, context=[task1])

agents = PraisonAIAgents(agents=[researcher, writer], tasks=[task1, task2])
agents.start()

Upgrading to the full wrapper later

If you start with praisonai-code and later need gateway or bot features, add pip install "praisonai-bot[gateway,bot]" for the standalone bot path, or pip install praisonai for the full wrapper. Nothing in your existing code needs to change — the same from praisonaiagents import Agent imports still work.

Installation Guide

Three-package comparison and decision guide

praisonai SDK

Full wrapper with gateway and bots

praisonaiagents SDK

Core agent SDK

Quick Start

Build your first agent

​Package Boundaries

​Four-tier ownership

​Wrapper bridge

​Quick Start

​What ships in this package

​Standalone limits (pip install praisonai-code only)

​What is not in this package

​When to pick this vs the wrapper vs the SDK

​Best Practices

​Related

Installation Guide

praisonai SDK

praisonaiagents SDK

Quick Start

Package Boundaries

Four-tier ownership

Wrapper bridge

Quick Start

What ships in this package

Standalone limits (`pip install praisonai-code` only)

What is not in this package

When to pick this vs the wrapper vs the SDK

Best Practices

Related