Merge pull request #93 from andyk/broken_agentos_run

fix bugs in quickstart code. fixes #88, #89, #90

Author: trush

README.rst

@@ -31,10 +31,12 @@ The AgentOS docs are at `agentos.org <https://agentos.org>`_.
 Install and Explore
 ===================
 
-AgentOS requires Python 3.5 - 3.8. To get started, use pip to install agentos,
-and then run a simple agent::
+AgentOS requires Python 3.5 - 3.8 and `conda`. To get started, use pip to
+install agentos, and then run a simple agent::
 
-  # First make sure you're using Python 3.5 - 3.8
+  # Make sure you're using Python 3.5 - 3.8
+  # Make sure you have miniforge, miniconda, or conda installed (required
+  # for `agentos run`, which uses MLflow Projects).
   pip install agentos
   agentos run agentos.agents.RandomAgent gym.envs.classic_control.CartPoleEnv
 

documentation/includes/install_and_try.rst

@@ -1,10 +1,12 @@
 Install and Explore
 ===================
 
-AgentOS requires Python 3.5 - 3.8. To get started, use pip to install agentos,
-and then run a simple agent::
+AgentOS requires Python 3.5 - 3.8 and `conda`. To get started, use pip to
+install agentos, and then run a simple agent::
 
-  # First make sure you're using Python 3.5 - 3.8
+  # Make sure you're using Python 3.5 - 3.8
+  # Make sure you have miniforge, miniconda, or conda installed (required
+  # for `agentos run`, which uses MLflow Projects).
   pip install agentos
   agentos run agentos.agents.RandomAgent gym.envs.classic_control.CartPoleEnv
 

documentation/quickstart.rst

@@ -23,14 +23,13 @@ The best way to understand AgentOS is to use it, so let’s install agentos and
 write a simple agent in Python that behaves randomly. Installation is easy::
 
   $ pip install agentos
-  $ pip install gym  # AgentOS uses OpenAI Gym's Environment abstraction.
 
 Writing a trivial agent is also easy::
 
   # Save this code in ./simple_agent.py
-  from agentos import Agent
+  import agentos
 
-  class SimpleAgent(Agent):
+  class SimpleAgent(agentos.Agent):
       def advance(self):
           obs, reward, done, _ = self.env.step(self.env.action_space.sample())
           print(f"Took a random step, done = {done}.")
@@ -76,7 +75,8 @@ Now, instead of writing our own while loop, let’s run the agent in a Python
 Thread using a convenience function that AgentOS provides which serves as an
 Agent Runner::
 
-  >>> agentos.run_agent(SimpleAgent, CartPoleEnv)
+  >>> from agentos import run_agent
+  >>> run_agent(SimpleAgent, CartPoleEnv)
   Took a random step, done = False.
   …
   Took a random step, done = False.
@@ -94,8 +94,9 @@ As an alternative to using our custom agent above, let’s go back into the
 Python shell we opened above and use AgentOS’s RandomAgent::
 
   >>> from agentos.agents import RandomAgent
+  >>> from agentos import run_agent
   >>> from gym.envs.classic_control import CartPoleEnv
-  >>> agentos.run_agent(RandomAgent, CartPoleEnv)
+  >>> run_agent(RandomAgent, CartPoleEnv)
   Took a random step, done = False.
   …
   Took a random step, done = False.