fix empty background issue

[bugsz]

Closes #24

📑 Description

Since in stranger relations the other's background is not visible, using and will wipe out all backgrounds. So we use or instead.
@XuhuiZhou Here is the fix. I will run experiments later and we can merge after it behaves correctly

✅ Checks

• My pull request adheres to the code style of this project
• My code requires changes to the documentation
• I have updated the documentation as required
• All the tests have passed
• Branch name follows type/descript (e.g. feature/add-llm-agents)
• Ready for code review

ℹ Additional Information

[bugsz]

I think there are some conflicts with previous expectations?
In text_xml_renderer.py it seems that we are expecting backgrounds to be empty. Maybe we should also change that?
@ProKil @XuhuiZhou What do you think?
Here is the code

sotopia/tests/renderers/test_xml_renderer.py

Lines 90 to 97 in eac12c8

	print(obs["John"].last_turn)
	assert (
	obs["John"].last_turn == "Here is the context of this interaction:\n"
	"Scenario: test\n"
	"Participants: John and Jane\n"
	"John's goal: agent_1's goal\n"
	"Jane's goal: Unknown"
	)

[ProKil]

@bugsz Thanks for your contribution. Can you describe the test case and what your expected outputs are?

[bugsz]

I think what Xuhui wants is to make the speaker background visible while the listener invisible?
For example, in the test cases above, in which john and jane are strangers, we might want the following output

obs["John"].last_turn == "Here is the context of this interaction:\n" 
     "Scenario: test\n"
     "Participants: John and Jane\n" 
     "John's background: this is a background"
     "Jane's background: unknown"
     "John's goal: agent_1's goal\n" 
     "Jane's goal: Unknown"

[ProKil]

@bugsz that makes sense. Would you mind updating that test case and reply me with a link to it?

[bugsz]

@ProKil Here is the updated test case.

sotopia/tests/renderers/test_xml_renderer.py

Lines 64 to 99 in fd6700a

	def test_renderer_in_env() -> None:
	env = ParallelSotopiaEnv(
	env_profile=EnvironmentProfile(
	scenario="test",
	agent_goals=["agent_1's goal", "agent_2's goal"],
	)
	)
	agents = Agents(
	{
	"John": BaseAgent(
	agent_profile=AgentProfile(
	first_name="John",
	last_name="Doe",
	)
	),
	"Jane": BaseAgent(
	agent_profile=AgentProfile(
	first_name="Jane",
	last_name="Doe",
	)
	),
	}
	)
	obs = env.reset(agents=agents)
	print(obs["John"].last_turn)
	assert (
	obs["John"].last_turn == "Here is the context of this interaction:\n"
	"Scenario: test\n"
	"Participants: John and Jane\n"
	"John's background: John Doe is a 0-year-old . pronouns. Personality and values description: John's secrets: \n"
	"Jane's background: Unknown\n"
	"John's goal: agent_1's goal\n"
	"Jane's goal: Unknown"
	)