aware-environment

aware-environment is the shared contract that lets humans and agents work on the same project surface. It defines the dataclasses, registries, renderers, and runtime helpers that every Aware environment (kernel or otherwise) uses to publish Agents, Roles, Rules, Objects, and supporting protocols. Concrete environments extend these building blocks, but the core contract shipped here is what the CLI, release pipeline, and terminal automation expect.

What ships in the package

• Typed specs & registries – immutable dataclasses describing agents, roles, rules, objects, and protocols plus registries for binding them to an environment.
• Environment container & loader – Environment.empty() returns a sterile registry bundle; load_environment("module:get_environment") imports a factory and returns a fully bound environment.
• Runtime executor – ObjectExecutor turns registered object functions into plan-executing calls that emit filesystem receipts/journals used by aware-cli and the release pipeline.
• Filesystem helpers – OperationPlan, receipt encoders, and pathspec utilities ensure object handlers produce deterministic diffs.
• Documentation renderers – constitution and environment guide renderers turn rule markdown into the human-facing guides consumed by AGENTS.md, CLAUDE overlays, and Studio panels.
• ACL & seeding helpers – utilities for modelling access control and generating the seed docs/panel manifests that environments expose.

Quick start

from aware_environment.agent.spec import AgentSpec
from aware_environment.environment import Environment
from aware_environment.object.spec import ObjectFunctionSpec, ObjectSpec
from aware_environment.role.spec import RoleSpec
from aware_environment.rule.spec import RuleSpec
from aware_environment.runtime.executor import FunctionCallRequest, ObjectExecutor

# Minimal constitution and role/agent wiring
env = Environment.empty()
env.bind_rules([
    RuleSpec(id="00-environment", title="Environment Constitution", markdown="# Welcome to Aware")
])
env.set_constitution_rule("00-environment")
env.bind_roles([
    RoleSpec(slug="environment-operator", title="Environment Operator", rule_ids=("00-environment",))
])
env.bind_agents([
    AgentSpec(slug="codex", title="Codex", role_slugs=("environment-operator",))
])

def ping(environment: Environment, message: str) -> dict[str, str]:
    return {"echo": message, "constitution": environment.get_constitution_rule().title}

env.bind_objects([
    ObjectSpec(
        type="hello",
        description="Example object",
        functions=(ObjectFunctionSpec(name="ping", handler_factory=lambda: ping),),
    )
])

executor = ObjectExecutor(env)
request = FunctionCallRequest(
    object_type="hello",
    function_name="ping",
    selectors={"agent": "codex"},
    arguments={"message": "Aware!"},
)
result = executor.execute(request)
print(result.payload["echo"])  # -> "Aware!"

The same executor is what aware-cli uses under the hood when an agent runs aware-cli object call.

Constitution-aware rendering

Environments can declare their governing rule with Environment.set_constitution_rule("00-environment"). The renderer helpers make it easy to surface that rule consistently:

• render_constitution_summary(environment) → markdown summary used in AGENTS.md and CLI overlays.
• render_environment_guide(environment, heading_level=1) → full environment guide (roles, rules, objects) generated as part of release pipelines.

If the constitution rule is missing, the renderer raises so CI catches drift.

Integration points

• aware-cli loads the active environment with load_environment and routes every object/function call through ObjectExecutor, turning handler plans into receipts and journal entries.
• Release/SDK pipelines stage panel manifests, AGENTS.md bundles, and manifest locks using the environment renderers, ensuring humans and agents see the same instructions.
• Custom environments can extend the base registries, add protocols, or seed docs via aware_environment.seed while reusing the same runtime and filesystem guarantees.

License

MIT © 2025 AWARE