from omnicoreagent import DeepAgent

# Create a DeepAgent
agent = DeepAgent(
    name="research_coordinator",
    system_instruction="You are a research coordinator specializing in tech analysis.",
    model_config={"provider": "openai", "model": "gpt-4o"},
)

# Initialize (registers subagent spawning tools)
await agent.initialize()

# Run a complex task - it will spawn subagents automatically
result = await agent.run("""
Research the benefits of Rust vs Go for cloud-native applications.
Consider performance, developer experience, and ecosystem maturity.
""")

print(result["response"])
await agent.cleanup()

# The lead agent can call this tool
{
    "name": "performance_researcher",
    "role": "Performance benchmarking specialist",
    "task": "Research runtime performance of Rust vs Go for web services",
    "output_path": "/memories/research/performance"
}

# The lead agent can call this tool
[
    {"name": "perf", "role": "Perf expert", "task": "Benchmark tests", "output_path": "/memories/perf"},
    {"name": "dx", "role": "DevEx expert", "task": "Compare tooling", "output_path": "/memories/dx"},
    {"name": "ecosystem", "role": "Ecosystem analyst", "task": "Library coverage", "output_path": "/memories/ecosystem"}
]

agent = DeepAgent(
    name="analyst",
    system_instruction="Your domain-specific instruction",
    model_config={"provider": "openai", "model": "gpt-4o"},
    agent_config={
        # These are the defaults (you can override):
        "max_steps": 50,  # More steps for orchestration
        "tool_call_timeout": 600,  # 10 min (subagents do deep work)
        "memory_tool_backend": "local",  # ENFORCED (cannot override)
        "context_management": {"enabled": True},
        "tool_offload": {"enabled": True},
    }
)

from omnicoreagent import DeepAgent

# 1. Create
agent = DeepAgent(
    name="coordinator",
    system_instruction="Your domain instruction",
    model_config={"provider": "openai", "model": "gpt-4o"},
    mcp_tools=[...],           # Optional MCP tools
    local_tools=ToolRegistry(),  # Optional local tools
    agent_config={...}          # Optional config overrides
)

# 2. Initialize (registers orchestration tools)
await agent.initialize()

# 3. Run
result = await agent.run(query, session_id="optional_session")

# 4. Properties
agent.is_initialized  # Check if ready
agent.prompt_builder  # Access custom prompt builder

# 5. Cleanup
await agent.cleanup()

from omnicoreagent import DeepAgent, ToolRegistry

# Define domain-specific tools
tools = ToolRegistry()

@tools.register_tool("run_sql_query")
def run_sql_query(query: str) -> dict:
    """Execute SQL query and return results."""
    # Your database logic
    return {"rows": [...]}

@tools.register_tool("fetch_api_data")
def fetch_api_data(endpoint: str) -> dict:
    """Fetch data from external API."""
    # Your API logic
    return {"data": {...}}

# Create DeepAgent with your tools
agent = DeepAgent(
    name="data_analyst",
    system_instruction="""
You are a data analyst. You can run SQL queries and fetch API data.
For complex analysis, spawn subagents to handle different data sources.
    """,
    model_config={"provider": "openai", "model": "gpt-4o"},
    local_tools=tools,  # Your tools + orchestration tools
)

await agent.initialize()

# The agent can now use your tools AND spawn subagents
result = await agent.run("""
Analyze Q4 2024 sales performance across all regions.
Compare with API data from our CRM system.
""")