Manifests & Versioning

from agents import Agent, function_tool
from decimalai.openai_agents import agent_config

@function_tool
def search(query: str) -> str:
    """Search the knowledge base."""
    return f"Results for {query}"

support = agent_config("research-agent")
agent = Agent(name=support.name, instructions=support.instructions, tools=[search, *support.tools])

import decimalai
from agents import Agent
from decimalai.openai_agents import instrument

decimalai.init(api_key="dai_sk_...")
agent = Agent(name="research-agent", instructions="Answer research questions.", model="gpt-4o", tools=[search])
instrument(agent=agent)

import decimalai

decimalai.init(api_key="dai_sk_...", otel=True)
# For CrewAI: decimalai.init(api_key="dai_sk_...", crewai=True)

import decimalai

decimalai.init(api_key="dai_sk_...")

# Option 1: declare it yourself
decimalai.register_manifest(
    agent_name="my-agent",
    tools=[
        {"name": "search", "schema": {"type": "object", "properties": {"query": {"type": "string"}}}},
    ],
    prompts={"system": "You are a helpful research assistant."},
    models={"default": {"provider": "openai", "model": "gpt-4o", "temperature": 0.7}},
)

# Option 2: let the trace build it. @decimalai.tool adds a schema from the type hints.
@decimalai.tool
def search(query: str) -> str:
    """Search the knowledge base."""
    return f"Results for {query}"

@decimalai.trace(agent_name="my-agent")
def run_agent(query: str) -> str:
    answer = search(query)
    decimalai.log_tool_call(name="search", input={"query": query}, output=answer)
    decimalai.log_llm_call(
        model="gpt-4o",
        input=[{"role": "user", "content": query}],
        output={"role": "assistant", "content": answer},
    )
    return answer