flatagents 0.3.2

A lightweight framework for building LLM-powered agents and hierarchical state machines with pluggable backends.

FlatAgents Python SDK

Python SDK for FlatAgents—YAML-configured AI agents and state machine orchestration.

LLM/machine readers: use MACHINES.md as a primary reference, it is more comprehensive and token efficient.

Install

pip install flatagents[litellm]

Quick Start

Single Agent

summarizer.yml

spec: flatagent
spec_version: "0.6.0"

data:
  name: summarizer
  model:
    provider: openai
    name: gpt-4o-mini
  system: You summarize text concisely.
  user: "Summarize: {{ input.text }}"
  output:
    summary:
      type: str
      description: A concise summary

from flatagents import FlatAgent, setup_logging, get_logger

# Optional: Enable internal logging to see agent progress
setup_logging(level="INFO")
logger = get_logger(__name__)

agent = FlatAgent(config_file="summarizer.yml")
result = await agent.execute(input={"text": "Long article..."})

logger.info(f"Summary: {result['summary']}")

State Machine

machine.yml

spec: flatmachine
spec_version: "0.1.0"

data:
  name: writer-critic
  context:
    product: "{{ input.product }}"
    score: 0
  agents:
    writer: ./writer.yml
    critic: ./critic.yml
  states:
    start:
      type: initial
      transitions:
        - to: write
    write:
      agent: writer
      output_to_context:
        tagline: "{{ output.tagline }}"
      transitions:
        - to: review
    review:
      agent: critic
      output_to_context:
        score: "{{ output.score }}"
      transitions:
        - condition: "context.score >= 8"
          to: done
        - to: write
    done:
      type: final
      output:
        tagline: "{{ context.tagline }}"

from flatagents import FlatMachine, setup_logging, get_logger

setup_logging(level="INFO")
logger = get_logger(__name__)

machine = FlatMachine(config_file="machine.yml")
result = await machine.execute(input={"product": "AI coding assistant"})

logger.info(f"Tagline: {result['tagline']}")

Configuration

Both YAML and JSON configs are supported. Pass config_file for file-based configs or config_dict for inline configs.

LLM Backends

from flatagents import LiteLLMBackend, AISuiteBackend

# LiteLLM (default)
agent = FlatAgent(config_file="agent.yml")

# AISuite
backend = AISuiteBackend(model="openai:gpt-4o")
agent = FlatAgent(config_file="agent.yml", backend=backend)

Hooks

Extend machine behavior with Python hooks:

from flatagents import FlatMachine, MachineHooks

class CustomHooks(MachineHooks):
    def on_state_enter(self, state: str, context: dict) -> dict:
        context["entered_at"] = time.time()
        return context

    def on_action(self, action: str, context: dict) -> dict:
        if action == "fetch_data":
            context["data"] = fetch_from_api()
        return context

machine = FlatMachine(config_file="machine.yml", hooks=CustomHooks())

Available hooks: on_machine_start, on_machine_end, on_state_enter, on_state_exit, on_transition, on_error, on_action

Built-in hooks: LoggingHooks, MetricsHooks, CompositeHooks

Execution Types

Configure how agents are executed in machine states:

execution:
  type: retry              # retry | parallel | mdap_voting
  backoffs: [2, 8, 16, 35] # Seconds between retries
  jitter: 0.1              # ±10% random variation

Type	Use Case
default	Single call
retry	Rate limit handling with backoff
parallel	Multiple samples (n_samples)
mdap_voting	Consensus voting (k_margin, max_candidates)

Schema Validation

from flatagents import validate_flatagent_config, validate_flatmachine_config

# Returns list of warnings/errors
warnings = validate_flatagent_config(config)
warnings = validate_flatmachine_config(config)

Examples

• helloworld — Minimal getting started
• writer_critic — Iterative refinement loop
• mdap — Multi-step reasoning with calibrated confidence
• error_handling — Error recovery patterns

Logging & Metrics

FlatAgents provides built-in standardized logging and OpenTelemetry-based metrics.

Logging

from flatagents import setup_logging, get_logger

# Configure once (respects FLATAGENTS_LOG_LEVEL env var)
setup_logging(level="INFO")
logger = get_logger(__name__)

logger.info("Agent starting...")

Environment Variables:

• FLATAGENTS_LOG_LEVEL: DEBUG, INFO, WARNING, ERROR
• FLATAGENTS_LOG_FORMAT: standard, json, simple

Metrics (OpenTelemetry)

Track performance, token usage, and costs. Metrics are opt-in.

pip install flatagents[metrics]
export FLATAGENTS_METRICS_ENABLED=true

from flatagents import AgentMonitor

with AgentMonitor("my-agent") as monitor:
    result = await agent.execute(input)
    # Automatically tracks duration, status, and can record custom metrics
    monitor.metrics["tokens"] = 1200

Supported backends via OTLP: Datadog, Honeycomb, StatsD (via collector), etc.

Specs

See flatagent.d.ts and flatmachine.d.ts for full specifications.

See MACHINES.md for state machine patterns and reference.

License

MIT — see LICENSE