본문으로 건너뛰기

Client Configuration

Configure the FluxLoop client to control trace collection, storage, and behavior.

Default Configuration​

By default, FluxLoop uses sensible defaults:

import fluxloop

# Uses default configuration:
# - File storage in ./traces/
# - Auto-flush enabled
# - JSON serialization
@fluxloop.agent()
def my_agent(prompt: str) -> str:
    return f"Response: {prompt}"

Explicit Client Configuration​

Initialize the client explicitly for custom settings:

from fluxloop import FluxLoopClient

client = FluxLoopClient(
    storage="file",              # Storage backend type
    base_path="./my_traces",     # Where to save traces
    auto_flush=True,             # Auto-save after each trace
    flush_interval=10,           # Flush every 10 seconds
    buffer_size=100,             # Buffer up to 100 traces
    project_name="my-agent",     # Project identifier
    enabled=True,                # Enable/disable tracing
)

Configuration Options​

storage​

Type of storage backend to use.

Options:

  • "file" (default) - Save to local JSONL files
  • "http" - Send to FluxLoop collector service
  • "memory" - Store in memory (for testing)
# File storage
client = FluxLoopClient(storage="file")

# HTTP storage
client = FluxLoopClient(
    storage="http",
    collector_url="http://localhost:8000"
)

base_path​

Directory for file storage (when storage="file").

Default: "./traces"

client = FluxLoopClient(
    storage="file",
    base_path="/var/log/fluxloop/traces"
)

auto_flush​

Automatically flush traces to storage after each function call.

Default: True

# Disable auto-flush for better performance
client = FluxLoopClient(auto_flush=False)

# Manually flush when needed
client.flush()

flush_interval​

Seconds between automatic flushes (when auto_flush=False).

Default: 10

client = FluxLoopClient(
    auto_flush=False,
    flush_interval=30  # Flush every 30 seconds
)

buffer_size​

Maximum number of traces to buffer before forced flush.

Default: 100

client = FluxLoopClient(
    buffer_size=500  # Buffer up to 500 traces
)

project_name​

Identifier for your project (included in all traces).

Default: None

client = FluxLoopClient(
    project_name="my-chatbot-v2"
)

enabled​

Enable or disable tracing globally.

Default: True

# Disable tracing (useful for production)
client = FluxLoopClient(enabled=False)

collector_url​

URL for HTTP collector (when storage="http").

Default: "http://localhost:8000"

client = FluxLoopClient(
    storage="http",
    collector_url="https://collector.fluxloop.dev"
)

api_key​

API key for HTTP collector authentication.

Default: None

client = FluxLoopClient(
    storage="http",
    collector_url="https://collector.fluxloop.dev",
    api_key="your-api-key-here"
)

Environment Variables​

You can configure FluxLoop using environment variables:

# Storage configuration
export FLUXLOOP_STORAGE=file
export FLUXLOOP_BASE_PATH=./traces

# HTTP collector
export FLUXLOOP_COLLECTOR_URL=http://localhost:8000
export FLUXLOOP_API_KEY=your-api-key

# Project settings
export FLUXLOOP_PROJECT_NAME=my-agent
export FLUXLOOP_ENABLED=true

# Buffer settings
export FLUXLOOP_AUTO_FLUSH=true
export FLUXLOOP_BUFFER_SIZE=100

Environment variables take precedence over code configuration.

Per-Environment Configuration​

Different settings for dev/staging/prod:

import os
from fluxloop import FluxLoopClient

env = os.getenv("ENVIRONMENT", "dev")

if env == "production":
    client = FluxLoopClient(
        storage="http",
        collector_url="https://prod-collector.fluxloop.dev",
        api_key=os.getenv("FLUXLOOP_API_KEY"),
        enabled=True,
    )
elif env == "staging":
    client = FluxLoopClient(
        storage="http",
        collector_url="https://staging-collector.fluxloop.dev",
        enabled=True,
    )
else:  # development
    client = FluxLoopClient(
        storage="file",
        base_path="./traces",
        enabled=True,
    )

Global vs Local Configuration​

Global Configuration​

Set once for your entire application:

# app/__init__.py
from fluxloop import FluxLoopClient

client = FluxLoopClient(project_name="my-app")

Local Override​

Override for specific functions:

from fluxloop import agent

# Uses global client settings
@agent()
def agent_1(prompt: str) -> str:
    return f"A1: {prompt}"

# Override storage for this function
@agent(storage="memory")
def agent_2(prompt: str) -> str:
    return f"A2: {prompt}"

Performance Tuning​

For high-throughput scenarios:

client = FluxLoopClient(
    auto_flush=False,       # Disable immediate writes
    buffer_size=1000,       # Large buffer
    flush_interval=60,      # Flush every minute
    storage="http",         # Async HTTP is faster
)

For low-latency scenarios:

client = FluxLoopClient(
    auto_flush=True,        # Immediate writes
    buffer_size=10,         # Small buffer
    storage="file",         # Fast local writes
)

Debugging Configuration​

Enable verbose logging:

import logging
from fluxloop import FluxLoopClient

logging.basicConfig(level=logging.DEBUG)

client = FluxLoopClient(
    project_name="debug-session",
    enabled=True,
)

Next Steps​