agent-config 0.1.0

Agent configuration loader and normalizer for AI/LLM applications

agent-config

A lightweight Python library and CLI for loading, merging, and validating agent configurations from files and environment variables.

Installation

pip install agent-config

For TOML support on Python < 3.11:

pip install agent-config[toml]

Quick Start

CLI Usage

Initialize a new configuration file:

agent-config init

This creates an agent.config.yaml file with sensible defaults.

Print the resolved configuration:

agent-config print --pretty

Get a specific value:

agent-config get model
agent-config get tools.0.name

Validate your configuration:

agent-config validate

Library Usage

from agent_config import load_config, validate_config

# Load configuration from file + environment
config = load_config()

# Access values
print(config["model"])
print(config["temperature"])

# Validate a configuration dict
issues = validate_config(config)
for issue in issues:
    print(f"{issue.level}: {issue.message}")

Configuration Sources

Configuration is loaded and merged from multiple sources in this order (later sources override earlier ones):

1. Built-in defaults - Sensible defaults for all fields
2. Configuration file - YAML, JSON, or TOML
3. Environment variables - Highest precedence

Configuration Files

The CLI automatically discovers configuration files in this order:

• agent.config.yaml
• agent.config.yml
• agent.config.json
• agent.config.toml
• .agent-configrc
• .agent-configrc.json

Example Configuration (YAML)

name: my-agent
model: gpt-4
temperature: 0.2
max_tokens: 4096

system_prompt: "You are a helpful assistant."

api_base: https://api.openai.com/v1
timeout_ms: 60000

tools:
  - name: web_search
    enabled: true
    config:
      max_results: 10

Example Configuration (JSON)

{
  "name": "my-agent",
  "model": "gpt-4",
  "temperature": 0.2,
  "max_tokens": 4096,
  "system_prompt": "You are a helpful assistant.",
  "timeout_ms": 60000,
  "tools": [
    {
      "name": "web_search",
      "enabled": true
    }
  ]
}

Environment Variables

Override any configuration value using environment variables:

Environment Variable	Config Key	Type
AGENT_NAME	name	string
AGENT_MODEL	model	string
AGENT_TEMPERATURE	temperature	float
AGENT_TOP_P	top_p	float
AGENT_MAX_TOKENS	max_tokens	int
AGENT_SYSTEM_PROMPT	system_prompt	string
AGENT_PROMPT_FILE	prompt_file	string
AGENT_API_BASE	api_base	string
AGENT_API_KEY	api_key	string
AGENT_TIMEOUT_MS	timeout_ms	int
AGENT_TOOLS_JSON	tools	JSON array

Example:

export AGENT_MODEL="gpt-4-turbo"
export AGENT_TEMPERATURE="0.5"
export AGENT_API_KEY="sk-..."
agent-config print

CLI Reference

agent-config print

Print the resolved configuration as JSON.

agent-config print [OPTIONS]

Options:
  -f, --file PATH    Path to configuration file
  -C, --cwd PATH     Working directory for config discovery
  -p, --pretty       Pretty-print JSON output
  --strict           Fail if no configuration file is found
  --no-env           Ignore environment variable overrides
  --redact           Redact sensitive values (default)
  --no-redact        Show sensitive values (use with caution)

agent-config get

Get a specific value by dot-notation path.

agent-config get PATH [OPTIONS]

Options:
  -f, --file PATH    Path to configuration file
  -C, --cwd PATH     Working directory for config discovery
  -d, --default VAL  Default value if path not found
  -j, --json         Output value as JSON
  --strict           Fail if no configuration file is found
  --no-env           Ignore environment variable overrides

Examples:

agent-config get model
agent-config get temperature --default 0.7
agent-config get tools.0.name
agent-config get tools --json

agent-config validate

Validate configuration and report issues.

agent-config validate [OPTIONS]

Options:
  -f, --file PATH    Path to configuration file
  -C, --cwd PATH     Working directory for config discovery
  -j, --json         Output issues as JSON
  --strict           Fail if no configuration file is found
  --no-env           Ignore environment variable overrides

Exit codes:

• 0 - Valid configuration
• 1 - Validation errors found

agent-config init

Create a starter configuration file.

agent-config init [OPTIONS]

Options:
  -t, --format FMT   File format: yaml or json (default: yaml)
  -f, --force        Overwrite existing configuration file
  -C, --cwd PATH     Directory to create configuration file in

Library API

load_config()

Load and merge configuration from all sources.

from agent_config import load_config

config = load_config(
    file=None,        # Explicit path to config file
    cwd=None,         # Working directory for discovery
    strict=False,     # Require config file to exist
    use_env=True,     # Apply environment overrides
    env=None,         # Custom environment dict
    redact=False,     # Mask sensitive values
)

validate_config()

Validate a configuration dictionary.

from agent_config import validate_config, ValidationIssue

issues = validate_config(config)
for issue in issues:
    print(f"[{issue.level}] {issue.path}: {issue.message}")

ValidationIssue

@dataclass
class ValidationIssue:
    level: Literal["error", "warn"]
    path: str | None
    message: str

ConfigError

Raised when configuration loading or validation fails.

from agent_config import load_config, ConfigError

try:
    config = load_config(strict=True)
except ConfigError as e:
    print(f"Config error: {e}")
    for issue in e.issues:
        print(f"  - {issue}")

Sensitive Data Handling

By default, the CLI redacts sensitive values (fields matching key, token, secret, password) when printing configuration. Use --no-redact to disable this behavior.

Warning: Be careful when using --no-redact in logs or shared environments.

Configuration Schema

Field	Type	Default	Description
name	string	null	Agent identifier
model	string	null	Model name/identifier
temperature	float	0.2	Sampling temperature
top_p	float	null	Nucleus sampling parameter
max_tokens	int	null	Maximum tokens to generate
system_prompt	string	null	System prompt text
prompt_file	string	null	Path to system prompt file
tools	array	[]	Tool configurations
api_base	string	null	API base URL
api_key	string	null	API key
timeout_ms	int	60000	Request timeout in milliseconds

Additional keys are preserved and passed through.

Import Name

Note: The PyPI package name is agent-config (with hyphen), but the Python import name uses underscores:

import agent_config

This follows Python naming conventions since hyphens are not valid in Python identifiers.