Documentation
¶
Overview ¶
Package config provides configuration management for the go-agents library. It defines structures for agent, model, provider, and transport configuration with support for human-readable duration strings and JSON serialization.
Configuration files use hierarchical JSON structure with transport-based organization. Example:
{
"name": "my-agent",
"system_prompt": "You are a helpful assistant",
"transport": {
"provider": {
"name": "ollama",
"base_url": "http://localhost:11434",
"model": {
"name": "llama3.2:3b",
"capabilities": {
"chat": {"format": "openai-chat", "options": {...}}
}
}
},
"timeout": "24s",
"connection_pool_size": 10
}
}
Duration values support human-readable strings ("24s", "1m", "2h") or numeric nanoseconds for programmatic configuration.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type AgentConfig ¶
type AgentConfig struct {
Name string `json:"name"`
SystemPrompt string `json:"system_prompt,omitempty"`
Format string `json:"format,omitempty"`
Client *ClientConfig `json:"client,omitempty"`
Provider *ProviderConfig `json:"provider"`
Model *ModelConfig `json:"model"`
}
AgentConfig defines the complete configuration for an agent. It includes the agent name, optional system prompt, optional client settings, provider configuration, and model configuration.
func DefaultAgentConfig ¶
func DefaultAgentConfig() AgentConfig
DefaultAgentConfig creates an AgentConfig with default values.
func LoadAgentConfig ¶
func LoadAgentConfig(filename string) (*AgentConfig, error)
LoadAgentConfig loads an AgentConfig from a JSON file and merges it with defaults. Returns an error if the file cannot be read or the JSON is invalid.
func (*AgentConfig) Merge ¶
func (c *AgentConfig) Merge(source *AgentConfig)
Merge combines the source AgentConfig into this AgentConfig. Non-empty values from source override the current values.
type ClientConfig ¶ added in v0.2.0
type ClientConfig struct {
Timeout Duration `json:"timeout"`
Retry RetryConfig `json:"retry"`
ConnectionPoolSize int `json:"connection_pool_size"`
ConnectionTimeout Duration `json:"connection_timeout"`
}
ClientConfig defines the configuration for the HTTP client layer. It includes timeout settings, retry behavior, and connection pooling parameters.
func DefaultClientConfig ¶ added in v0.2.0
func DefaultClientConfig() *ClientConfig
DefaultClientConfig creates a ClientConfig with default values.
func (*ClientConfig) Merge ¶ added in v0.2.0
func (c *ClientConfig) Merge(source *ClientConfig)
Merge combines the source ClientConfig into this ClientConfig. Positive values from source override the current values. Zero values are ignored.
type Duration ¶
Duration is a wrapper around time.Duration that supports JSON unmarshaling from both string format ("24s", "1m", "2h") and numeric nanoseconds.
Example JSON formats:
"timeout": "24s" // string format "timeout": 24000000000 // numeric nanoseconds
When marshaling to JSON, Duration outputs string format.
func (Duration) MarshalJSON ¶
MarshalJSON implements json.Marshaler for Duration. It outputs the duration in string format (e.g., "24s", "1m30s").
func (Duration) ToDuration ¶
ToDuration converts Duration to time.Duration.
func (*Duration) UnmarshalJSON ¶
UnmarshalJSON implements json.Unmarshaler for Duration. It accepts both string duration format ("24s", "1m", "2h") and numeric nanoseconds. Returns an error if the value cannot be parsed as either format.
type ModelConfig ¶
type ModelConfig struct {
Name string `json:"name,omitempty"`
Capabilities map[string]map[string]any `json:"capabilities,omitempty"`
}
ModelConfig defines the configuration for an LLM model. Name is the model identifier (e.g., "gpt-4o", "claude-3-opus", "llama3.1:8b"). Capabilities maps protocol names to their default options.
Example JSON:
{
"name": "gpt-4o",
"capabilities": {
"chat": {
"temperature": 0.7,
"max_tokens": 4096
},
"vision": {
"temperature": 0.5,
"max_tokens": 2048
}
}
}
func DefaultModelConfig ¶
func DefaultModelConfig() *ModelConfig
DefaultModelConfig creates a ModelConfig with initialized empty capabilities.
func (*ModelConfig) Merge ¶
func (c *ModelConfig) Merge(source *ModelConfig)
Merge combines the source ModelConfig into this ModelConfig. Non-empty name from source overrides the current value. Capabilities are merged at the protocol level.
type ProviderConfig ¶
type ProviderConfig struct {
Name string `json:"name"`
BaseURL string `json:"base_url"`
Options map[string]any `json:"options"`
}
ProviderConfig defines the configuration for an LLM provider. It includes the provider name, base URL, and provider-specific options (e.g., deployment, API version, authentication type).
func DefaultProviderConfig ¶
func DefaultProviderConfig() *ProviderConfig
DefaultProviderConfig creates a ProviderConfig with Ollama defaults. BaseURL is intentionally empty — providers auto-construct their default base URL when not explicitly configured (e.g., Ollama defaults to localhost:11434, Bedrock constructs from region).
func (*ProviderConfig) Merge ¶
func (c *ProviderConfig) Merge(source *ProviderConfig)
Merge combines the source ProviderConfig into this ProviderConfig. Non-empty name, base_url, and options from source override the current values.
type RetryConfig ¶ added in v0.2.0
type RetryConfig struct {
MaxRetries int `json:"max_retries"`
InitialBackoff Duration `json:"initial_backoff"`
MaxBackoff Duration `json:"max_backoff"`
BackoffMultiplier float64 `json:"backoff_multiplier"`
Jitter bool `json:"jitter"`
}
RetryConfig configures retry behavior for failed requests. Implements exponential backoff with jitter for transient failures.
func DefaultRetryConfig ¶ added in v0.2.0
func DefaultRetryConfig() RetryConfig
DefaultRetryConfig creates a RetryConfig with default values. Retries up to 3 times with exponential backoff starting at 1s, capped at 30s.
Source Files