Struct OpenAiClient 

pub struct OpenAiClient {
    api_key: Option<SecretString>,
    endpoint: String,
    model: String,
    max_tokens: Option<u32>,
    client: Client,
}

OpenAI LLM client.

API keys are stored using SecretString which zeroizes memory on drop, preventing sensitive credentials from lingering in memory after use.

Fields

api_key: Option<SecretString>

API key (zeroized on drop for security).

endpoint: String

API endpoint.

model: String

Model to use.

max_tokens: Option<u32>

Maximum completion tokens (default: 8192).

client: Client

HTTP client.

Implementations

impl OpenAiClient

pub const DEFAULT_ENDPOINT: &'static str = "https://api.openai.com/v1"

Default API endpoint.

pub const DEFAULT_MODEL: &'static str = "gpt-5-mini"

Default model.

pub const DEFAULT_MAX_TOKENS: u32 = 8192

Default max completion tokens.

pub fn new() -> Self

Creates a new OpenAI client.

pub fn with_api_key(self, key: impl Into<String>) -> Self

Sets the API key.

pub fn with_endpoint(self, endpoint: impl Into<String>) -> Self

Sets the API endpoint.

pub fn with_model(self, model: impl Into<String>) -> Self

Sets the model.

pub fn without_api_key(self) -> Self

Clears the API key (for testing scenarios).

pub fn with_http_config(self, config: LlmHttpConfig) -> Self

Sets HTTP client timeouts for LLM requests.

pub const fn with_max_tokens(self, max_tokens: u32) -> Self

Sets the maximum completion tokens.

fn validate(&self) -> Result<()>

Validates that the client is configured with a valid API key (SEC-M1).

Checks both presence and format of the API key to prevent injection attacks.

fn is_gpt5_model(&self) -> bool

Checks if the model is a GPT-5 family model.

GPT-5 models use max_completion_tokens instead of max_tokens and only support temperature=1 (default).

fn is_valid_api_key_format(key: &str) -> bool

Validates OpenAI API key format (SEC-M1).

OpenAI API keys follow the format: sk- prefix followed by alphanumeric characters. This prevents injection attacks via malformed keys.

fn request(&self, messages: Vec<ChatMessage>) -> Result<String>

Makes a request to the OpenAI API.

Trait Implementations

impl Default for OpenAiClient

fn default() -> Self

Returns the “default value” for a type.

impl LlmProvider for OpenAiClient

fn name(&self) -> &'static str

The provider name.

fn complete(&self, prompt: &str) -> Result<String>

Generates a completion for the given prompt.

fn complete_with_system(&self, system: &str, user: &str) -> Result<String>

Generates a completion with a system prompt.

fn analyze_for_capture(&self, content: &str) -> Result<CaptureAnalysis>

Analyzes content for memory capture.

fn analyze_for_capture_extended( &self, content: &str, existing_memories: Option<&str>, ) -> Result<ExtendedCaptureAnalysis>

Analyzes content for memory capture with extended security analysis.

fn classify_search_intent(&self, prompt: &str) -> Result<ExtendedSearchIntent>

Classifies search intent with namespace weights.

fn analyze_for_consolidation( &self, memories: &str, ) -> Result<ConsolidationAnalysis>

Analyzes memories for consolidation.