Struct LlmClientAdapter 

pub struct LlmClientAdapter { /* private fields */ }

Adapt a generic Arc<dyn LlmClient> into the narrow ExtractMemoriesLlm surface so the boot loop can construct an ExtractMemories from any provider’s concrete client.

The adapter packages the system prompt + serialized transcript into a ChatRequest, calls the upstream LLM, and pulls the first ResponseContent::Text block out. The trait’s max_tokens argument lands on ChatRequest::max_tokens so the extractor’s per-call cap reaches the provider — important when the conversation is long and the extract prompt itself already consumed a lot of input tokens.

Provider-agnostic: the adapter has no Anthropic / MiniMax / OpenAI / Gemini / DeepSeek / xAI / Mistral specifics. Switch the underlying LlmClient impl at agent boot and the same extract behaviour runs unchanged.

Splitting via LlmClientAdapter keeps the trait surface narrow and lets the driver-loop crate stay decoupled from the binary’s LlmAgentBehavior call sites.

Implementations

impl LlmClientAdapter

pub fn new(llm: Arc<dyn LlmClient>, model: impl Into<String>) -> Self

Trait Implementations

impl ExtractMemoriesLlm for LlmClientAdapter

fn chat<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, system_prompt: &'life1 str, user_messages: &'life2 str, max_tokens: u32, ) -> Pin<Box<dyn Future<Output = Result<String, String>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Send a chat request and return the text content of the response.