Struct ContextBuilder 

pub struct ContextBuilder<P: Platform> { /* private fields */ }

Builder for assembling LLM context from multiple sources.

Combines configuration, memory, skills, and session history into a structured message list for the LLM pipeline. Each source is read asynchronously through the platform abstraction.

Usage

let ctx = ContextBuilder::new(config, memory, skills, platform);
let system_prompt = ctx.build_system_prompt().await;
let messages = ctx.build_messages(&session, &["research"]).await;

Implementations

impl<P: Platform> ContextBuilder<P>

pub fn new( config: AgentsConfig, memory: Arc<MemoryStore<P>>, skills: Arc<SkillsLoader<P>>, platform: Arc<P>, ) -> Self

Create a new context builder.

Arguments

• config – Agent configuration (system prompt, memory window, etc.)
• memory – Shared memory store for reading long-term memory
• skills – Shared skills loader for reading active skill prompts
• platform – Platform abstraction for filesystem/env access

pub async fn build_system_prompt(&self) -> String

Build the system prompt from configuration.

Assembles the core identity prompt. This includes:

• A static identity header describing the assistant
• Workspace path information
• Memory file paths for the agent to reference

The returned string is suitable for a role="system" message.

pub async fn build_messages( &self, session: &Session, active_skills: &[String], ) -> Vec<LlmMessage>

Build the complete message list for an LLM call.

Assembles messages in the canonical order:

1. System prompt (role="system")
2. Active skill prompts (role="system", one per skill)
3. Long-term memory context (role="system", if non-empty)
4. Conversation history from session.get_history(memory_window)

The current user message is not included – the caller adds it after calling this method.

Arguments

• session – Current conversation session
• active_skills – Names of skills to include in context

pub fn config(&self) -> &AgentsConfig

Get a reference to the agent configuration.

pub async fn build_system_prompt_for_agent( &self, agent: &AgentDefinition, args: &str, ) -> String

Build a system prompt message for a specific agent definition.

Prepends the agent’s system_prompt (with template rendering) to the standard system prompt. If the agent has no custom system prompt, only the standard prompt is returned.

Arguments

• agent – The agent definition to inject
• args – Arguments string for template variable expansion

pub async fn build_messages_for_agent( &self, session: &Session, agent: &AgentDefinition, args: &str, extra_skill_instructions: Option<&str>, ) -> Vec<LlmMessage>

Build the complete message list for an agent-aware LLM call.

Like build_messages but additionally:

• Uses the agent’s system prompt (template-rendered)
• Activates the agent’s declared skills
• Appends any extra skill instructions as a system message

Arguments

• session – Current conversation session
• agent – Agent definition to use
• args – Arguments string for template rendering
• extra_skill_instructions – Optional additional skill text to inject