Module context 

Context management module for agent memory and knowledge systems

This module provides the core infrastructure for managing agent contexts, including hierarchical memory systems, knowledge bases, and persistent context storage with session continuity.

Architecture

The context system is built around several key components:

• AgentContext: The main container for an agent’s memory and knowledge
• HierarchicalMemory: Multi-layered memory system (short-term, long-term, working, episodic, semantic)
• KnowledgeBase: Structured storage for facts, procedures, and learned patterns
• ContextManager: Trait and implementation for context operations

Usage

use symbi_runtime::context::{StandardContextManager, ContextManagerConfig};
use symbi_runtime::types::AgentId;

let config = ContextManagerConfig::default();
let context_manager = StandardContextManager::new(config, "my-agent").await?;
context_manager.initialize().await?;

let agent_id = AgentId::new();
let session_id = context_manager.create_session(agent_id).await?;

Features

• Hierarchical Memory: Multiple memory types with different retention policies
• Knowledge Management: Structured storage and retrieval of facts, procedures, and patterns
• Semantic Search: Vector-based similarity search across memory and knowledge
• Session Management: Persistent context across agent sessions
• Knowledge Sharing: Secure sharing of knowledge between agents
• Retention Policies: Automatic archiving and cleanup of old context data
• Access Control: Policy-driven access control for context operations

Re-exports

pub use types::AccessLevel;

pub use types::AgentContext;

pub use types::ContextError;

pub use types::ContextId;

pub use types::ContextPersistence;

pub use types::ContextQuery;

pub use types::FilePersistenceConfig;

pub use types::HierarchicalMemory;

pub use types::Knowledge;

pub use types::KnowledgeBase;

pub use types::KnowledgeId;

pub use types::KnowledgeItem;

pub use types::KnowledgeSource;

pub use types::KnowledgeType;

pub use types::MemoryItem;

pub use types::MemoryType;

pub use types::QueryType;

pub use types::RetentionPolicy;

pub use types::SessionId;

pub use types::StorageStats;

pub use types::VectorBatchItem;

pub use types::VectorBatchOperation;

pub use types::VectorContentType;

pub use types::VectorId;

pub use types::VectorMetadata;

pub use types::VectorOperationType;

pub use types::VectorSearchResult;

pub use manager::ContextManager;

pub use manager::ContextManagerConfig;

pub use manager::FilePersistence;

pub use manager::StandardContextManager;

pub use markdown_memory::MarkdownMemoryStore;

pub use embedding::create_embedding_service;

pub use embedding::create_embedding_service_from_env;

pub use embedding::EmbeddingConfig;

pub use embedding::EmbeddingProvider;

pub use embedding::OllamaEmbeddingService;

pub use embedding::OpenAiEmbeddingService;

pub use vector_db::EmbeddingService;

pub use vector_db::MockEmbeddingService;

pub use vector_db::NoOpVectorDatabase;

pub use vector_db::QdrantConfig;

pub use vector_db::QdrantDistance;

pub use vector_db::TfIdfEmbeddingService;

pub use vector_db::VectorDatabase;

pub use vector_db::VectorDatabaseStats;

pub use compaction::CompactionConfig;

pub use compaction::CompactionResult;

pub use compaction::CompactionTier;

pub use token_counter::context_limit_for_model;

pub use token_counter::create_token_counter;

pub use token_counter::HeuristicTokenCounter;

pub use token_counter::TiktokenCounter;

pub use token_counter::TokenCounter;

pub use vector_db_factory::create_vector_backend;

pub use vector_db_factory::resolve_vector_config;

pub use vector_db_factory::VectorBackendConfig;

pub use vector_db_lance::LanceDbBackend;

pub use vector_db_lance::LanceDbConfig;

pub use vector_db_trait::DistanceMetric;

pub use vector_db_trait::VectorDb;

Modules

compaction

Context compaction pipeline for reclaiming tokens.

embedding

Embedding service providers for generating vector embeddings

manager

Context Manager implementation for agent memory and knowledge management

markdown_memory

Markdown-backed agent memory persistence

token_counter

Multi-model token counting for context compaction.

types

Core data structures for the Context & Knowledge Systems

vector_db

Vector Database integration for Qdrant

vector_db_factory

Vector backend factory.

vector_db_lance

LanceDB embedded vector backend.

vector_db_trait

Backend-agnostic vector database trait.