agents-features-memory
Provides AgentMemory feature that allows to store and persist facts from LLM history between agent runs and even between multiple agents
Overview
The agents-features-memory module provides memory capabilities for AI agents, allowing them to store, retrieve, and share information between conversations and even between different agents. This enables agents to maintain context, remember user preferences, and build knowledge over time.
Key features include:
Storage and retrieval of facts with concepts and values
Memory organization with subjects and scopes
Secure storage with encryption options
Memory sharing between agents
Automatic fact detection from agent history
Using in your project
To use the memory feature in your project, add the following dependency:
dependencies {
implementation("ai.koog.agents:agents-features-memory:$version")
}Then, install the AgentMemory feature when creating your agent:
val myAgent = AIAgents(
// other configuration parameters
) {
install(AgentMemory) {
memoryProvider = LocalFileMemoryProvider(
config = LocalMemoryConfig("my-memory"),
storage = SimpleStorage(JVMFileSystemProvider),
root = Path("memory/data")
)
featureName = "my-feature"
organizationName = "my-organization"
}
}Using in unit tests
For testing agents with memory capabilities, you can use an in-memory storage implementation:
// Create an in-memory storage for testing
val testMemoryProvider = LocalFileMemoryProvider(
config = LocalMemoryConfig("test-memory"),
storage = SimpleStorage(InMemoryFileSystemProvider),
root = Path("test/memory")
)
// Create a test agent with memory
val testAgent = AIAgents(
// other test configuration
) {
install(AgentMemory) {
memoryProvider = testMemoryProvider
featureName = "test-feature"
}
// Enable testing mode
withTesting()
}This approach allows you to test memory operations without writing to the actual file system.
Example of usage
Here's an example of using memory in an agent strategy graph:
// Define concepts for storing information
val projectStructureConcept = Concept(
"project-structure",
"Structure of the project, including modules and important files",
FactType.MULTIPLE
)
val userPreferencesConcept = Concept(
"user-preferences",
"User's preferred settings and configurations",
FactType.SINGLE
)
// Create a strategy with memory operations
val strategy = strategyGraph<Unit, String> {
// Node to load facts from memory
val loadFromMemory by nodeLoadFromMemory(
concepts = listOf(projectStructureConcept, userPreferencesConcept)
)
// Node to process user request
val processRequest by nodeLLM<Unit, String> {
// LLM processing logic
}
// Node to save facts to memory
val saveProjectInfo by nodeSaveToMemory(
projectStructureConcept,
subject = MemorySubjects.Project,
scope = MemoryScopeType.PRODUCT
)
// Connect the nodes
start - loadFromMemory - processRequest - saveProjectInfo - end
}This example demonstrates loading facts from memory, processing a request, and saving updated information back to memory.
Provides AgentMemory feature that allows to store and persist facts from LLM history between agent runs and even between multiple agents
Overview
The agents-features-memory module provides memory capabilities for AI agents, allowing them to store, retrieve, and share information between conversations and even between different agents. This enables agents to maintain context, remember user preferences, and build knowledge over time.
Key features include:
Storage and retrieval of facts with concepts and values
Memory organization with subjects and scopes
Secure storage with encryption options
Memory sharing between agents
Automatic fact detection from agent history
Using in your project
To use the memory feature in your project, add the following dependency:
dependencies {
implementation("ai.koog.agents:agents-features-memory:$version")
}Then, install the AgentMemory feature when creating your agent:
val myAgent = AIAgents(
// other configuration parameters
) {
install(AgentMemory) {
memoryProvider = LocalFileMemoryProvider(
config = LocalMemoryConfig("my-memory"),
storage = SimpleStorage(JVMFileSystemProvider),
root = Path("memory/data")
)
featureName = "my-feature"
organizationName = "my-organization"
}
}Using in unit tests
For testing agents with memory capabilities, you can use an in-memory storage implementation:
// Create an in-memory storage for testing
val testMemoryProvider = LocalFileMemoryProvider(
config = LocalMemoryConfig("test-memory"),
storage = SimpleStorage(InMemoryFileSystemProvider),
root = Path("test/memory")
)
// Create a test agent with memory
val testAgent = AIAgents(
// other test configuration
) {
install(AgentMemory) {
memoryProvider = testMemoryProvider
featureName = "test-feature"
}
// Enable testing mode
withTesting()
}This approach allows you to test memory operations without writing to the actual file system.
Example of usage
Here's an example of using memory in an agent strategy graph:
// Define concepts for storing information
val projectStructureConcept = Concept(
"project-structure",
"Structure of the project, including modules and important files",
FactType.MULTIPLE
)
val userPreferencesConcept = Concept(
"user-preferences",
"User's preferred settings and configurations",
FactType.SINGLE
)
// Create a strategy with memory operations
val strategy = strategyGraph<Unit, String> {
// Node to load facts from memory
val loadFromMemory by nodeLoadFromMemory(
concepts = listOf(projectStructureConcept, userPreferencesConcept)
)
// Node to process user request
val processRequest by nodeLLM<Unit, String> {
// LLM processing logic
}
// Node to save facts to memory
val saveProjectInfo by nodeSaveToMemory(
projectStructureConcept,
subject = MemorySubjects.Project,
scope = MemoryScopeType.PRODUCT
)
// Connect the nodes
start - loadFromMemory - processRequest - saveProjectInfo - end
}This example demonstrates loading facts from memory, processing a request, and saving updated information back to memory.
Provides AgentMemory feature that allows to store and persist facts from LLM history between agent runs and even between multiple agents
Overview
The agents-features-memory module provides memory capabilities for AI agents, allowing them to store, retrieve, and share information between conversations and even between different agents. This enables agents to maintain context, remember user preferences, and build knowledge over time.
Key features include:
Storage and retrieval of facts with concepts and values
Memory organization with subjects and scopes
Secure storage with encryption options
Memory sharing between agents
Automatic fact detection from agent history
Using in your project
To use the memory feature in your project, add the following dependency:
dependencies {
implementation("ai.koog.agents:agents-features-memory:$version")
}Then, install the AgentMemory feature when creating your agent:
val myAgent = AIAgents(
// other configuration parameters
) {
install(AgentMemory) {
memoryProvider = LocalFileMemoryProvider(
config = LocalMemoryConfig("my-memory"),
storage = SimpleStorage(JVMFileSystemProvider),
root = Path("memory/data")
)
featureName = "my-feature"
organizationName = "my-organization"
}
}Using in unit tests
For testing agents with memory capabilities, you can use an in-memory storage implementation:
// Create an in-memory storage for testing
val testMemoryProvider = LocalFileMemoryProvider(
config = LocalMemoryConfig("test-memory"),
storage = SimpleStorage(InMemoryFileSystemProvider),
root = Path("test/memory")
)
// Create a test agent with memory
val testAgent = AIAgents(
// other test configuration
) {
install(AgentMemory) {
memoryProvider = testMemoryProvider
featureName = "test-feature"
}
// Enable testing mode
withTesting()
}This approach allows you to test memory operations without writing to the actual file system.
Example of usage
Here's an example of using memory in an agent strategy graph:
// Define concepts for storing information
val projectStructureConcept = Concept(
"project-structure",
"Structure of the project, including modules and important files",
FactType.MULTIPLE
)
val userPreferencesConcept = Concept(
"user-preferences",
"User's preferred settings and configurations",
FactType.SINGLE
)
// Create a strategy with memory operations
val strategy = strategyGraph<Unit, String> {
// Node to load facts from memory
val loadFromMemory by nodeLoadFromMemory(
concepts = listOf(projectStructureConcept, userPreferencesConcept)
)
// Node to process user request
val processRequest by nodeLLM<Unit, String> {
// LLM processing logic
}
// Node to save facts to memory
val saveProjectInfo by nodeSaveToMemory(
projectStructureConcept,
subject = MemorySubjects.Project,
scope = MemoryScopeType.PRODUCT
)
// Connect the nodes
start - loadFromMemory - processRequest - saveProjectInfo - end
}This example demonstrates loading facts from memory, processing a request, and saving updated information back to memory.