Agent Memory is a class that manages conversation history, session summaries, and long-term memories for AI agents. It provides comprehensive memory management capabilities including message tracking, run history, session summarization, and persistent memory storage.
Parameters
| Parameter | Type | Description | Default |
|---|
runs | List[AgentRun] | List of conversation runs between user and agent | [] |
messages | List[Message] | List of messages sent to the model | [] |
update_system_message_on_change | bool | Whether to update system message when changed | False |
summary | Optional[SessionSummary] | Current session summary | None |
create_session_summary | bool | Whether to create session summaries | False |
update_session_summary_after_run | bool | Whether to update summary after each run | True |
summarizer | Optional[MemorySummarizer] | Summarizer for generating session summaries | None |
create_user_memories | bool | Whether to create personalized memories | False |
update_user_memories_after_run | bool | Whether to update memories after each run | True |
db | Optional[MemoryDb] | Database for storing personalized memories | None |
user_id | Optional[str] | User identifier for personalized memories | None |
retrieval | MemoryRetrieval | Memory retrieval strategy | MemoryRetrieval.last_n |
memories | Optional[List[Memory]] | List of retrieved memories | None |
num_memories | Optional[int] | Number of memories to retrieve | None |
classifier | Optional[MemoryClassifier] | Classifier for memory importance | None |
manager | Optional[MemoryManager] | Manager for memory operations | None |
Agent Run
| Parameter | Type | Description | Default |
|---|
message | Optional[Message] | Single message associated with the run | None |
messages | Optional[List[Message]] | List of messages associated with the run | None |
response | Optional[RunResponse] | Response generated during the run | None |
Session Summary
| Parameter | Type | Description | Default |
|---|
summary | str | Concise summary of the session focusing on important information | Required |
topics | Optional[List[str]] | List of topics discussed in the session | None |
Memory Summarizer
| Parameter | Type | Description | Default |
|---|
model | Optional[Model] | Model used for generating summaries | None |
use_structured_outputs | bool | Whether to use structured outputs from the model | False |
Memory Retrieval
| Parameter | Type | Description |
|---|
last_n | str | Retrieve the last N memories from history |
first_n | str | Retrieve the first N memories from history |
semantic | str | Retrieve memories based on semantic similarity |
Memory
| Parameter | Type | Description | Default |
|---|
memory | str | The actual memory content | Required |
id | Optional[str] | Unique identifier for the memory | None |
topic | Optional[str] | Topic or category of the memory | None |
input | Optional[str] | Original input that generated the memory | None |
Memory Classifier
| Parameter | Type | Description | Default |
|---|
model | Optional[Model] | Model used for classifying memories | None |
system_prompt | Optional[str] | Custom system prompt for the classifier | None |
existing_memories | Optional[List[Memory]] | List of existing memories to check against | None |
Memory Manager
| Parameter | Type | Description | Default |
|---|
model | Optional[Model] | Model used for managing memories | None |
user_id | Optional[str] | Unique identifier for the user | None |
system_prompt | Optional[str] | Custom system prompt for the memory manager | None |
db | Optional[MemoryDb] | Database for storing memories | None |
input_message | Optional[str] | Current input message being processed | None |
