Agent Memory

How memory systems integrate with the Agent class.

See Agent Overview for the core Agent class, roles, and configuration. See Memory Module for the full memory system documentation.

Memory Architecture

Cogent provides five composable memory systems:

System	Parameter	Purpose
Conversation	conversation=True	Thread-based message history (default ON)
ACC	acc=True	Agentic Context Compression — prevents drift
Knowledge	memory=True	Long-term memory with remember/recall tools
Episodic	episodic=True	Graph-backed temporal recall across sessions
Cache	cache=True	Semantic cache for tool outputs

Episodic memory supports optional reflection — EpisodicMemory(reflection=True) (every turn) or EpisodicMemory(reflection=N) (every N turns) — which synthesizes observations into reusable lessons stored in both the episodic graph and semantic memory.

See Memory Module for detailed explanation of how each system works.

Long-term memory with tools

agent = Agent(name="Assistant", model="gpt4", memory=True)

Agent gets remember(), recall(), forget() tools

Semantic cache for expensive tools

agent = Agent(name="Assistant", model="gpt4", cache=True)

All systems together

agent = Agent(
    name="SuperAgent",
    model="gpt4",
    acc=True,     # Prevents context drift
    memory=True,  # Long-term facts
    cache=True,   # Cache tool outputs
)

ACC (Agentic Context Compression)

For long conversations (>10 turns), enable ACC to prevent memory drift:

from cogent.memory.acc import AgentCognitiveCompressor

# Simple: Enable with defaults
agent = Agent(name="Assistant", model="gpt4", acc=True)

# Advanced: Custom ACC with specific bounds
acc = AgentCognitiveCompressor(max_constraints=5, max_entities=20)
agent = Agent(name="Assistant", model="gpt4", acc=acc)

See docs/acc.md for detailed ACC documentation.

Semantic Cache

For expensive tools, enable semantic caching to avoid redundant calls:

from cogent.memory import SemanticCache
from cogent.models import create_embedding

# Simple: Enable with defaults
agent = Agent(name="Assistant", model="gpt4", cache=True)

# Advanced: Custom SemanticCache instance
embed = create_embedding("openai", "text-embedding-3-small")
cache = SemanticCache(
    embedding=embed,
    similarity_threshold=0.90,  # Stricter matching
    max_entries=5000,
    default_ttl=3600,  # 1 hour
)
agent = Agent(name="Assistant", model="gpt4", cache=cache)

See docs/memory.md#semantic-cache for detailed cache documentation.