zerolatency 0.2.0

Python SDK for the 0Latency agent memory API with drop-in wrappers for Anthropic, OpenAI, and Gemini

0Latency Python SDK

A lightweight Python client for the 0Latency agent memory API.

Installation

pip install zero-latency

Quick Start

from zero_latency import ZeroLatencyClient

# Initialize the client
client = ZeroLatencyClient(api_key="your_api_key_here")

# Extract memories from a conversation
result = client.extract(
    agent_id="my-agent",
    human_message="I love hiking in the mountains",
    agent_message="That's great! Mountain hiking is wonderful exercise."
)

# Recall relevant memories
memories = client.recall(
    agent_id="my-agent",
    query="What outdoor activities does the user enjoy?",
    limit=5
)

for memory in memories:
    print(f"- {memory['content']} (confidence: {memory['confidence']})")

Features

• Extract: Automatically extract and store memories from conversations
• Recall: Semantic search to find relevant memories
• Search: Keyword-based text search
• Seed: Bulk import existing facts or notes
• Graph: Explore memory relationships and connections
• Entities: Extract and track people, places, concepts
• Sentiment: Analyze emotional tone across memories
• Consolidate: Merge and deduplicate similar memories

Usage

Context Manager (Recommended)

with ZeroLatencyClient(api_key="your_api_key") as client:
    memories = client.recall(agent_id="my-agent", query="user preferences")

Direct Instantiation

client = ZeroLatencyClient(api_key="your_api_key")
try:
    memories = client.recall(agent_id="my-agent", query="user preferences")
finally:
    client.close()

Error Handling

from zero_latency import ZeroLatencyClient, AuthenticationError, ValidationError

client = ZeroLatencyClient(api_key="your_api_key")

try:
    result = client.extract(
        agent_id="my-agent",
        human_message="Hello!",
        agent_message="Hi there!"
    )
except AuthenticationError:
    print("Invalid API key")
except ValidationError as e:
    print(f"Invalid request: {e}")

API Methods

extract(agent_id, human_message, agent_message)

Extract and store memories from a conversation turn.

recall(agent_id, query, limit=10)

Recall relevant memories using semantic search.

search(agent_id, q, limit=20)

Search memories by keyword.

list_memories(agent_id, limit=50)

List all memories for an agent.

seed(agent_id, facts)

Bulk import a list of facts.

get_graph(agent_id, memory_id, depth=2)

Get memory graph traversal.

get_entities(agent_id, limit=50)

List extracted entities.

get_sentiment_summary(agent_id)

Get sentiment breakdown.

consolidate(agent_id, auto_merge=False)

Run consolidation to merge similar memories.

delete_memory(memory_id)

Delete a specific memory.

Documentation

Full documentation available at docs.0latency.ai

License

MIT License - see LICENSE file for details.

Support

• Email: [e-mail]
• Issues: GitHub Issues