Skip to main content

Python SDK

Official Python SDK for the Tensorheart Memory API.

Installation​

pip install tensorheart-memory

Quick Start​

from tensorheart_memory import MemoryClient

# Initialize client
client = MemoryClient(api_key="mem_live_your_key")

# Create a memory
memory = client.memories.create(
    text="User prefers dark mode interfaces",
    metadata={"category": "preference"}
)
print(f"Created: {memory.id}")

# Query memories
results = client.query(
    context="What are the user's UI preferences?"
)

for mem in results.memories:
    print(f"[{mem.relevance_score:.2f}] {mem.text}")

# Get an answer
answer = client.query_answer(
    context="What theme should I use for this user?"
)
print(answer.answer)

Client Configuration​

from tensorheart_memory import MemoryClient

client = MemoryClient(
    api_key="mem_live_...",
    base_url="https://api.memory.tensorheart.com",  # Optional
    timeout=30,  # Request timeout in seconds
)

Memory Operations​

Create​

memory = client.memories.create(
    text="Memory content",
    metadata={"key": "value"},
    memory_id="optional_custom_id"  # Optional
)

List​

memories = client.memories.list(page=1, per_page=50)

for mem in memories.data:
    print(mem.text)

print(f"Total: {memories.pagination.total}")

Get​

memory = client.memories.get("mem_abc123")

Update​

memory = client.memories.update(
    "mem_abc123",
    text="Updated content",
    metadata={"updated": True}
)

Delete​

client.memories.delete("mem_abc123")

# Delete all (requires confirmation)
client.memories.delete_all(confirm=True)

Query Operations​

Basic Query​

results = client.query(
    context="What does the user do for work?",
    max_memories=10,
    relevance_threshold=0.5
)

Query with Answer​

response = client.query_answer(
    context="What programming languages does the user know?"
)

print(response.answer)
print(f"Based on {len(response.memories_used)} memories")

Extract Memories​

result = client.extract(
    content="User: Hi, I'm John from Google.",
    content_type="conversation",
    metadata={"source": "chat"}
)

print(f"Extracted {result.extracted_count} memories")

Error Handling​

from tensorheart_memory import MemoryClient
from tensorheart_memory.exceptions import (
    AuthenticationError,
    RateLimitError,
    NotFoundError,
    ValidationError,
)

client = MemoryClient(api_key="...")

try:
    memory = client.memories.get("mem_xyz")
except NotFoundError:
    print("Memory not found")
except RateLimitError as e:
    print(f"Rate limited. Retry after {e.retry_after} seconds")
except AuthenticationError:
    print("Invalid API key")
except ValidationError as e:
    print(f"Invalid request: {e.message}")

Async Support​

import asyncio
from tensorheart_memory import AsyncMemoryClient

async def main():
    client = AsyncMemoryClient(api_key="mem_live_...")

    # All methods are async
    memory = await client.memories.create(text="Hello")
    results = await client.query(context="greeting")

    await client.close()

asyncio.run(main())