Skip to content

Python SDK

The official Suprflo SDK for Python. Wraps the REST API.

Install

bash
pip install suprflo
# with the LangChain integration:
pip install "suprflo[langchain]"

Quickstart

python
from suprflo import MemoryClient   # MemoryGClient is an alias

client = MemoryClient(api_key="YOUR_API_KEY")   # base_url defaults to https://api.suprflo.com

# Add memories from a message (string, {role, content}, or a list)
client.add("I'm vegetarian and allergic to peanuts", user_id="alice")

# Recall relevant memories
hits = client.search("what can this user eat?", user_id="alice", top_k=5)
for m in hits["results"]:
    print(m["memory"], m["score"])

Use it as a context manager to close the HTTP connection cleanly:

python
with MemoryClient(api_key="…") as client:
    client.add(messages, user_id="alice")

Methods

MethodREST call
add(messages, *, user_id=…, agent_id=…, run_id=…, app_id=…, metadata=…, infer=True, async_mode=False, timestamp=…)POST /api/memories
search(query, *, top_k=10, filters=…, threshold=…, rerank=…, memory_types=…, as_of=…, user_id=…)POST /api/memories/search
get_all(*, user_id=…, agent_id=…, run_id=…, app_id=…, limit=100, offset=0, memory_types=…)GET /api/memories
get(memory_id)GET /api/memories/{id}
update(memory_id, data)PUT /api/memories/{id}
delete(memory_id)DELETE /api/memories/{id}
delete_all(*, user_id=…, …)DELETE /api/memories
history(memory_id)GET /api/memories/{id}/history
users()GET /api/memories/users

Errors

Non-2xx responses raise APIError(status_code, message) (subclass of SuprfloError), with message taken from the API's detail. 503 is a retryable "upstream busy" signal.

Full source: the suprflo package on GitHub.

The memory layer for AI agents.