praqtordb-apps 0.1.0

Python SDK for PraqtorDB for Apps — per-user memory, knowledge graph, and Deep Memory for your application's end users.

PraqtorDB for Apps — Python SDK

Give your application's end users persistent, isolated memory — a per-user knowledge graph and AI-generated Deep Memory briefing — over two HTTP calls. This SDK wraps the PraqtorDB for Apps data-plane API (/v1).

Install

pip install -e sdk/python      # local/dev (from this repo)
# pip install praqtordb-apps   # once published

Quickstart

from praqtordb_apps import Client

c = Client(api_key="pk_live_...")          # your app's secret key
c.add(end_user_id="sarah", text="On the Pro plan at Acme.")
print(c.search(query="plan", end_user_id="sarah"))

That's it — the memory is chunked, embedded, and auto-extracted into notes, entities, and a Deep Memory briefing for sarah, isolated from every other user in your app.

Configuration

Client(api_key="pk_live_...", base_url="https://praqtordb-for-apps.fly.dev", timeout=30.0)

• api_key — your app's secret key (sent as the X-API-Key header). Required.
• base_url — the API host. Defaults to the hosted API; override for self-hosted/staging.

Methods

Method	Endpoint	Description
add(end_user_id, text, source_type="chat")	POST /v1/memory	Store a memory for one user.
search(query, end_user_id=None, limit=10)	POST /v1/search	3-layer retrieval (semantic + keyword + graph), RRF-fused. Omit end_user_id to search the whole app.
list_users()	GET /v1/users	List the app's end users.
get_user(end_user_id)	GET /v1/users/{id}	Full detail: briefing, memories, entities, conversations.
delete_user(end_user_id) / forget(...)	DELETE /v1/users/{id}	GDPR cascade erasure for one user.
preload(end_user_id, notes=[...], ...)	POST /v1/users/{id}/profile	Pre-load a user's memory before they ever chat (CSV/CRM import).

Pre-load (the "known on message one" pattern)

c.preload(
    end_user_id="jordan_cfo",
    notes=["Plan: Enterprise", "Company: Acme Robotics", "Renewal: 2026-03-15"],
    entities=[{"name": "Acme Robotics", "type": "COMPANY"}],
    topics=["enterprise", "onboarding"],
)

The user now has a populated Deep Memory briefing and searchable history — so the very first time they message your bot, it already knows them. Supply briefing=... to set the profile verbatim (no LLM call); omit it and one is synthesized from the notes.

Async

Same surface, for asyncio apps and agents:

import asyncio
from praqtordb_apps import AsyncClient

async def main():
    async with AsyncClient(api_key="pk_live_...") as c:
        await c.add(end_user_id="sarah", text="On the Pro plan")
        print(await c.search(query="plan", end_user_id="sarah"))

asyncio.run(main())

Errors

from praqtordb_apps import PraqtorAPIError, PraqtorError

try:
    c.get_user("nobody")
except PraqtorAPIError as e:
    print(e.status_code, e.detail)   # e.g. 404 "unknown end_user_id"
except PraqtorError as e:
    print("network/transport error:", e)

Requirements

Python 3.9+, httpx.