JSON

JSON file storage

JSON

Simple JSON file-based storage for development.

Quick Start

from praisonaiagents import Agent

agent = Agent(
    name="Assistant",
    instructions="You are a helpful assistant.",
    memory={
        "backend": "file",
        "session_id": "my-session"
    }
)

response = agent.start("Hello!")
print(response)

File Location

By default, JSON files are stored in:

• ~/.praisonai/memory/ for memory data
• ~/.praisonai/sessions/ for session data

When to Use

✅ Good for:

• Local development
• Testing
• Quick prototyping

❌ Not recommended for:

• Production deployments
• Multi-user applications
• Large datasets

Storage Backend (Advanced)

For training data, sessions, and general persistence, use the FileBackend:

BaseJSONStore

from praisonaiagents.storage import FileBackend, BaseJSONStore

backend = FileBackend(storage_dir="~/.praisonai/data")
store = BaseJSONStore("session.json", backend=backend)
store.save({"messages": ["Hello"]})
data = store.load()

BaseJSONStore.save() writes atomically: on success the staging .tmp is renamed over the real file; on serialization or encoding failure the original error propagates, the previous store is left intact, and the staging file is cleaned up. See Thread Safety → BaseJSONStore atomic save.

TrainingStorage is a context manager. FileBackend holds no open connection, so .close() is a safe no-op here — but using with keeps the pattern identical to the SQLite and Redis backends, where closing releases a file or network handle (essential on Windows and for long-running processes).

FileBackend.load() returns None on corrupted files (invalid UTF-8, malformed JSON, unreadable) and leaves the bytes untouched on disk. The same instance then refuses an ordinary save() on that key until you call backend.delete(key), the source is removed, or a subsequent load() succeeds — so a transient-looking fallback never silently overwrites forensic evidence. See Error Handling and Save-side guard for details.

CLI Usage

# Training with file backend (default)
praisonai train agents --input "Hello" --storage-backend file --storage-path ~/.praisonai/train

# Session list with file backend
praisonai session list --storage-backend file --storage-path ~/.praisonai/sessions

See Storage Backends for more details.