Surreal Commands

A distributed task queue system similar to Celery, built with Python, SurrealDB, and LangChain. This system allows you to define, submit, and execute asynchronous commands/tasks with real-time processing capabilities.

Features

• Real-time Processing: Uses SurrealDB LIVE queries for instant command pickup
• Concurrent Execution: Configurable concurrent task execution with semaphore controls
• Type Safety: Pydantic models for input/output validation
• LangChain Integration: Commands are LangChain Runnables for maximum flexibility
• Dynamic CLI: Auto-generates CLI from registered commands
• Status Tracking: Track command status through lifecycle (new � running � completed/failed)
• Persistent Queue: Commands persist in SurrealDB across worker restarts
• Comprehensive Logging: Built-in logging with loguru

Architecture Overview

graph TD
    CLI[CLI Interface] --> SurrealDB[(SurrealDB Queue)]
    SurrealDB --> Worker[Worker Process]
    Worker --> Registry[Command Registry]
    Registry --> Commands[Registered Commands]
    Worker --> |Execute| Commands
    Commands --> |Results| SurrealDB

Installation

Install the package using pip:

pip install surreal-commands

Set up environment variables in .env:

# SurrealDB Configuration
SURREAL_URL=ws://localhost:8000/rpc
SURREAL_USER=root
SURREAL_PASSWORD=root
SURREAL_NAMESPACE=test
SURREAL_DATABASE=test

4. Ensure SurrealDB is running:

# Using Docker
docker run --rm -p 8000:8000 surrealdb/surrealdb:latest start --user root --pass root

# Or install locally
# See: https://surrealdb.com/install

Quick Start

1. Define Commands

Create your commands using the @command decorator:

# my_app/tasks.py
from surreal_commands import command, submit_command
from pydantic import BaseModel

class ProcessInput(BaseModel):
    message: str
    uppercase: bool = False

class ProcessOutput(BaseModel):
    result: str
    length: int

@command("process_text")  # Auto-detects app name as "my_app"
def process_text(input_data: ProcessInput) -> ProcessOutput:
    result = input_data.message.upper() if input_data.uppercase else input_data.message
    return ProcessOutput(result=result, length=len(result))

# Alternative: explicit app name
@command("analyze", app="analytics")
def analyze_data(input_data: ProcessInput) -> ProcessOutput:
    return ProcessOutput(result=f"Analyzed: {input_data.message}", length=len(input_data.message))

2. Submit and Monitor Commands

from surreal_commands import submit_command, wait_for_command_sync

# Submit a command
cmd_id = submit_command("my_app", "process_text", {
    "message": "hello world", 
    "uppercase": True
})

print(f"Submitted command: {cmd_id}")

# Wait for completion
result = wait_for_command_sync(cmd_id, timeout=30)
if result.is_success():
    print(f"Result: {result.result}")

3. Start the Worker

# Start the worker process (import modules from environment variable)
export SURREAL_COMMANDS_MODULES="tasks"
surreal-commands-worker

# Or specify modules directly via CLI
surreal-commands-worker --import-modules "tasks"

# With debug logging
surreal-commands-worker --debug --import-modules "tasks"

# With custom concurrent task limit
surreal-commands-worker --max-tasks 10 --import-modules "tasks"

# Import multiple modules
surreal-commands-worker --import-modules "tasks,my_app.commands"

4. Monitor with CLI Tools

# View command dashboard
surreal-commands-dashboard

# View real-time logs
surreal-commands-logs

Library Structure

surreal-commands/
├── apps/                    # Your command applications
│   └── text_utils/         # Example app
│       ├── __init__.py
│       └── commands.py     # Command definitions
├── cli/                    # CLI components
│   ├── __init__.py
│   ├── launcher.py        # Dynamic CLI generator
│   ├── dashboard.py       # (Future) Dashboard UI
│   └── logs.py           # (Future) Log viewer
├── commands/              # Core command system
│   ├── __init__.py
│   ├── command_service.py # Command lifecycle management
│   ├── executor.py       # Command execution engine
│   ├── loader.py         # Command discovery
│   ├── registry.py       # Command registry (singleton)
│   ├── registry_types.py # Type definitions
│   └── worker.py         # Worker process
├── repository/           # Database layer
│   └── __init__.py      # SurrealDB helpers
├── cli.py               # CLI entry point
├── run_worker.py        # Worker entry point
└── .env                 # Environment configuration

Core Components

Command Registry

• Singleton pattern for global command management
• Stores commands as LangChain Runnables
• Organizes commands by app namespace

Command Service

• Manages command lifecycle
• Validates arguments against schemas
• Updates command status in real-time

Worker

• Long-running process polling SurrealDB
• Processes existing commands on startup
• Listens for new commands via LIVE queries
• Configurable concurrency limits

Executor

• Handles sync/async command execution
• Type conversion and validation
• Streaming support for long-running tasks

Advanced Usage

Custom Command with Complex Types

from typing import List, Optional
from datetime import datetime
from pydantic import BaseModel, Field

class AnalysisInput(BaseModel):
    data: List[float]
    method: str = Field(default="mean", description="Analysis method")
    threshold: Optional[float] = None

class AnalysisOutput(BaseModel):
    result: float
    method_used: str
    items_processed: int
    warnings: List[str] = []

def analyze_data(input_data: AnalysisInput) -> AnalysisOutput:
    # Your analysis logic here
    pass

Async Commands

async def async_process(input_data: MyInput) -> MyOutput:
    # Async processing
    await some_async_operation()
    return MyOutput(...)

# LangChain handles both sync and async
command = RunnableLambda(async_process)

Working with Execution Context

Commands can access execution metadata (command_id, execution time, etc.) using the CommandInput and CommandOutput base classes. This is the recommended approach that works with all registration methods.

Using CommandInput and CommandOutput

from surreal_commands import command, CommandInput, CommandOutput, ExecutionContext

# Input that can access execution context
class ProcessInput(CommandInput):
    message: str
    uppercase: bool = False

# Output that includes execution metadata
class ProcessOutput(CommandOutput):
    result: str
    # command_id, execution_time, and execution_metadata are inherited

@command("process_with_context")
def process_with_context(input_data: ProcessInput) -> ProcessOutput:
    # Access execution context from input
    ctx = input_data.execution_context
    
    if ctx:
        command_id = ctx.command_id
        app_name = ctx.app_name
        user_context = ctx.user_context or {}
        user_id = user_context.get("user_id", "anonymous")
    else:
        command_id = "unknown"
        user_id = "anonymous"
    
    # Process the message
    result = input_data.message.upper() if input_data.uppercase else input_data.message
    result = f"Processed by {user_id}: {result}"
    
    # Return output - framework automatically populates:
    # - command_id (from execution context)
    # - execution_time (measured by framework)
    # - execution_metadata (additional context info)
    return ProcessOutput(result=result)

Benefits of the New Pattern

• Works with all registration methods (decorator and direct registry.register())
• Type-safe with full IDE support
• Automatic metadata population in outputs
• Backward compatible - existing commands continue to work
• Clean API that aligns with LangChain's expectations

Base Class Usage Options

1. CommandInput only: Access execution context in your command
2. CommandOutput only: Get automatic execution metadata in outputs
3. Both: Full execution context support with automatic metadata
4. Neither: Regular commands without execution context (backward compatible)

# Option 1: Access context only
class MyInput(CommandInput):
    data: str

@command("context_input_only")
def my_command(input_data: MyInput) -> RegularOutput:
    ctx = input_data.execution_context
    # ... use context ...
    return RegularOutput(result="done")

# Option 2: Output metadata only
class MyOutput(CommandOutput):
    result: str

@command("context_output_only")
def my_command(input_data: RegularInput) -> MyOutput:
    # ... process ...
    return MyOutput(result="done")  # Gets auto-populated metadata

# Option 3: Full context support
class MyInput(CommandInput):
    data: str

class MyOutput(CommandOutput):
    result: str

@command("full_context")
def my_command(input_data: MyInput) -> MyOutput:
    # ... access context and get metadata ...

ExecutionContext Properties

• command_id: Database ID of the command
• execution_started_at: When execution began
• app_name: Application name
• command_name: Command name
• user_context: Optional CLI context (user_id, scope, etc.)

CommandOutput Auto-populated Fields

• command_id: Automatically set from execution context
• execution_time: Measured execution time in seconds
• execution_metadata: Additional execution information

Migration from Old Pattern

If you have commands using the old pattern with execution_context as a parameter:

# OLD (causes errors):
def old_command(input_data: MyInput, execution_context: ExecutionContext):
    command_id = execution_context.command_id
    # ...

# NEW (recommended):
class MyInput(CommandInput):
    # your fields here

def new_command(input_data: MyInput):
    command_id = input_data.execution_context.command_id if input_data.execution_context else None
    # ...

See examples/migration_example.py for detailed migration examples.

Monitoring

Check Command Status

# check_results.py
import asyncio
from repository import db_connection

async def check_status():
    async with db_connection() as db:
        commands = await db.query(
            "SELECT * FROM command ORDER BY created DESC LIMIT 10"
        )
        for cmd in commands:
            print(f"{cmd['id']}: {cmd['status']}")

asyncio.run(check_status())

View Logs

# Worker logs with debug info
uv run python run_worker.py --debug

# Filter logs by level
LOGURU_LEVEL=INFO uv run python run_worker.py

Database Schema

Commands are stored in SurrealDB with the following structure:

{
  id: "command:unique_id",
  app: "app_name",
  name: "command_name",
  args: { /* command arguments */ },
  context: { /* optional context */ },
  status: "new" | "running" | "completed" | "failed",
  result: { /* command output */ },
  error_message: "error details if failed",
  created: "2024-01-07T10:30:00Z",
  updated: "2024-01-07T10:30:05Z"
}

Development

Adding New Commands

1. Create a new app directory under apps/
2. Define your command models and logic
3. Register with the command registry
4. Restart the worker to pick up new commands

Running Tests

# Run tests (when implemented)
uv run pytest

# With coverage
uv run pytest --cov=commands

Debugging

Use the debug mode to see detailed logs:

# Debug CLI
LOGURU_LEVEL=DEBUG uv run python cli.py text_utils uppercase --text "test"

# Debug Worker
uv run python run_worker.py --debug

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

Retry Logic

Surreal Commands includes built-in retry logic using the tenacity library. Commands can be configured with automatic retry behavior for handling transient failures.

Basic Retry Configuration

from surreal_commands import command

# Simple retry with 3 attempts
@command("process_data", retry={"max_attempts": 3, "wait_strategy": "fixed", "wait_time": 1})
def process_data(input_data: MyInput) -> MyOutput:
    # This command will retry up to 3 times with 1 second wait between attempts
    return MyOutput(result="processed")

Wait Strategies

Choose the right wait strategy for your use case:

# Fixed wait time (good for rate-limited APIs)
@command("api_call", retry={"max_attempts": 3, "wait_strategy": "fixed", "wait_time": 2})
def api_call(input_data: MyInput) -> MyOutput:
    pass

# Exponential backoff (recommended for network calls)
@command("network_call", retry={
    "max_attempts": 5,
    "wait_strategy": "exponential",
    "wait_min": 1,
    "wait_max": 60,
    "wait_multiplier": 2
})
def network_call(input_data: MyInput) -> MyOutput:
    # Wait times: 1s, 2s, 4s, 8s, 16s (capped at 60s)
    pass

# Exponential with jitter (prevents thundering herd)
@command("distributed_call", retry={"wait_strategy": "exponential_jitter"})
def distributed_call(input_data: MyInput) -> MyOutput:
    pass

Exception Filtering

Control which exceptions trigger retries:

# Retry only on specific exceptions
@command("selective_retry", retry={
    "max_attempts": 3,
    "retry_on": [ConnectionError, TimeoutError]
})
def selective_retry(input_data: MyInput) -> MyOutput:
    pass

# Don't retry on permanent errors
@command("skip_permanent", retry={
    "max_attempts": 3,
    "stop_on": [ValueError, TypeError, KeyError]
})
def skip_permanent(input_data: MyInput) -> MyOutput:
    pass

Controlling Retry Log Verbosity

Control the log level for retry attempt messages to reduce noise in high-concurrency scenarios:

# Reduce logging for bulk operations with expected transient failures
@command("bulk_embed", retry={
    "max_attempts": 5,
    "wait_strategy": "exponential",
    "retry_log_level": "debug"  # Only show retries in debug mode
})
def bulk_embed(input_data: MyInput) -> MyOutput:
    pass

# Suppress all retry logs for high-volume background tasks
@command("background_sync", retry={
    "max_attempts": 3,
    "retry_log_level": "none"  # No retry logging
})
def background_sync(input_data: MyInput) -> MyOutput:
    pass

# Use warning level for critical operations
@command("critical_api_call", retry={
    "max_attempts": 3,
    "retry_log_level": "warning"  # Make retries more visible
})
def critical_api_call(input_data: MyInput) -> MyOutput:
    pass

Available log levels: "debug", "info" (default), "warning", "error", "none"

Global Retry Configuration

Set default retry behavior via environment variables:

# Enable retries globally
SURREAL_COMMANDS_RETRY_ENABLED=true
SURREAL_COMMANDS_RETRY_MAX_ATTEMPTS=3
SURREAL_COMMANDS_RETRY_WAIT_STRATEGY=exponential
SURREAL_COMMANDS_RETRY_WAIT_MIN=1
SURREAL_COMMANDS_RETRY_WAIT_MAX=60
SURREAL_COMMANDS_RETRY_LOG_LEVEL=info

Per-command configuration always overrides global defaults.

Disabling Retry

# Explicitly disable retry for a command
@command("no_retry", retry=None)
def no_retry(input_data: MyInput) -> MyOutput:
    # Fails immediately without retry
    pass

# Or use enabled flag
@command("also_no_retry", retry={"enabled": False})
def also_no_retry(input_data: MyInput) -> MyOutput:
    pass

Type-Safe Configuration

Use RetryConfig for IDE autocomplete and validation:

from surreal_commands import command, RetryConfig, RetryStrategy, RetryLogLevel

config = RetryConfig(
    enabled=True,
    max_attempts=5,
    wait_strategy=RetryStrategy.EXPONENTIAL,
    wait_min=1,
    wait_max=60,
    retry_log_level=RetryLogLevel.DEBUG  # Control log verbosity
)

@command("typed_retry", retry=config)
def typed_retry(input_data: MyInput) -> MyOutput:
    pass

See examples/retry_examples.py for more examples.

Future Enhancements

• Web dashboard for monitoring
• Command scheduling (cron-like)
• Priority queues
• Result callbacks
• Retry mechanisms (completed!)
• Command chaining/workflows
• Metrics and monitoring
• REST API endpoint
• Command result TTL
• Dead letter queue

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Inspired by Celery's design patterns
• Built with SurrealDB for real-time capabilities
• Leverages LangChain for flexible command definitions