signal-fabric-client 0.1.28

Client library for Signal Fabric

Signal Fabric Client

Official Python client library for Signal Fabric - a lightweight handler-based framework for generating market signals on demand.

Installation

pip install signal-fabric-client

Quick Start

Live Signal Request

from signal_fabric import GrpcClient

with GrpcClient(host='localhost', port=9090, use_tls=False) as client:
    outcome = client.process_signal(
        target='BTCUSDT',
        signal_name='rsi_binance_spot',
        signal_op='compute_rsi',
        handler_request={'period': 14, 'timeframe': '1h'}
    )

    if outcome.has_errors():
        print(f"Error: {outcome.errors}")
    else:
        print(f"RSI: {outcome.result['latest_rsi']}, Regime: {outcome.result['regime']}")

Backtest Replay Mode

from signal_fabric import GrpcClient

with GrpcClient(host='localhost', port=9090, use_tls=False) as client:
    # Get dataset metadata
    metadata = client.get_backtest_metadata(
        signal_name='rsi_binance_mf',
        target='BTCUSDT',
        timeframe='1h'
    )

    if metadata.has_errors():
        print(f"Error: {metadata.errors}")
    else:
        print(f"Total positions: {metadata.total_positions}")
        print(f"Start position: {metadata.start_position} (after {metadata.required_lookback} lookback)")

        # Iterate through positions for walk-forward backtest
        for position, timestamp in metadata.iter_positions():
            outcome = client.process_signal(
                target='BTCUSDT',
                signal_name='rsi_binance_mf',
                signal_op='compute_rsi',
                handler_request={
                    '_position': position,
                    'timeframe': '1h',
                    'period': 14
                }
            )
            # Evaluate outcome against known historical data...

API Reference

GrpcClient

Constructor

GrpcClient(host: str = 'localhost', port: int = 50051, timeout: int = 30, ca_cert_path: str = None)

Parameters:

• host (str): Server hostname or IP address (default: 'localhost')
• port (int): Server port number (default: 50051)
• timeout (int): Request timeout in seconds (default: 30)
• ca_cert_path (str): Path to CA certificate for server verification (PEM format)

Methods

connect()

Establish connection to the server.

disconnect()

Close the connection to the server.

is_connected() -> bool

Check if client is currently connected.

process_signal(target, signal_name, signal_op, handler_request=None) -> SignalOutcome

Process a signal request.

Parameters:

• target (str): Target for signal computation (e.g., 'BTC', 'ETH', 'AAPL')
• signal_name (str): Signal handler name or profile name
• signal_op (str): Operation to perform (e.g., 'analyze', 'greet')
• handler_request (dict, optional): Request parameters as dictionary

Returns: SignalOutcome object

get_backtest_metadata(signal_name, target, timeframe) -> BacktestMetadata

Get metadata for backtest dataset (positions, timestamps, lookback).

Parameters:

• signal_name (str): Signal profile name (e.g., 'rsi_binance_mf')
• target (str): Target symbol (e.g., 'BTCUSDT')
• timeframe (str): Timeframe string (e.g., '1h', '5m')

Returns: BacktestMetadata object

SignalOutcome

Result object containing the signal computation outcome.

Attributes

• result (str): Signal result value
• computation (str): Description of computation performed
• computed_at (float): Unix timestamp when computed
• errors (Dict[str, str]): Error codes and messages (or empty)
• details (Dict[str, str]): Additional computation details (or empty)

Methods

has_errors() -> bool

Returns True if the outcome contains errors.

is_detailed() -> bool

Returns True if the outcome has errors or additional details.

BacktestMetadata

Metadata for backtest dataset, used for walk-forward backtesting.

Attributes

• total_positions (int): Total number of candles in dataset
• start_position (int): First valid position (after handler lookback)
• required_lookback (int): Handler's required lookback period
• timeframe_ms (int): Timeframe interval in milliseconds
• first_timestamp (int): First candle timestamp (ms since epoch)
• last_timestamp (int): Last candle timestamp (ms since epoch)
• symbol (str): Confirmed symbol from dataset
• timeframe_str (str): Confirmed timeframe string
• errors (Dict[str, str]): Error messages (or empty)

Methods

has_errors() -> bool

Returns True if metadata request returned errors.

get_timestamp(position: int) -> int

Calculate timestamp for a given position using formula: first_timestamp + (position * timeframe_ms)

iter_positions() -> Iterator[Tuple[int, int]]

Iterate through valid backtest positions. Yields (position, timestamp) tuples.

Requirements

• Python 3.8+
• grpcio >= 1.76.0
• protobuf >= 4.0.0

Server Setup

This client requires a running Signal Fabric server. To set up the server:

1. Clone the Signal Fabric repository:

   git clone https://github.com/phasequant/signal-fabric.git
   cd signal-fabric
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Start the server:

   python src/server/main.py --config config.yaml
   

Version

Current version: 0.1.28

License

See LICENSE file for details.

Links

• Signal Fabric Repository
• Documentation
• Issue Tracker

Support

For questions, issues, or contributions, please visit the GitHub repository.