forta-toolkit 0.4.5

Tools to help with the common problems of Forta bot development

Forta Toolkit

Various tools to help with the common problems of Forta bot development.

Table of Contents

• Installation
• Usage
  • Alert statistics
  • Logging execution events
  • Indexing
  • Preprocessing
  • Improving performances
  • Load balancing
  • Profiling
  • Recommendation And Warning
• Development
  • Changelog
  • Todo
• Credits
• License

Installation

# globally
pip install forta_toolkit

# in a local environment
poetry add forta_toolkit

Usage

Bot setup

The Forta often require initialization steps to adapt to a given chain or use external tools.

The toolkit reads the OS environment and local files to load the settings:

import forta_toolkit.parsing.env

forta_toolkit.parsing.env.get_bot_version() # reads "package.json" in the parent directory
forta_toolkit.parsing.env.load_secrets() # read the file "secrets.json", in the parent directory
forta_toolkit.parsing.env.load_chain_id(provider=w3) # load the chain_id from the env variables or query the provider if it is not set

Alert statistics

This is an alternative to querying the Zetta API for alert statistics. It saves a local history of the alerts in memory and use it to calculate the rates. The main motivation is to improve performance by avoiding web requests.

To use it, just wrap handle_block / handle_transaction / handle_alert as follows:

import forta_toolkit

@forta_toolkit.alerts.alert_history(size=10000)
def handle_block(log: BlockEvent) -> list:
    pass

@forta_toolkit.alerts.alert_history(size=10000)
def handle_transaction(log: TransactionEvent) -> list:
    pass

@forta_toolkit.alerts.alert_history(size=10000)
def handle_alert(log: AlertEvent) -> list:
    pass

The decorator will automatically add the anomaly_score in the metadata of the Finding objects. It will use the field alert_id from the Finding objects to identify them.

make sure the history size is big enough to contain occurences of the bot alerts!

For example, if your bot triggers ALERT-1 every 2k transactions and ALERT-2 every 10k on average: @alert_history(size=100000) would gather enough alerts to have a relevant estimation of the rate of both alerts.

Parsing logs / traces / transactions

The forta-agent returns slightly different objects compared to a direct query on a RPC endpoint.

The parsing functions convert these objects into plain dictionaries, with only HEX string data. Instead of a mix of bytes, HexBytes, str, int with irregular formats.

Transactions are represented like the following:

{
    'block_number': '0x00',
    'chain_id': '0x01',
    'from_address': '0x584Df055A35acf2d2183d37fC7F42cb4c502Dc51',
    'gas_limit': '0x00',
    'gas_price': '0x0a68fde78f',
    'gas_used': '0x035672',
    'input': '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',
    'max_fee_per_gas': '0x00',
    'max_priority_fee_per_gas': '0x00',
    'nonce': '0x3863',
    'success': '0x01',
    'to_address': '0x1D2B04F5008295918BA9c78B1e664fEE8b444007',
    'transaction_hash': '0x030ceaef086d4f8cb8d19ac6f6e0d38522e731db357b6e73a2f7642cf0c9e7fa',
    'transaction_index': '0x00',
    'transaction_type': '0x02',
    'value': '0x00'}

Each log is like:

{
    'address': '0x7262a43c94258e6071d3bAC00eBeC6d7c9B4Ef30',
    'blockHash': '0x27fad9215f66a8d3c1d5bd4009bb0af5f3bad76abcf11aadabb2586b33aca837',
    'blockNumber': '0x011cca51',
    'block_hash': '0x27fad9215f66a8d3c1d5bd4009bb0af5f3bad76abcf11aadabb2586b33aca837',
    'block_number': '0x011cca51',
    'chain_id': '0x01',
    'data': '0x000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000006d38666f93bbc74000000000000000000000000000000000000000000032de60fd4b32140f5447b0000000000000000000000000000000000000000000000000000000000000000',
    'logIndex': '0x8c',
    'log_index': '0x8c',
    'topics': [
        b"\xd7\x8a\xd9_\xa4l\x99KeQ\xd0\xda\x85\xfc'_\xe6\x13\xce7"
        b'e\x7f\xb8\xd5\xe3\xd10\x84\x01Y\xd8"',
        b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x1d+\x04\xf5'
        b'\x00\x82\x95\x91\x8b\xa9\xc7\x8b\x1efO\xee\x8bD@\x07',
        b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00$o]\xf4'
        b'z\xf6\xc4\x07\xe4\x14\x997\xfb\xddt\xbdB8\xbb\xb5'],
    'transactionHash': '0x030ceaef086d4f8cb8d19ac6f6e0d38522e731db357b6e73a2f7642cf0c9e7fa',
    'transactionIndex': '0x2c',
    'transaction_hash': '0x030ceaef086d4f8cb8d19ac6f6e0d38522e731db357b6e73a2f7642cf0c9e7fa',
    'transaction_index': '0x2c'}

The key aliases in camel-case are meant to satisfy the web3 package expectations.

The topics are the only data encoded as bytes instead of HEX strings.

And finally the each trace looks like this call:

{
    'action_call_type': 'staticcall',
    'action_from': '0x1D2B04F5008295918BA9c78B1e664fEE8b444007',
    'action_gas': '0x00',
    'action_init': '0x',
    'action_input': '0x70a08231000000000000000000000000246f5df47af6c407e4149937fbdd74bd4238bbb5',
    'action_reward_type': '',
    'action_to': '0x41EA5d41EEACc2D5c4072260945118a13bb7EbCE',
    'action_type': 'call',
    'action_value': '0x00',
    'block_hash': '0x27fad9215f66a8d3c1d5bd4009bb0af5f3bad76abcf11aadabb2586b33aca837',
    'block_number': '0x011cca51',
    'chain_id': '0x01',
    'error': None,
    'result_address': '',
    'result_code': '0x',
    'result_gas_used': '0x0253',
    'result_output': '0x000000000000000000000000000000000000000000356d47e966a8ef1b7c49e3',
    'subtraces': '0x00',
    'trace_address': [0, 8],
    'transaction_hash': '0x030ceaef086d4f8cb8d19ac6f6e0d38522e731db357b6e73a2f7642cf0c9e7fa',
    'transaction_index': '0x2c'}

The keys remain ("from", "to", "input", "output") when the trace type is suicide or create:

{
    'block': 'fe6b17',
    'from': '0x45F50A4aC2c4e636191ADcfBB347Ec2a3079FC02',
    'gas': '',
    'hash': '3bfcc1c5838ee17eec1ddda2f1ff0ac1c1ccdbd30dd520ee41215c54227a847f',
    'input': '5860208158601c335a63aaf10f428752fa158151803b80938091923cf3',
    'output': '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',
    'to': '0x6b75d8AF000000e20B7a7DDf000Ba900b4009A80',
    'type': 'create',
    'value': ''}

Logging execution events

The logging level and message template can be setup with:

import forta_toolkit.logging
import forta_toolkit.parsing.env

forta_toolkit.logging.setup_logger(level=logging.INFO, version=forta_toolkit.parsing.env.get_bot_version())

Which will produce messages with the bot version and log level:

[0.1.17 - INFO] Metamorphism: 0x212728A4567F63e41eCD57A7dc329dbF2081B370 is deploying a factory contract at 0xF20e35e946C95ea4fcdadbEd1d79f28f2B8F44DE

Indexing

Serialization (Pickle)

The input arguments and the output findings can be automatically saved to the disk with:

import forta_toolkit.indexing.dump

@forta_toolkit.indexing.pickle.serialize_io()
def handle_transaction(log: TransactionEvent) -> list:
    pass

The decorator accepts a few optional arguments:

@forta_toolkit.indexing.pickle.serialize_io(arguments=False, results=True, filter=True, compress=False, path='.data/{alert}/{txhash}/')
def handle_transaction(log: TransactionEvent) -> list:
    pass

Database (Parquet)

The blockchain data can be indexed in a database with:

import forta_toolkit.indexing.parquet

@forta_toolkit.indexing.parquet.export_to_database(chain_id=1, dataset='contracts', path='.data/contracts/', chunksize=2**10, compress=True)
def handle_transaction(log: TransactionEvent) -> list:
    pass

Currently, only the contracts dataset is supported. The module will soon cover transaction, logs and traces data too.

The database is saved using the parquet file format and using the library pyarrow.

This historic data can then be imported with:

@forta_toolkit.indexing.parquet.import_from_database(chain_id=CHAIN_ID, dataset='contracts', path='.data/contracts/')

Which would lead to the final declaration:

@forta_toolkit.indexing.parquet.export_to_database(chain_id=1, dataset='contracts', path='.data/contracts/', chunksize=2**10, compress=True)
@forta_toolkit.indexing.parquet.import_from_database(chain_id=CHAIN_ID, dataset='contracts', path='.data/contracts/')
def handle_transaction(log: TransactionEvent, dataset: 'pyarrow._dataset.FileSystemDataset') -> list:
    pass

Note that the dataset is automatically passed as argument to handle_transaction.

Preprocessing

The decorator parse_forta_arguments processes the input TransactionEvent and returns the transaction, logs and traces objects.

These objects are automatically sanitized and parsed into fixed structures and base types (mostly int, HEX str, list and bytes).

import forta_toolkit.preprocessing

@forta_toolkit.preprocessing.parse_forta_arguments
def handle_transaction(transaction: dict, logs: list, traces: list) -> list:
    pass

This decorator can only be placed right above a function with the signature (transaction: dict, logs: list, traces: list) -> list.

Improving performances

Load balancing

Profiling

The bots have to follow the pace of the blockchain, so they need to process transactions relatively quickly.

You can leverage the profiling tools to find the performance bottlenecks in your bots:

from forta_toolkit.profiling import test_performances, display_performances

test_performances(func=handle_transaction, data=some_tx_log)
display_performances(logpath='./test_performances')

Otherwise, you can monitor the performances directly when processing mainnet transactions. Just decorate the handle_block / handle_transaction / handle_alert as follows:

@forta_toolkit.alerts.profile
def handle_transaction(tx: TransactionEvent) -> list:
    pass

Then you can parse the profile logs manually with pstats or:

display_performances(logpath='some/path/to/the/logs/handle_transaction')

Recommendation And Warning

All the above decorators can be mixed and matched.

However the order in which the decorator are composed matters:

@forta_toolkit.profiling.timeit
@forta_toolkit.alerts.alert_history(size=history_size)
@forta_toolkit.preprocessing.parse_forta_arguments
@forta_toolkit.indexing.pickle.serialize_io(arguments=True, results=True)
def handle_transaction(transaction: dict, logs: list, traces: list) -> list:
    pass

In the configuration above, the serialize_io decorator will save each of the transaction, logs and traces objects. However if the decorators were switched:

@forta_toolkit.indexing.pickle.serialize_io(arguments=True, results=True)`
@forta_toolkit.preprocessing.parse_forta_arguments

serialize_io would save to disk the arguments of the function returned by parse_forta_arguments: a single TransactionEvent would be serialized to the disk.

Be wary of this composition and test your setup!

The recommended order is the one written at the start of this section.

Development

Contributions welcome!

Changelog

See CHANGELOG.

Todo

See TODO.

Credits

The RPC request queue was inspired by the TS module forta-helpers by Artem Kovalchuk.

License

Licensed under the aGPL v3.