Skip to content

On-chain deployment checklist

This section details a requirement checklist that need to be satisfied so that your AI agent can be properly deployed on-chain.

Required FSM App skill models

Ensure that your skill extends the following classes in the models.py file:

packages.valory.skills.abstract_round_abci.models.BenchmarkTool
packages.valory.skills.abstract_round_abci.models.BaseParams

You can define custom arguments for the skill, if required.

from packages.valory.skills.abstract_round_abci.models import BenchmarkTool
from packages.valory.skills.abstract_round_abci.models import BaseParams

# (...)

YourBenchmarkTool = BenchmarkTool
YourSkillParams = BaseParams

from packages.valory.skills.abstract_round_abci.models import BenchmarkTool
from packages.valory.skills.abstract_round_abci.models import BaseParams

# (...)

YourBenchmarkTool = BenchmarkTool

class YourSkillParams(BaseParams):
    def __init__(self, *args: Any, **kwargs: Any) -> None:
        self.your_custom_arg_1: str = self._ensure("your_custom_arg_1", kwargs, str)
        self.your_custom_arg_2: str = self._ensure("your_custom_arg_2", kwargs, str)            

        # (...)

        super().__init__(*args, **kwargs)

Required arguments and overrides

Ensure that your FSM App skill, agent blueprint and AI agent configuration files (skill.yaml, aea-config.yaml, and service.yaml, respectively) define the appropriate arguments and overrides:

skill.yaml
Must define default/placeholder values for the arguments associated to the YourSkillParams class.
aea-config.yaml
Must define overrides for valory/abci connection, valory/ledger connection, valory/p2p_libp2p_client connection, and your FSM App skill. Environment variables used for agent-blueprint-level overrides can use the simplified syntax ${<type>:<default_value>}.
service.yaml
Must define overrides for valory/ledger connection and your FSM App skill (optionally, also for valory/p2p_libp2p_client connection). Environment variables used for ai-agent-level overrides use the syntax ${<env_var_name>:<type>:<default_value>}. They will be exported as their upper case JSON path in the agent blueprint Docker container. See also the AI agent level overrides and multiple overrides sections for more information.
name: <your_skill_name>
author: <author>
version: <version>
type: skill    
# (...)
models:
  benchmark_tool:
    args:
      log_dir: /logs
    class_name: YourBenchmarkTool

  params:
    args:
      setup:
        all_participants:
        - '0x0000000000000000000000000000000000000000'
        safe_contract_address: '0x0000000000000000000000000000000000000000'
        consensus_threshold: null            
      tendermint_url: http://localhost:26657
      tendermint_com_url: http://localhost:8080
      service_registry_address: null
      share_tm_config_on_startup: false
      on_chain_service_id: null
      # (...)
    class_name: YourSkillParams

# (...)
---
public_id: valory/abci:0.1.0
type: connection
config:
  target_skill_id: <author>/<your_skill_name>:<version>
  host: ${ABCI_HOST:str:localhost}
  port: ${ABCI_PORT:int:26658}
  use_tendermint: ${ABCI_USE_TENDERMINT:bool:false}

---
public_id: valory/ledger:0.19.0
type: connection
config:
  ledger_apis:
    ethereum:
      address: ${str:http://localhost:8545}
      chain_id: ${int:31337}
      poa_chain: ${bool:false}
      default_gas_price_strategy: ${str:eip1559}

---
public_id: valory/p2p_libp2p_client:0.1.0
type: connection
config:
  nodes:
 - uri: ${P2P_URI:str:acn.autonolas.tech:9005}
   public_key: ${P2P_PUBLIC_KEY:str:02d3a830c9d6ea1ae91936951430dee11f4662f33118b02190693be835359a9d77}
 - uri: ${P2P_URI:str:acn.autonolas.tech:9006}
   public_key: ${P2P_PUBLIC_KEY:str:02e741c62d706e1dcf6986bf37fa74b98681bc32669623ac9ee6ff72488d4f59e8}
cert_requests:
- identifier: acn
  ledger_id: ethereum
  message_format: '{public_key}'
  not_after: '2025-01-01'
  not_before: '2024-01-01'
  public_key: ${P2P_PUBLIC_KEY:str:02d3a830c9d6ea1ae91936951430dee11f4662f33118b02190693be835359a9d77}
  save_path: .certs/acn_cosmos_9005.txt

---
public_id: <author>/<your_skill_name>:<version>
type: skill    
models:
  benchmark_tool:
    args:
      log_dir: ${str:/benchmarks}

  params:
    args:
      setup:
        all_participants: ${list:[]}
        safe_contract_address: ${str:'0x0000000000000000000000000000000000000000'}
        consensus_threshold: ${int:null}
      tendermint_url: ${str:http://localhost:26657}
      tendermint_com_url: ${str:http://localhost:8080}
      service_registry_address: ${str:null}
      share_tm_config_on_startup: ${bool:false}
      on_chain_service_id: ${int:null}
      # (...)

# (...)
---
public_id: valory/ledger:0.19.0
type: connection
config:
  ledger_apis:
    ethereum:
      address: ${ETHEREUM_ADDRESS:str:http://localhost:8545}
      chain_id: ${ETHEREUM_CHAIN_ID:int:31337}
      poa_chain: ${ETHEREUM_POA_CHAIN:bool:false}
      default_gas_price_strategy: ${DEFAULT_GAS_PRICE_STRATEGY:str:eip1559}

---
public_id: <author>/<your_skill_name>:<version>
type: skill    
models:
  benchmark_tool:
    args:
      log_dir: ${LOG_DIR:str:/benchmarks}

  params:
    args:
      setup:
        all_participants: ${ALL_PARTICIPANTS:list:["0x...","0x...","0x...","0x..."]}      
        safe_contract_address: ${SAFE_CONTRACT_ADDRESS:str:0x...}
        consensus_threshold: ${CONSENSUS_THRESHOLD:int:null}
      tendermint_url: ${str:http://localhost:26657}
      tendermint_com_url: ${str:http://localhost:8080}
      service_registry_address: ${SERVICE_REGISTRY_ADDRESS:str:0x...}
      share_tm_config_on_startup: ${SHARE_TM_CONFIG_ON_STARTUP:bool:false}
      on_chain_service_id: ${ON_CHAIN_SERVICE_ID:int:1}
      # (...)

Important

Recall that when deploying an on-chain AI agent using autonomy deploy from-token, a number of arguments (under setup) are overridden with the values registered in the Autonolas Protocol:

service.yaml
# (...)
models:
  params:
    args:
      setup:
        all_participants: # Overridden with the registered values
        safe_contract_address: # Overridden with the registered values
        consensus_threshold: # Overridden with the registered values

For local deployments, the argument consensus_threshold can take the value:

  • null: then the framework will automatically calculate consensus_threshold as \(\lceil (2N+1)/3 \rceil\), where \(N=\)len(all_participants).
  • Any value \(M\) such that \(\lceil (2N+1)/3 \rceil \leq M \leq N\).

Otherwise, the framework will raise an error and the app will not start.

Publish and mint packages

Ensure that your components, agent blueprints and AI agents packages are published to the IPFS registry:

Ensure that your components, agent blueprints and AI agents packages are minted on-chain in the Autonolas Protocol.

Check the deployment readiness of the AI agent using

$ autonomy analyse service --public-id PUBLIC_ID

or if you want to check deployment readiness of an on-chain AI agent

$ autonomy analyse service --token-id TOKEN_ID

Publish Docker images (optional)

You can build the Docker images for the AI agent using the autonomy build-image command. Alternatively, the images are built automatically when the AI agent is deployed using autonomy deploy from-token command

If you want to use an image with a stable hash or a stable version of a runtime image, you can provide the hash/version using --image-version on the autonomy deploy build command.

Ensure that the image exists before running the deployment:

docker pull <author>/oar_runtime_<ai_agent_name>:<ipfs_hash_ai_agent>