Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
Python agentagentic-aiagentic-frameworkagentic-workflowaiautomationclaudedeepseekgenaigptllmlow-codemcpnextjsno-codeopenaipythonskillsworkflow
from base64 import b64decode, b64encode
from typing import Protocol
from core.human_input_v2.entities import IMProvider
from core.human_input_v2.im_integration.adapters.credentials import (
IMProviderCredentials,
IMProviderCredentialsAdapter,
)
from repositories.human_input_v2.channel_credentials import ChannelEncryptedCredentials
_SAFE_ERROR = "IM credential configuration is unavailable"
class BoundCredentialCipher(Protocol):
"""A credential cipher bounded to a tenant or the whole deployment."""
def encrypt(self, plaintext: str) -> bytes: ...
def decrypt(self, ciphertext: bytes) -> str: ...
class IMCredentialError(ValueError): ...
class IMCredentialCodec:
def __init__(self, cipher: BoundCredentialCipher) -> None:
self._cipher = cipher
def seal(self, credentials: IMProviderCredentials) -> ChannelEncryptedCredentials:
encrypted = self._cipher.encrypt(credentials.model_dump_json())
return ChannelEncryptedCredentials(ciphertext=b64encode(encrypted).decode())
def load(
self,
provider: IMProvider,
envelope: ChannelEncryptedCredentials,
) -> IMProviderCredentials:
if envelope.version != 1:
raise IMCredentialError(_SAFE_ERROR)
try:
decrypted = self._cipher.decrypt(b64decode(envelope.ciphertext))
credentials = IMProviderCredentialsAdapter.validate_json(decrypted)
except Exception as exc:
raise IMCredentialError(_SAFE_ERROR) from exc
if credentials.provider is not provider:
raise IMCredentialError(_SAFE_ERROR)
return credentials