Migrated from agentspan-ai/agentspan#167 (originally by @v1r3n) as part of the Agentspan → Conductor merge. The Agentspan repo is archived; the agent runtime now lives here.
Currently, all @tool functions get a hardcoded Conductor task definition with retry_count=2 and retry_delay_seconds=2 (set in _default_task_def in runtime.py). Users cannot override this from the SDK.
Add retry_count and retry_delay_seconds as optional parameters on the @tool decorator:
@tool(retry_count=10, retry_delay_seconds=5, retry_policy="linear_backoff|expoential") # similar to task def
def call_flaky_api(query: str) -> str:
...
@tool(retry_count=0) # fail immediately, no retries
def process_payment(amount: float) -> dict:
...These values should be passed through to the Conductor TaskDef when the tool is registered.
Follow same configurations as Conductor TaskDef https://github.com/conductor-oss/conductor/blob/main/common/src/main/java/com/netflix/conductor/common/metadata/tasks/TaskDef.java for the retry policy. Give an easy way to configure with sensible defaults and allow users to configure as required.
Migrated from agentspan-ai/agentspan#167 (originally by @v1r3n) as part of the Agentspan → Conductor merge. The Agentspan repo is archived; the agent runtime now lives here.
Currently, all @tool functions get a hardcoded Conductor task definition with retry_count=2 and retry_delay_seconds=2 (set in _default_task_def in runtime.py). Users cannot override this from the SDK.
Add retry_count and retry_delay_seconds as optional parameters on the @tool decorator:
These values should be passed through to the Conductor TaskDef when the tool is registered.
Follow same configurations as Conductor TaskDef https://github.com/conductor-oss/conductor/blob/main/common/src/main/java/com/netflix/conductor/common/metadata/tasks/TaskDef.java for the retry policy. Give an easy way to configure with sensible defaults and allow users to configure as required.