Added account for cost info #360
Conversation
paperqa/agents/env.py
Outdated
| @@ -151,21 +151,22 @@ def export_frame(self) -> Frame: | |||
| async def step( | |||
| self, action: ToolRequestMessage | |||
| ) -> tuple[list[Message], float, bool, bool]: | |||
|
|
|||
| # add usage for actions that have usage | |||
There was a problem hiding this comment.
| # add usage for actions that have usage | |
| # add usage if the action has usage |
There's just one action
| @@ -239,6 +239,12 @@ class AgentSettings(BaseModel): | |||
| default="gpt-4o-2024-08-06", | |||
| description="Model to use for agent", | |||
| ) | |||
|
|
|||
| agent_llm_config: dict | None = Field( | |||
There was a problem hiding this comment.
A few lines below this there is an agent_config that is basically kwargs for the Agent/ToolSelector
Do you mind adjusting their names and/or descriptions so it's clear what is the difference between agent_llm_config and agent_config? They're similar enough in naming right now I think we should fix
There was a problem hiding this comment.
👍 Tried to make the descriptions more clear in distinction
There was a problem hiding this comment.
Thanks for your efforts! Being honest, it's still not quite intuitive enough for my tastes, but for now it's good.
I think we can make this clearer by moving to_aviary_tool_selector and to_ldp_agent to be methods of AgentSettings, since all their information is derived from there.
| @@ -239,6 +239,12 @@ class AgentSettings(BaseModel): | |||
| default="gpt-4o-2024-08-06", | |||
| description="Model to use for agent", | |||
| ) | |||
|
|
|||
| agent_llm_config: dict | None = Field( | |||
There was a problem hiding this comment.
Thanks for your efforts! Being honest, it's still not quite intuitive enough for my tastes, but for now it's good.
I think we can make this clearer by moving to_aviary_tool_selector and to_ldp_agent to be methods of AgentSettings, since all their information is derived from there.
Waiting on https://github.com/Future-House/aviary/pull/28
Checks messages from agent for usage info and if present, will attach to current answer objects.