Use new template repo #7
Conversation
staverm
left a comment
staverm
left a comment
There was a problem hiding this comment.
Agent impls look good. Both scenarios run smoothly on my machine. I'm on board for getting these in but I'm not sure what to do with the old code.
We can either delete it or leave it as alternative impls, but then it would be good to mention that in some README. Keeping alternative impls seems fine for a tutorial repo. What do you think?
Talking about these files:
Tau2
- tau2/tau2_agent.py
- tau2/tau2_evaluator.py
Debate
- debate/debater.py
- debate/debate_judge_common.py
- debate/debate_judge.py
- debate/adk_debate_judge.py
- debate/Dockerfile.adk-debate-judge
we can e.g. add a debate/README file and briefly explain the different impls
The old 'SDK'
- src/agentbeats/green_executor.py
- src/agentbeats/tool_provider.py
it's weird that this is shared code, but the new template Executor is not
| break | ||
| time.sleep(0.5) | ||
| else: | ||
| time.sleep(1.0) |
There was a problem hiding this comment.
sleep here looks unnecessary, we're already waiting for agents to be ready above
| load_dotenv(override=True) | ||
|
|
||
|
|
||
| def _connect_host(host: str) -> str: |
There was a problem hiding this comment.
This mapping is only done in ensure_endpoints_unused and not in wait_for_agents or client_cli.py. Not sure if this is actually useful, maybe consider removing it
| json.dump(output_data, f, indent=2) | ||
| print(f"Results written to {output_path}") | ||
|
|
||
| if error_status: |
There was a problem hiding this comment.
ℹ️ this change is to write the output json file even when green agent fails (writes role-id mapping and empty results)
| self.messages.append({"role": "assistant", "content": assistant_content}) | ||
|
|
||
| await updater.add_artifact( | ||
| parts=[Part(root=TextPart(text=assistant_content))], |
There was a problem hiding this comment.
Can also use DataPart here since the output is json
| ``` | ||
|
|
||
| If you see `Error: Some agent endpoints are already in use`, update the `endpoint` + `--port` values in `scenarios/tau2/scenario.toml`. | ||
|
|
There was a problem hiding this comment.
add uv sync --extra tau2-agent --extra tau2-evaluator in the Setup section
| 1. Receives task descriptions with available tools from the green agent | ||
| 2. Decides which tool to call or how to respond | ||
| The tau2 purple agent now lives in `scenarios/tau2/agent/src/`. | ||
| This file is kept so external links and older commands keep working. |
There was a problem hiding this comment.
we can just delete this file right? if we delete this, we should also delete scenarios/tau2/tau2_evaluator.py. Or we leave both files as alternative impls
No description provided.