[DO NOT MERGE] Amazon Bedrock AgentCore exploration #3098
Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. 🚀 New features to boost your workflow:
|
There was a problem hiding this comment.
I think this file was necessary for the AgentCore deployment to work...we couldn't just rely on pyproject.toml. But I should double-check.
There was a problem hiding this comment.
Most of my learnings are in this file, not the code.
kompfner
force-pushed
the
pk/aws-agentcore-exploration
branch
from
fabfb8f to
0c27230
Compare
| from pipecat.transports.base_transport import BaseTransport, TransportParams | ||
| from pipecat.transports.daily.transport import DailyParams | ||
|
|
||
| app = BedrockAgentCoreApp() |
There was a problem hiding this comment.
Here's an AgentCore-specific thing
| await runner.run(task) | ||
|
|
||
|
|
||
| @app.entrypoint |
There was a problem hiding this comment.
Here's an AgentCore-specific method
| await run_bot(transport, RunnerArguments()) | ||
|
|
||
|
|
||
| if __name__ == "__main__": |
There was a problem hiding this comment.
Here we have some AgentCore-specific logic
|
|
||
| Follow the prompts to complete the configuration. | ||
|
|
||
| **IMPORTANT:** when asked if you want to use "Direct Code Deploy" or "Container", choose "Container". Today there is an incompatibility between Pipecat and "Direct Code Deploy". |
There was a problem hiding this comment.
Here's a friction point I hit
| PIPECAT_LOCAL_DEV=1 uv run python bot.py -t daily -d | ||
| ``` | ||
|
|
||
| > Ideally you should be able to use `agentcore launch --local`, but it doesn't currently appear to be working (even with [this workaround](https://github.com/aws/bedrock-agentcore-starter-toolkit/issues/156) applied), at least not for this project. |
There was a problem hiding this comment.
Here's another friction point I hit.
There was a problem hiding this comment.
The biggest friction point of all, though, is that once the agent is running it doesn't seem to be able to receive WebRTC traffic. It can join and leave the Daily call at the appropriate times, and it can print its initial LLM text to console, but it can't actually hear the user.
This is just an exploration, meant to establish patterns and share my findings so far. Eventually if we get to the point where a Pipecat bot can be deployed to AgentCore Runtime, this example should live in the
pipecat-examplesrepo, not this one.