Add API via agent-protocol SDK

[ValentaTomas]

Background

After a conversation with @merwanehamadi we came up with a minimal agent communication protocol that we think could be useful for the community — if more agents use a universal protocol people can way more easily benchmark, develop, deploy and use agents.

We formalized the spec in a repository and also added SDK wrappers to get rid of the boilerplate when implementing the spec. We also started adding the protocol support to agents like smol developer, beebot and babyagi.

As I mentioned, the spec is minimal — we plan to support workspaces/artifacts and other functionality, but the core has to be usable and make sense, that's why we are starting with a minimal spec and plan to iteratively add more.

Changes

I made changes to the cli_web_app runner that was not actively developed — I reused the original CLI methods and let the user start a server via the agent-protocol package.

The user can then interact (this is the minimal interaction) with the API via POST /tasks for creating a task that AutoGPT should run and POST /agent/tasks/{task_id}/steps for a repeated execution of steps from the AutoGPT loop while optionally adding user_input that the execution can process.
For this, I used a modified run_interaction_loop function that only executes one step of the loop and waits for another API POST /agent/tasks/{task_id}/steps request to continue.

Now the next step is to implement calling this API via a CLI as was the original plan of the CLI web app runner.

Documentation

I need to document the changes and the new functionality.

Test Plan

I still need to thoroughly test these changes.

PR Quality Checklist

• My pull request is atomic and focuses on a single change.
• I have thoroughly tested my changes with multiple different prompts.
• I have considered potential risks and mitigations for my changes.
• I have documented my changes clearly and comprehensively.
• I have not snuck in any "extra" small tweaks changes.
• I have run the following commands against my code to ensure it passes our linters:

  black .
  isort .
  mypy
  autoflake --remove-all-unused-imports --recursive --ignore-init-module-imports --ignore-pass-after-docstring autogpt tests --in-place

[github-actions]

This pull request has conflicts with the base branch, please resolve those so we can evaluate the pull request.

[netlify]

❌ Deploy Preview for auto-gpt-docs failed.

Name	Link
🔨 Latest commit	49eace4
🔍 Latest deploy log	https://app.netlify.com/sites/auto-gpt-docs/deploys/64bd9786acdc0000081b2e2e

[waynehamadi]

@ValentaTomas here is how we're going to do this one.
I have a lot of questions about this, I didn't realize there would be an interaction_step and all that.

But I know :

• e2b has our back if we have any issue.
• this web server is not even used now.

So let's trust. Let's iterate and adjust.

And we'll ask e2b to help us adjust as we go.

So I am going to merge but see this as an iterative process.

I will write tests against this interface, using agbenchmark.