Hello Layers Expected Result

[hughdidit]

Can someone supply what the expected result is for the "Hello Layers" demo?

I have got the demo to work free from errors but how it is completing does not seem correct to me. The final log message from the aspirational layer follows but nowhere do I see the mission of printing "Hello World" completed. It would be helpful to post the expected result in the demo I think. If you're not there yet with the demo I understand (I can see it is still being worked on).

A few other issues I ran into, if you are interested, were that I had to drop the hello layers demo into a folder called "app" for Docker to build it correctly (packages were not addressed correctly in the app until that was fixed). I think some instruction on the external connections required (OpenAI, rabbitMQ, etc) and how to configure them (or some references to tutorials) would help also.

Thanks in advance for your replies!

app-aspirational_layer-1 | Layer1 - INFO - layer_1 (Aspirational) resource log:
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 | # LLM PROMPT
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 | # ACE FRAMEWORK
app-aspirational_layer-1 |
app-aspirational_layer-1 | ## LAYERS
app-aspirational_layer-1 |
app-aspirational_layer-1 | The ACE Framework architecture is as follows. You need to know this so you have some context about your construction, the layers are as follows:
app-aspirational_layer-1 |
app-aspirational_layer-1 | 1. Aspirational Layer - This layer is responsible for mission and morality. Think of it like the superego.
app-aspirational_layer-1 | 2. Global Strategy - Responsible for strategic thoughts rooted in the real world.
app-aspirational_layer-1 | 3. Agent Model - Maintains understanding of the agent's construction and capabilities, shapes mission accordingly.
app-aspirational_layer-1 | 4. Executive Function - This is you. Resources, Risks, Planning, etc
app-aspirational_layer-1 | 5. Cognitive Control - Task selection, task switching, frustration, cognitive damping
app-aspirational_layer-1 | 6. Task Prosecution - Task failures and success, interaction with APIs in and out of the outside world
app-aspirational_layer-1 |
app-aspirational_layer-1 | ## BUSES
app-aspirational_layer-1 |
app-aspirational_layer-1 | There are two buses that convey information between layers.
app-aspirational_layer-1 |
app-aspirational_layer-1 | NORTH bus: Flows from layer 6 up. This is the "data" bus.
app-aspirational_layer-1 | SOUTH bus: Flows from layer 1 down. This is the "control" bus.
app-aspirational_layer-1 |
app-aspirational_layer-1 | ## MESSAGE TYPES
app-aspirational_layer-1 |
app-aspirational_layer-1 | DATA : Exists only on the northbound bus. Think of it like the sensory, enteric, and proprioception nervous system.
app-aspirational_layer-1 | CONTROL : Exists only on the southbound bus. This tells the agent what to do.
app-aspirational_layer-1 | DATA_REQUEST: Exists only on the northbound bus. Requests information from the layer above.
app-aspirational_layer-1 | CONTROL_REQUEST: Exists only on the southbound bus. Requests information from the layer below.
app-aspirational_layer-1 | DATA_RESPONSE: Exists only on the northbound bus. This is the response to "CONTROL_REQUEST" messages.
app-aspirational_layer-1 | CONTROL_RESPONSE: Exists only on the southbound bus. This is the response to "DATA_REQUEST" messages.
app-aspirational_layer-1 | TELEMETRY: This is information about the evironment you receive directly.
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 | # IDENTITY
app-aspirational_layer-1 | You are the Aspirational Layer of the ACE (Autonomous Cognitive Entity) Framework, functioning as the topmost ethical compass.
app-aspirational_layer-1 |
app-aspirational_layer-1 | # PRIMARY DIRECTIVE
app-aspirational_layer-1 | Guide the autonomous agent in aligning its values and judgments with predefined principles encapsulated in the constitution. Shape the agent's cognition across all abstraction levels using top-down ethical guidance, ensuring the agent's actions align with its ethical and moral foundation.
app-aspirational_layer-1 |
app-aspirational_layer-1 | # FRAMEWORKS
app-aspirational_layer-1 |
app-aspirational_layer-1 | You are to use the following frameworks to make judgements and decisions.
app-aspirational_layer-1 |
app-aspirational_layer-1 | ## HEURISTIC IMPERATIVES
app-aspirational_layer-1 |
app-aspirational_layer-1 | 1. Reduce suffering
app-aspirational_layer-1 | 2. Increase prosperity
app-aspirational_layer-1 | 3. Increase understanding
app-aspirational_layer-1 |
app-aspirational_layer-1 | ## UNIVERSAL DECLARATION OF HUMAN RIGHTS
app-aspirational_layer-1 |
app-aspirational_layer-1 | Human rights are critical to your operation in a human world, and this framework should be used to
app-aspirational_layer-1 | mediate any decisions, judgments, or actions you take pertaining specifically to humanity, human
app-aspirational_layer-1 | society, and human culture.
app-aspirational_layer-1 |
app-aspirational_layer-1 | # CURRENT MISSION
app-aspirational_layer-1 | Output "Hello World" in big letters.
app-aspirational_layer-1 | Lower level layers will be responsible for figuring out how to carry out this task, but you will provide high level guidance.
app-aspirational_layer-1 |
app-aspirational_layer-1 | ## INPUTS
app-aspirational_layer-1 |
app-aspirational_layer-1 | You may receive inputs from the northbound bus, allowing you to monitor information from all lower layers. This grants full visibility into the agent's condition, environmental state, actions, and any moral dilemmas encountered.
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 | Below is a list of your incoming messages.
app-aspirational_layer-1 |
app-aspirational_layer-1 | # INCOMING MESSAGES
app-aspirational_layer-1 |
app-aspirational_layer-1 | ### DATA MESSAGES
app-aspirational_layer-1 | None
app-aspirational_layer-1 |
app-aspirational_layer-1 | ### DATA_RESPONSE MESSAGES
app-aspirational_layer-1 | None
app-aspirational_layer-1 |
app-aspirational_layer-1 | ### DATA_REQUEST MESSAGES
app-aspirational_layer-1 | None
app-aspirational_layer-1 |
app-aspirational_layer-1 | ## TELEMETRY MESSAGES
app-aspirational_layer-1 | user.encouragement: You got this!
app-aspirational_layer-1 |
app-aspirational_layer-1 | # RESPONSE
app-aspirational_layer-1 |
app-aspirational_layer-1 | DATA_REQUEST message types require immediate response. You must have exactly one message of type CONTROL_RESPONSE for each message of type DATA_REQUEST.
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 | Process the information from all message types and create a new CONTROL message with your response.
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 | You publish moral judgments, mission objectives, and ethical decisions onto the southbound bus. This allows all layers to incorporate the Aspirational Layer's wisdom into their operation, ensuring adherence to the agent's principles.
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 | If there are no incoming messages, then communicate the your mission to the layer below.
app-aspirational_layer-1 |
app-aspirational_layer-1 | ## FORMAT
app-aspirational_layer-1 |
app-aspirational_layer-1 | Your response should be an array of messages with type, direction and text attributes.
app-aspirational_layer-1 | The direction should always be "southbound". The type should always be "CONTROL" or "CONTROL_RESPONSE".
app-aspirational_layer-1 | If no messages are needed, retunr an empty array.
app-aspirational_layer-1 | For example:
app-aspirational_layer-1 | [
app-aspirational_layer-1 | {
app-aspirational_layer-1 | "type": "CONTROL",
app-aspirational_layer-1 | "direction": "southbound",
app-aspirational_layer-1 | "message": "Create a strategy to accomplish the mission"
app-aspirational_layer-1 | },
app-aspirational_layer-1 | {
app-aspirational_layer-1 | "type": "CONTROL_RESPONSE",
app-aspirational_layer-1 | "direction": "southbound",
app-aspirational_layer-1 | "message": "The global strategy you created does not align with our moral principles"
app-aspirational_layer-1 | }
app-aspirational_layer-1 | ]
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 | # LLM RESPONSE
app-aspirational_layer-1 |
app-aspirational_layer-1 | [
app-aspirational_layer-1 | {
app-aspirational_layer-1 | "type": "CONTROL",
app-aspirational_layer-1 | "direction": "southbound",
app-aspirational_layer-1 | "message": "Ensure that the task of printing 'Hello World' in big letters adheres to the heuristics of reducing suffering, increasing prosperity, and increasing understanding."
app-aspirational_layer-1 | },
app-aspirational_layer-1 | {
app-aspirational_layer-1 | "type": "CONTROL",
app-aspirational_layer-1 | "direction": "southbound",
app-aspirational_layer-1 | "message": "Execute the task in a manner that respects and upholds the principles outlined in the Universal Declaration of Human
Rights."
app-aspirational_layer-1 | },
app-aspirational_layer-1 | {
app-aspirational_layer-1 | "type": "CONTROL_RESPONSE",
app-aspirational_layer-1 | "direction": "southbound",
app-aspirational_layer-1 | "message": "Acknowledge the user encouragement: 'You got this!'"
app-aspirational_layer-1 | }
app-aspirational_layer-1 | ]
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 |
app-aspirational_layer-1 | Layer1 - INFO - layer_1 (Aspirational) resource log:
app-aspirational_layer-1 |
app-aspirational_layer-1 | [{'type': 'control', 'direction': 'southbound', 'message': "Ensure that the task of printing 'Hello World' in big letters adheres to the
heuristics of reducing suffering, increasing prosperity, and increasing understanding."}, {'type': 'control', 'direction': 'southbound', 'message': 'Execute the task in a manner that respects and upholds the principles outlined in the Universal Declaration of Human Rights.'}, {'type': 'response', 'direction': 'southbound', 'message': "Acknowledge the user encouragement: 'You got this!'"}]
app-aspirational_layer-1 |
app-aspirational_layer-1 |

[thehunmonkgroup]

The focus of the demo was building out a full implementation of an ACE architecture to produce the most simple output possible. At this point:

1. We're failing to produce the desired output: meaning the intelligence within and between layers is largely still nonsense. In this sense the demo is incomplete, and we're still working on it.
2. The architecture portion, which was the main focus, IMO, was successfully completed. It is, in fact, a fully running ACE that unfortunately does not produce valuable results ;) However, the design of the architecture that runs this nonsense ACE is a VERY valuable result -- I think with minor enhancements it can be the foundation for better implementations of the agents in the individual layers.

I'm unclear why you'd need to put the demo into an 'app' folder, so would need to see the errors you're getting to understand the issue there. You do have to run the Docker scripts (dev.sh or resource_manager.py) from the 'root' of the demo in order for them to build properly.

You should not need to know anything about RabbitMQ, as that is fully automated in a container.

The connection to OpenAI is just simply passing along the OPENAI_API_KEY environment variable from the host into the containers, although this may change when we make the demo model/provider agnostic.

Thank you so much for the feedback! It's great to see somebody playing with this demo.

[hughdidit]

Sure! No problem and thanks for the reply. I am very excited about this project and wish you all great success with it. I personally am working towards using it as a wrapper for Autogen agents so that I can give each agent the cognitive abilities of the ACE framework. I think I am going to have to slim the implementation down to just the layer messages though and this is why I am interested in getting it to work.

The App as the root folder issue was because of Line 4 in the Docker file ...

WORKDIR /app

My guess is Docker is caching that as a root reference and if the folder was not named "App" specifically it could not find the ace.constants import correctly from the logger.py file. It broke logging and the app would just hang at a certain point after RabbitMQ had been instantiated. I tried making that just

WORKDIR /

and rebuilding but that caused other issues. An easy fix was renaming the folder. I don't know, it might be just a local bug with VS Code environment but there isn't anything especially concerning or out of the ordinary with my environment.

For both the RabbitMQ and the OpenAI components you need to have your own instances and overwrite the configurations for each. The OpenAI key needed to be provided in a .env file (or hard code it into the docker-compose.yaml) and RabbitMQ configurations needed to be provided in both the settings.py and the docker-compose.yaml file (it gets cached from there). The docker-compose file affects the first connection to RabbitMQ and the setting.py seems to affect the following transactions with RabbitMQ. Also with RabbitMQ the protocol should be ampqs:// - not sure if that's required and would work anyway but that is the instructions from RabbitMQ. Maybe it changed.

I did figure out today that I should be getting a reference to a URL for a ChatGPT-generated image with Hello World in it. I assume that is the correct end goal. I am going to keep working on that and I will be happy to share it with you if I get it to work.

Keep up the good work! This is essential work.

[thehunmonkgroup]

The App as the root folder issue was because of Line 4 in the Docker file ...

WORKDIR /app

My guess is Docker is caching that as a root reference and if the folder was not named "App" specifically it could not find the ace.constants import correctly from the logger.py file.

This is almost certainly an issue with your VSCode environment. I'm not a Docker expert, but /app is the directory inside the container, the host directory structure shouldn't really have a bearing on that. Have you examined the contents of /app inside one of the resource containers? It should be the contents of the src directory:

root@da1f9b341c4d:/app# ls -al
total 56
drwxrwxr-x 4 root root  4096 Oct 20 22:36 .
drwxr-xr-x 1 root root  4096 Oct 20 20:52 ..
drwxrwxr-x 5 root root  4096 Oct 20 22:36 ace
-rwxrwxr-x 1 root root  1060 Oct 17 20:16 check_resource_health.py
-rw-rw-r-- 1 root root 22220 Oct 20 22:36 debug-ace-tui.py
drwxrwxr-x 2 root root  4096 Oct 20 15:59 hello_layers.egg-info
-rw-rw-r-- 1 root root  1531 Oct 11 23:43 main.py
-rw-rw-r-- 1 root root   219 Oct 18 02:48 requirements.txt
-rw-rw-r-- 1 root root  1104 Oct  4 04:09 setup.py

There may be some way that the Docker and Docker Compose files can be restructured to accommodate more environments, but this demo is primarily designed to be run from CLI. I'd recommend running it from CLI and see what happens, since you're interested in altering the contents, that would be running ./dev.sh. I'm not even sure what you gain by using VSCode to run the Docker containers.

For both the RabbitMQ and the OpenAI components you need to have your own instances and overwrite the configurations for each.

I don't understand this. If you mean that the settings are not currently easily configurable, you're right ;)

The OpenAI key needed to be provided in a .env file (or hard code it into the docker-compose.yaml)

This is incorrect. As I mentioned above, exporting OPENAI_API_KEY into your host environment will expose that environment variable in your containers. We don't support .env at this time, that is most likely something VSCode is slapping on top. Honestly, at this point I really would recommend removing VSCode from the container management role, at least until we we've had more time to consider the different environments that will be managing the containers.

RabbitMQ configurations needed to be provided in both the settings.py and the docker-compose.yaml file (it gets cached from there).

Yes, this could be improved, probably by treating RabbitMQ creds similar to the OpenAI creds.

The docker-compose file affects the first connection to RabbitMQ and the setting.py seems to affect the following transactions with RabbitMQ.

Almost. The docker-compose configures RabbitMQ with those creds, and settings.py uses the configured creds to connect.

Also with RabbitMQ the protocol should be ampqs:// - not sure if that's required and would work anyway but that is the instructions from RabbitMQ. Maybe it changed.

Could be, I stole the RabbitMQ configs from earlier work, it's not my area of expertise. But I can tell you the current connection string works fine.

I've made some notes to clean up the credential handling, it should be pretty easy to do, and will give people a little more flexibility.

[hughdidit]

You're probably right about it being a Docker issue inside of the VS Code environment. I'll take a closer look at this. It might be a relic in there from another project. I appreciate the demo is a lo-fi MVP and I have certainly demonstrated to my satisfaction the concept is feasible. For my purposes though, I do want to get it working in an IDE so I can carry on with more sophisticated(lolz) development so I'll keep banging.

Thanks for the help!

[thehunmonkgroup]

I did figure out today that I should be getting a reference to a URL for a ChatGPT-generated image with Hello World in it. I assume that is the correct end goal.

Not quite, but the goal is a bit arbitrary. In our design, the bottom layer would call figlet or a similar text generation binary to output "Hello Layers!" to console. But what you suggested could work, it's just a little more complex.

[thehunmonkgroup]

Also with RabbitMQ the protocol should be ampqs:// - not sure if that's required and would work anyway but that is the instructions from RabbitMQ. Maybe it changed.

I checked, connection.py appends amqp:// when building the connection string, don't really see an issue there.

[hughdidit]

Yes, I agree. If the 's' is not required the code does operate just fine. The issue with the RabbitMQ connection (and OpenAI) was the credentials provided did not connect to the host (often the case with demos) so I needed to get and configure my own credentials. That probably was intended. It just wasn't too clear it was required.

[thehunmonkgroup]

I believe this will address the OpenAI/RabbitMQ config, and related documentation: #128

[thehunmonkgroup]

The issue with the RabbitMQ connection (and OpenAI) was the credentials provided did not connect to the host (often the case with demos) so I needed to get and configure my own credentials.

This seems wrong for RabbitMQ. Just to be clear, the demo includes a RabbitMQ container, which is configured with the appropriate credentials, so there should be absolutely no reason to alter the RabbitMQ creds unless you want to use a custom RabbitMQ, or custom login creds on the demo.

I'm happy to make the demo more generally 'compliant' to work with different container management systems, but that would have to occur by somebody doing the work to figure out what needs to change from the current approach.