rnadigital/agentcloud

Agent Cloud is like having your own GPT builder with a bunch extra goodies. The GUI features 1) RAG pipeline which can natively embed 260+ datasources 2) Create Conversational apps (like GPTs) 3) Create Multi Agent process automation apps (crewai) 4) Tools 5) Teams+user permissions. Get started fast with Docker and our install.sh

agentcloud's Issues

Deployment script for cloud hosting [AWS, Azure…]

Need a deployment script for cloud hosting; especially because of the Docker component. This will enable us to deploy this stack on AWS, Azure, or similar. Prefer to use Terraform; open source alternatives are also possible; regardless it needs to support ephemeral states [pre-deploy, active, dormant, inactive].

FYI that Vercel doesn’t support Docker; might file another issue for a non-Docker version of this stack.

UX cleanup

Sign up

• emails are not verified in local mode info box
• Implement change password
• get rid of agree to terms of service on login (change to By signing in you agree to the terms of service)
• remove start a 14 day free trial (change to sign up and put in box)

Home page

• on / url, if signed in go to sessions, otherwise go to signup
  Put terms of service in footer
• After login Change first page to be /sessions
• Hid Ids in team selector
• Change color of team switcher (add opacity)

Accounts

• Handle empty account context on back button
  Change the account page to table

Agents

• Temporarily disable new agent button (coming soon feature) For now agentcloud auto generates agents and teams
• Build a teams page (teams are collections of agents)

Session

• Add session prompt on the session card
  

There are three modes/statuses that a session can be in - either

1. Team building
2. Task execution
3. User input required

Team building

• If it's the same person as the previous message don't create a new row
• timestamp in bottom right of message in grey (like whatsapp)
• Add task title in the top bar with the create date timestamp on left side
• make session box one line by default and max to five with a scroller
  Move loading state to the submit button
• Disable message button when message box is empty
• Dont allow to submit a session if the agent_backend isnt in the task_queue room
• Enter to submit, shift enter to start new line
  change first message logic in session to re-send if there are no incoming:false messages in the message list
• Change Do the thing to Start Session
• Add collapsible code blocks
• Pause autoscroll if manually scroll up
• Show the loading state whenever the last message is incoming true or is a feedback message
• #20
• #19
• Represent the team in a clean format (use reactflow cards to show a hierarchy instead of pure code) - let users access code if necessary by a switcher code mode vs preview mode
• #21
• add subtitle that says whether we are generating team or executing (status light in top bar)
• When an agent is pending human feedback, change the status light (Pending human feedback)
• If there is an outstanding task for the human adjust the typical system message Press enter to skip and use auto-reply, or type 'exit' to end the conversation: to Press the 👍Proceed button to continue or send feedback in the text box
  If in team planner mode - exit = 👍Proceed
  If in task mode- exit = End session
• Show outstanding tasks as text and action button in the text box area
• Give agents and users a default icon with this

Task execution

• The system repeats itself, make sure system prompts are not repeated
• Provide feedback to chat_manager.
  
• Current feedback options need to be converted to action buttons
  Press enter to skip and use auto-reply, or type 'exit' to end the conversation
  Convert to buttons/actions

Errors

• #22

Back end - Python

• send date in message in_send_from_socket

My Org/My Team

Is your feature request related to a problem? Please describe.
'My Org' has no landing page or functionality.

Describe the solution you'd like
Enable internal storage to upload, catalog, and modify company documents. This would be a good place to store multiple API Keys an determine organizational customizations/capabilities. I'd imagine it would also allow each Agent to refine their capabilities as well as learn new skills. Furthermore, subscription based storage could be a good source of revenue generation for you.

Show the Groups that agents are part of (to help distinguish)

Is your feature request related to a problem? Please describe.
Hard to tell which agent on the agent screen is linked to what group

Describe the solution you'd like
Show what groups the agents are linked to

Describe alternatives you've considered
N/A

Additional context

Mongo restarting docker build error

Describe the bug
See https://discord.com/channels/1165866460745314304/1167801582923030538/1171861403615903784

To Reproduce
Docker build

Additional context
Add any other context about the problem here.

AgentCloud Agents in list view - great UI improvement

Hi @andrew

I'd like to share a possible improvement at least visually that I'm sure users would appreciate. This same improvement has just been implemented in another type of platform and it is more than relevant. See Image 1, card view, very similar to current AgentCloud Agents section. (Image 2)

As you can see in the attached Images 3 - 4 being able to have 2 views "card view" and "list view" along with the ability to search by Agent names and categories adds invaluable value.

Consider that many developers will be managing dozens or hundreds of Agents, so to keep an overview of all of them, the order is very important.

Many will choose to define categories such as customers, other areas or departments, and the search functionality allows you to quickly identify the Agents you need to edit. Image 5 - 6

I think this would be a great UI improvement.🚀

Please let me know your thougts. Many thanks in advance.

Image 1:

Image 2:

Image 3:

Image 4:

Image 5:

Image 6:

Platform breaks when accessing group with deleted agent

Describe the bug
Deleting an agent that is already in a group will lead to an error when accessing groups

To Reproduce
Steps to reproduce the behavior:

1. Delete an agent that is already part of a group
2. Go to Groups
3. Click View on a group
4. See error

Expected behavior
No error

Screenshots

Desktop (please complete the following information):

• OS: Mac os Ventura
• Browser Chrome
• Version 118

Additional context
Docker v0.0.1-a298fe0

Remove all group agent type selector fields, just put them in one field

Describe the bug
Users only need to specify the agents in a group. Agents contain all necessary config already.
Change Group screen to just be two fields

• Group Name
• Agents (Chips)

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots

Additional context
Docker v0.0.1-cfe4073

Github and google oauth login

Azure credential adding does not have a field for endpoint

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:

1. Go to Credentials
2. Create new credential
3. Select Azure
4. See issue

Expected behavior
Expect a field for endpoint as azure implementations have custom endpoints

Screenshots

Additional context
Docker v0.0.3-b961d55

When system agents are talking to eachother, collapse their work into an expanding section (like when chatGPT plugins are working)

Executor reverting to user proxy agent on create and save

Describe the bug
One thing as well regarding setting up an executor agent - I think there's currently no way to set the type to Executor, it gives "Saved" feedback when set up, but then reverts to User Proxy Agent

To Reproduce
Steps to reproduce the behavior:

1. Create agent
2. Select executor
3. Save
4. See error

Expected behavior
Stays as executor

Additional context
Docker
Alpha cloud 16nov23 8pm

Let users assign agents to groups when creating an agent

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
While creating agent, Let me select/create new group and add agent to it

Describe alternatives you've considered
Creating agents then going to Group screen to add takes a longer time

Multiple output messages from executor

Describe the bug
At the end of my interaction I can see multiple output messages (not sure if bug or expected)

To Reproduce
Steps to reproduce the behavior:

1. Create Arxiv agents (like those mentioned here)
2. Create a Arxiv group
3. Create a session with prompt: find papers on LLM applications from arxiv in the last week, create a markdown table of different domains.
4. Approve the task, get to the end and see duplicate entries

Expected behavior
Single markdown

Screenshots

Desktop (please complete the following information):

• OS: Mac OS ventura
• Browser Chrome
• Version 118

Additional context
Docker v0.0.1-e6812d3

A documentation for the front-end webapp environment variables

If it is possible, can there be a documentation for setting up the environment variables? To a back-end developer, it can be a bit confusing.

Thanks!

Chat prompt field not clearing after sending

Describe the bug
Chat messages are not being cleared after giving feedback

Screenshots

Additional context
v0.0.3-fc8ed45

Action buttons appearing are not applicable

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:

1. Start session
2. Get past team creation stage
3. Wait until a prompt arrives
4. See Approve team button (not required)

Expected behavior
Buttons should be applicable to the context.

Screenshots

Desktop (please complete the following information):

• OS: Mac Ventura
• Browser Chrome
• Version 118

Additional context
Docker
v0.0.1-356345b

Add new message red line and autoscroll like slack '10 new messages'

Backend '<' not supported between instances of 'NoneType' and 'int'

Describe the bug
While running a session I get a bug that prevents me from progressing (feels like the app hangs).

To Reproduce
Steps to reproduce the behavior:
Boot up app using docker

1. Start a session by providing a prompt
2. Progress past the team generation point
3. Error appears in terminal and UI loader keeps going
4. See error

Expected behavior
App should not hang and error should not appear

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: 13.0 (22A380) MacBook Air M2
• Browser: Chrome
• Version 118.0.5993.117

Additional context
Run using docker

Can't run session

Describe the bug
I am an Alpha tester, have created 28 Agents, and 8 groups. After assigning the very first task (to group containing all 28 Agents, asking to define their goals, teams, capabilities, and mission) the UI shows 'running' but nothing happens and no tokens have been used. I have signed out and retried multiple times.

Calculating actual tokens used

I tried two sessions with the model GPT-4-preview-1106, which ended with about 4k and 7k tokens. However, the rate usage of OpenAI API reached $15, which is equivalent to 1500k tokens. It seems the token count only adds the tokens in agent responses in the conversation but ignores the input context.

I am not sure if it is complicate to calculate the tokens consumed in AutoGen, but providing a rough estimation should be also helpful!

[FEATURE] Build team from individual agent in mongo

Is your feature request related to a problem? Please describe.
Currently, the agent-backend expects that the complete team json will be in mongo. However, this limits the ability of users to edit and/or create team members

Describe the solution you'd like
the agent-backend needs to pull individual agents from mongo that are related to a given team id and construct the team JSON at runtime. This will allow users to edit agents and add/remove team members dynamically

Don't display the real messag econtent of feedback response messages

e.g. "End session" should not show "TERMINATE":

Should set a flag when submitting a message through the buttons, and display them as a different format inline like " selected ".

• update chat db/model with new flag
• handle flag in socketio.ts
• update rendering of chat/message component

Allow other GPT models to be used

Is your feature request related to a problem? Please describe.
At the moment, agentcloud defaults to GPT-4 for all agents.
However, some agents might not need GPT-4 and would work fine on GPT-3.5 or maybe the user wants to use GPT-3.5 for cost reasons.

Describe the solution you'd like
Lets the user choose what models agents should use.

Describe alternatives you've considered
none

Additional context
none

Subscribe to unlock sessions is appearing for local version

Describe the bug
Subscribe to unlock sessions is appearing for local version

To Reproduce
Steps to reproduce the behavior:

1. Sign in
2. Start session
3. See modal

Expected behavior
No modal in local version if no stripe credentials entered

Screenshots

Docker v0.0.3-57fe82f

Late arriving chunks caused message to duplicate

I tried the finding arxiv paper example twice. The first time is going well, but the second time raised a issue that one agent (critic) was continually repeating the response when reviewing the markdown table. After 5 times repetition, it triggers a admin feedback requirement. I replied that there was a repetition problem. However, the critic agent repeated its response again.

I am not sure if this bug is reproducible or I should report it in AutoGen. It seems there is a problem calling the agent in this session.
I only entered three prompts in this session:
First is the task initialization: "find papers on LLM applications from arxiv in the last week, create a markdown table of different domains."
Second I entered "confirmed" when it asked for plan feedback.
Third is pointing out the repeating issue: "critic is repeating the response"

[All agent and model settings are the same as the arxiv paper example]

Db consistency tasks

• when deleting an in-use credential, unassign from agents after deleting.
• when deleting an agent that is in a group, remove it from the group
• when deleting an in-use tool, unassign from agents after deleting.
• when a group doesn't have the required admin and at least 1 agent, filter it out from the start session chatbox
• probably more stuff... @ragyabraham @anada10 if you can think of any

Return budget/token count on the sessions screen per message level

TypeError: argument of type 'NoneType' is not iterable

Describe the bug
TypeError: argument of type 'NoneType' is not iterable

To Reproduce
Steps to reproduce the behavior:

1. Start session
2. Enter prompt
3. Start watching stream
4. See error

Expected behavior
No error
Screenshots

Desktop (please complete the following information):

• OS: Mac Ventura
• Browser Chrome
• Version 118

Additional context
Docker
v0.0.1-812b094

Module not found ubuntu22.04

Describe the bug
Get module not found during docker compose up on Ubuntu 22.04

To Reproduce
Steps to reproduce the behavior:
1.start app on docker

Expected behavior
Runs

Desktop (please complete the following information):

• OS: Ubuntu 22.04

Posted on discord
https://discord.com/channels/1165866460745314304/1165866461198295131/1173905333878538321

hello,
new to agentcloud, I don't get it up on ubu22.04:

git clone…
SHORT_COMMIT_HASH=$(git rev-parse --short HEAD) docker compose up --build ## commit: f6558b8

agentcloud-webapp_next-1    | 
agentcloud-webapp_next-1    | > [email protected] start
agentcloud-webapp_next-1    | > NODE_ENV=production node dist/server.js
agentcloud-webapp_next-1    | 
agentcloud-webapp_next-1    | node:internal/modules/cjs/loader:1147
agentcloud-webapp_next-1    |   throw err;
agentcloud-webapp_next-1    |   ^
agentcloud-webapp_next-1    | 
agentcloud-webapp_next-1    | Error: Cannot find module '/app/dist/server.js'
agentcloud-webapp_next-1    |     at Module._resolveFilename (node:internal/modules/cjs/loader:1144:15)
agentcloud-webapp_next-1    |     at Module._load (node:internal/modules/cjs/loader:985:27)
agentcloud-webapp_next-1    |     at Function.executeUserEntryPoint [as runMain] (node:internal/modules/run_main:135:12)
agentcloud-webapp_next-1    |     at node:internal/main/run_main_module:28:49 {
agentcloud-webapp_next-1    |   code: 'MODULE_NOT_FOUND',
agentcloud-webapp_next-1    |   requireStack: []
agentcloud-webapp_next-1    | }
agentcloud-webapp_next-1    | 
agentcloud-webapp_next-1    | Node.js v21.1.0
agentcloud-webapp_next-1 exited with code 1

webapp can't start - SyntaxError: Unexpected end of JSON input

Describe the bug
Starting from a clean install, I updated the key in the docker-compose.yml file and started it in Docker Desktop. agentcloud-webapp is in a startup loop because of this error:

agentcloud-webapp_next-1 | SyntaxError: Unexpected end of JSON input
agentcloud-webapp_next-1 | at JSON.parse ()
agentcloud-webapp_next-1 | at ReadStream. (/app/node_modules/google-auth-library/build/src/auth/googleauth.js:431:43)
agentcloud-webapp_next-1 | at ReadStream.emit (node:events:515:28)
agentcloud-webapp_next-1 | at ReadStream.emit (node:domain:488:12)
agentcloud-webapp_next-1 | at endReadableNT (node:internal/streams/readable:1589:12)
agentcloud-webapp_next-1 | at process.processTicksAndRejections (node:internal/process/task_queues:82:21)
agentcloud-docker_mongo-1 | {"t":{"$date":"2023-10-26T14:12:49.180+00:00"},"s":"I", "c":"NETWORK", "id":22944, "ctx":"conn1","msg":"Connection ended","attr":{"remote":"172.19.0.4:43936","uuid":{"uuid":{"$uuid":"99e18779-aaca-4c04-a937-0898c1c3bf3a"}},"connectionId":1,"connectionCount":0}}
dependency failed to start: container agentcloud-webapp_next-1 is unhealthy

To Reproduce
Just install using the instructions?

Expected behavior
Expected the app to load correctly.

Screenshots
(attached error log earlier)

Desktop (please complete the following information):

• OS: Windows 11
• Browser : I never got to the point that I could load it.
• Version: NA

Additional context
Happy to provide any additional information about my environment, but I am not sure what would be relevant. It is a 5950x with 128GB of RAM and a 3090 GPU. Docker runs containers normally, so I don't think that's the issue.

Force stop session (so that it stops hitting open AI)

Is your feature request related to a problem? Please describe.
There is no way of stopping then editing a session if you've made a mistake

Describe the solution you'd like
Allow people to stop mid response and then edit their session/question

• Editing and submitting would be like starting a new session with the same group
• Stopping would make sure no more calls are made to open AI.

Describe alternatives you've considered
Using the TERMINATE command is sometimes too late (sometimes agents have begn extensive back and forth using up tokens, and you can't interrupt them.

Additional context
Stop

Edit button - Button only appears on hover of prompt

Edit state - Save/Submit & Cancel

Platform is creating new blank agents in DB when existing agents are re-used in a new group's session

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:

1. Create a new group
2. Add a mix of new agents and pre-existing agents
3. Run a session
4. Go back to agents screen
5. See new (duplicated) agents

Expected behavior
No additional agents are created

Additional context
Docker v0.0.1-a298fe0
Alpha

Let users socially share their groups/outputs on X (Twitter) or Agent Exchange

Is your feature request related to a problem? Please describe.
Users should be able to socialise their Groups to agent exchange, given it takes time to optimise them

Describe the solution you'd like
Tweet<script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script>

KeyError: 'retry_wait_time'

Describe the bug
Hi. I got an error after I built the docker. The UI is opened, but when I tried to send a message, I got this:

agentcloud-agent_backend-1  | ERROR:root:'retry_wait_time'
agentcloud-agent_backend-1  | Traceback (most recent call last):
agentcloud-agent_backend-1  |   File "/app/agents/base.py", line 24, in task_execution
agentcloud-agent_backend-1  |     create_run_chat(session_id, group.get("roles"), group.get("group_chat"), task, {})
agentcloud-agent_backend-1  |   File "/app/build/build_group.py", line 53, in create_run_chat
agentcloud-agent_backend-1  |     user_proxy.initiate_chat(recipient=manager, message=prompt, clear_history=True, **history)
agentcloud-agent_backend-1  |   File "/usr/local/lib/python3.11/site-packages/autogen/agentchat/conversable_agent.py", line 565, in initiate_chat
agentcloud-agent_backend-1  |     self.send(self.generate_init_message(**context), recipient, silent=silent)
agentcloud-agent_backend-1  |   File "/usr/local/lib/python3.11/site-packages/autogen/agentchat/conversable_agent.py", line 360, in send
agentcloud-agent_backend-1  |     recipient.receive(message, self, request_reply, silent)
agentcloud-agent_backend-1  |   File "/usr/local/lib/python3.11/site-packages/autogen/agentchat/conversable_agent.py", line 496, in receive
agentcloud-agent_backend-1  |     reply = self.generate_reply(messages=self.chat_messages[sender], sender=sender)
agentcloud-agent_backend-1  |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
agentcloud-agent_backend-1  |   File "/usr/local/lib/python3.11/site-packages/autogen/agentchat/conversable_agent.py", line 940, in generate_reply
agentcloud-agent_backend-1  |     final, reply = reply_func(self, messages=messages, sender=sender, config=reply_func_tuple["config"])
agentcloud-agent_backend-1  |                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
agentcloud-agent_backend-1  |   File "/usr/local/lib/python3.11/site-packages/autogen/agentchat/groupchat.py", line 182, in run_chat
agentcloud-agent_backend-1  |     speaker = groupchat.select_speaker(speaker, self)
agentcloud-agent_backend-1  |               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
agentcloud-agent_backend-1  |   File "/usr/local/lib/python3.11/site-packages/autogen/agentchat/groupchat.py", line 92, in select_speaker
agentcloud-agent_backend-1  |     final, name = selector.generate_oai_reply(
agentcloud-agent_backend-1  |                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
agentcloud-agent_backend-1  |   File "/usr/local/lib/python3.11/site-packages/autogen/agentchat/conversable_agent.py", line 650, in generate_oai_reply
agentcloud-agent_backend-1  |     response = client.create(
agentcloud-agent_backend-1  |                ^^^^^^^^^^^^^^
agentcloud-agent_backend-1  |   File "/usr/local/lib/python3.11/site-packages/autogen/oai/client.py", line 262, in create
agentcloud-agent_backend-1  |     response = self._completions_create(client, params, **extra_config)
agentcloud-agent_backend-1  |                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
agentcloud-agent_backend-1  |   File "/usr/local/lib/python3.11/site-packages/autogen/oai/client.py", line 277, in _completions_create
agentcloud-agent_backend-1  |     params.pop("retry_wait_time")
agentcloud-agent_backend-1  | KeyError: 'retry_wait_time'

Adding support for Retrival chat and function calling

Hi,

Saw your project really cool work. I was able to set it up locally. I am currently working on adding support for Retrival chat agents. Autogen already has examples for this. I would also like to add support for function calling in the group chat, do you guys have idea what would be the best way to approach this?

SyntaxError: invalid decimal literal

Describe the bug
Bug in agent-backend SyntaxError: invalid decimal literal

To Reproduce
Steps to reproduce the behavior:

Edit the base.py file line 66 to

Return the team using the JSON structure as a template:

1. Start session with Create tekken using unreal engine
2. Provided feedback If you will be writing code can you make sure to include a code executor and make sure ONLY the code executor can execute code. Also make sure that when code is produced, a tester can test it and pass back feedback to the developer
3. Accept Team
4. Error

Expected behavior
No error

Screenshots

Desktop (please complete the following information):

• OS: Chrome
• Browser [e.g. chrome, safari]
• Version [e.g. 22]

Additional context
Docker
v0.0.1-356345b

Chat prompt text is partially hidden by send button

Describe the bug
Chat prompt text is partially hidden by send button
To Reproduce
Steps to reproduce the behavior:

1. Start a session
2. Start typing in the prompt box
3. Notice text under the send button
4. See error

Expected behavior
Text should be to the left of the send button

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: Mac Ventura
• Browser Chrome
• Version 118

Additional context
Running docker

Ability to choose pre-defined agents

Is your feature request related to a problem? Please describe.
I can't seem to choose the agents I have pre-defined in the agents tab when starting a new session.

Describe the solution you'd like
When starting a new session, I would like to choose the agents I've predefined under agents instead of adding new agents again.

Describe alternatives you've considered
The alternatives considered is to ignore the agents created and start the session and (1) use the roles suggested by the roles planner or (2) re-add agents from "start a session manually" both are not ideal.

Additional context

"Start a session manually" should first ask us whether we want to pick agents or add agents.

Agent cloud hard to read in night mode

Describe the bug
Hard to read in night mode

To Reproduce
Steps to reproduce the behavior:

1. Change system to dark mode
2. View screen

Expected behavior
Inverted text

Screenshots

Additional context
Alpha Docker

Gracefully handle errors during the chat (such as timeouts, rate limits with Open AI)

Add db migrations

Add more Open AI models for Agents to use

Is your feature request related to a problem? Please describe.
Allow agents to use latest open ai models

Describe the solution you'd like
Models are found here

Provide more options here

On sign in Loading... appears

Describe the bug
When logging in it says Loading... until refreshing

To Reproduce

1. Sign in
2. See error

Expected behavior
Go straight to sessions screen

Screenshots

Additional context
Alpha Docker

[FEATURE] allow users to stop and resume sessions

Is your feature request related to a problem? Please describe.
Right now users can not stop and/or resume a session.

Describe the solution you'd like
Give users the ability to cancel their session and if they wish they can come back and resume at any time.

Additional context
work has been done to allow users to STOP a session. What is remaining is that they can no resume a session at any given time

Can't save team without code executor

Describe the bug
I can't create a team without a code executor

To Reproduce
Steps to reproduce the behavior:

1. Go Groups
2. Click Create new
3. Enter members, don't add code executor
4. See error

Expected behavior
Can create a team that does not have a coder

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: Mac
• Browser Chrome
• Version 118

Additional context
Docker v0.0.1-cfe4073

Session does not terminate when the end session button is pressed

Describe the bug
Session does not terminate when the end session button is pressed
To Reproduce
Steps to reproduce the behavior:

1. Get to the end of a session
2. End session button
3. Session doesn't terminate

Expected behavior
Session should terminate.
I typed exit and it worked

Screenshots
Pressing end session did not terminate

Pressing exit terminated the session (noticed status is not accurate)

Desktop (please complete the following information):

• OS: Mac Ventura
• Browser Chrome
• Version 118

Smartphone (please complete the following information):

• Device: [e.g. iPhone6]
• OS: [e.g. iOS8.1]
• Browser [e.g. stock browser, safari]
• Version [e.g. 22]

Additional context
Add any other context about the problem here.

UX: Code box is not full width and the collapse button is too high up

Describe the bug

the code box is not full with and the
To Reproduce
Steps to reproduce the behavior:

1. Start session
2. Get past team generation
3. Get a code output
4. See error

Expected behavior
Code box should span width of the header
The collapse button should sit at least three lines down

Screenshots

Desktop (please complete the following information):

• OS: Mac OS Ventura
• Browser Chrome
• Version 118

Additional context
Docker

403 when creating new agent

Describe the bug
Click create new agent and got 403

To Reproduce
Steps to reproduce the behavior:

1. Go to 'Agents'
2. Click Create an Agent
3. See error

This is hard to replicate. Seems like it happens after new builds?

Expected behavior
New agent screen loads

Screenshots

Desktop (please complete the following information):

• OS: Mac
• Browser Chrome
• Version 118

Additional context
Docker v0.0.1-cfe4073

Cancel button does not work on agents

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:

1. Go to Agents
2. Create new agent
3. Click Cancel
4. See error

Expected behavior
Cancel goes back to agents screen

Screenshots

Desktop (please complete the following information):

• OS: Mac os ventura
• Browser Chrome
• Version 118

Additional context
Docker v0.0.1-cfe4073