Error parsing JSON response with literal_eval invalid syntax (<unknown>, line 1) JSON Validation Error: 'thoughts' is a required property

imfleabagg commented 1 year ago

⚠️ Search for existing issues first ⚠️

[X] I have searched the existing issues, and there is no existing issue for my problem

Which Operating System are you using?

Other

Which version of Auto-GPT are you using?

Latest Release

Do you use OpenAI GPT-3 or GPT-4?

GPT-3.5

Which area covers your issue best?

Installation and setup

Describe your issue.

Hi All !

I have been playing around with AutoGPT for a few days now, and it had been running quite well until yesterday afternoon. I am very much unsure what change but this is the error i keep getting now :

Not sure why , as It started fairly randomly. I am no developper whatsoever , so if anyone could help, I would very much appreciate it !! thanks

Upload Activity Log Content

No response

Upload Error Log Content

No response

kamacau commented 1 year ago

same here `Using memory of type: JSONFileMemory Using Browser: chrome Error parsing JSON response with literal_eval invalid syntax (, line 1)
JSON Validation Error: 'thoughts' is a required property

Failed validating 'required' in schema: {'$schema': 'http://json-schema.org/draft-07/schema#', 'additionalProperties': False, 'properties': {'command': {'additionalProperties': False, 'properties': {'args': {'type': 'object'}, 'name': {'type': 'string'}}, 'required': ['name', 'args'], 'type': 'object'}, 'thoughts': {'additionalProperties': False, 'properties': {'criticism': {'description': 'constructive ' 'self-criticism', 'type': 'string'}, 'plan': {'description': '- ' 'short ' 'bulleted\n' '- ' 'list ' 'that ' 'conveys\n' '- ' 'long-term ' 'plan', 'type': 'string'}, 'reasoning': {'type': 'string'}, 'speak': {'description': 'thoughts ' 'summary ' 'to ' 'say ' 'to ' 'user', 'type': 'string'}, 'text': {'description': 'thoughts', 'type': 'string'}}, 'required': ['text', 'reasoning', 'plan', 'criticism', 'speak'], 'type': 'object'}}, 'required': ['thoughts', 'command'], 'type': 'object'}

On instance: {} JSON Validation Error: 'command' is a required property

Failed validating 'required' in schema: {'$schema': 'http://json-schema.org/draft-07/schema#', 'additionalProperties': False, 'properties': {'command': {'additionalProperties': False, 'properties': {'args': {'type': 'object'}, 'name': {'type': 'string'}}, 'required': ['name', 'args'], 'type': 'object'}, 'thoughts': {'additionalProperties': False, 'properties': {'criticism': {'description': 'constructive ' 'self-criticism', 'type': 'string'}, 'plan': {'description': '- ' 'short ' 'bulleted\n' '- ' 'list ' 'that ' 'conveys\n' '- ' 'long-term ' 'plan', 'type': 'string'}, 'reasoning': {'type': 'string'}, 'speak': {'description': 'thoughts ' 'summary ' 'to ' 'say ' 'to ' 'user', 'type': 'string'}, 'text': {'description': 'thoughts', 'type': 'string'}}, 'required': ['text', 'reasoning', 'plan', 'criticism', 'speak'], 'type': 'object'}}, 'required': ['thoughts', 'command'], 'type': 'object'}

On instance: {}

NEXT ACTION: COMMAND = None ARGUMENTS = None Enter 'y' to authorise command, 'y -N' to run N continuous commands, 's' to run self-feedback commands, 'n' to exit program, or enter feedback for ServerlessGPT... Asking user via keyboard... Input:`

uneuro commented 1 year ago

fixed on my side with creating a fresh ai_settings.yaml

uneuro commented 1 year ago

fixed on my side with creating a fresh ai_settings.yaml

hum, got it again a bit after in another request (master branch)

imfleabagg commented 1 year ago

Not sure what is causing this , I am using stable branch, would you mind sharing how you "fixed" it previously? Again sorry, fairly new to this

ewave-design commented 1 year ago

I keep getting that error no matter what i do as well!

Update: So seems like the work around for this issue is to use Pinecone or Redis as the (MEMORY_BACKEND=)

If your using AutoGPT via Git or local install you will need to create accounts with the above mentioned or Use Docker setup which includes Redis setup.

Personally just made an account with Pincone as docker setup was a headace as a novice to these things. hope it helps

erik-megarad commented 1 year ago

Much of this will be fixed when we move to OpenAI functions. In the mean time we can try to figure out a solution.

On Tue, Jun 20, 2023 at 9:05 AM ewave-design @.***> wrote:

I keep getting that error no matter what i do as well!

— Reply to this email directly, view it on GitHub https://github.com/Significant-Gravitas/Auto-GPT/issues/4752#issuecomment-1599089207, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAABQXF2MVO4SNVC6DRU3Q3XMHC3XANCNFSM6AAAAAAZNAX7RE . You are receiving this because you are subscribed to this thread.Message ID: @.***>

chenkc805 commented 1 year ago

I am also getting this error. I'm also not a developer and not familiar with what Redis or Pinecone are.

chenkc805 commented 1 year ago

fixed on my side with creating a fresh ai_settings.yaml

How do you create a fresh ai_settings.yaml and what does that exactly do?

AngryFinn commented 1 year ago

FWIW, I too am experiencing this. Worked well for 20ish iterations before bombing out.

This is my first time running it, so I'm not sure if it's a recent commit issue, or a bug that has existed for a while.

VectorZhao commented 1 year ago

same error

NEXT ACTION:  COMMAND = read_file ARGUMENTS = {'filename': '/home/Auto-GPT/autogpt/auto_gpt_workspace/pdf2word.py'}
  Enter 'y' to authorise command, 'y -N' to run N continuous commands, 's' to run self-feedback commands, 'n' to exit program, or enter feedback for PDF2WordGPT...
  Asking user via keyboard...
Input:y
-=-=-=-=-=-=-= COMMAND AUTHORISED BY USER -=-=-=-=-=-=-= 
  Text length: 219 tokens
  Max chunk length: 3545 tokens
  Failed to update API costs: KeyError: 'gpt-35-turbo'
  Failed to update API costs: KeyError: 'ada'
  Failed to update API costs: KeyError: 'ada'
SYSTEM:  Command read_file returned: import PyPDF2 import docx def convert_pdf_to_word(pdf_file_path, word_file_path): # Open the PDF file pdf_file = open(pdf_file_path, 'rb') pdf_reader = PyPDF2.PdfReader(pdf_file) # Create a new Word document doc = docx.Document() # Loop through each page of the PDF file for page_num in range(pdf_reader.numPages): # Extract the text from the page page = pdf_reader.getPage(page_num) text = page.extractText() # Add the text to the Word document doc.add_paragraph(text) # Save the Word document doc.save(word_file_path) if __name__ == '__main__': # Get the PDF file path and Word file path from the user pdf_file_path = input('Enter the path of the PDF file: ') word_file_path = input('Enter the path of the Word file: ') # Convert the PDF file to a Word document convert_pdf_to_word(pdf_file_path, word_file_path)
  Error parsing JSON response with literal_eval unterminated string literal (detected at line 1) (<unknown>, line 1)
  Error parsing JSON response with literal_eval '{' was never closed (<unknown>, line 1)
  Error parsing JSON response with literal_eval '{' was never closed (<unknown>, line 1)
  Error parsing JSON response with literal_eval '{' was never closed (<unknown>, line 1)
  Error parsing JSON response with literal_eval '{' was never closed (<unknown>, line 1)
  Failed to update API costs: KeyError: 'gpt-35-turbo'
  Failed to update API costs: KeyError: 'gpt-35-turbo'
  Error parsing JSON response with literal_eval '{' was never closed (<unknown>, line 1)

Dids commented 1 year ago

For what it's worth, I reverted from the latest version (v0.4.1) back to v0.4.0 (Docker container in my case) and that seems to still work fine, unless this is some kind of external sporadic issue.

kelteseth commented 1 year ago

@Dids same reverting to v0.4.0 worked for me

lucasmocellin commented 1 year ago

@Dids @kelteseth how did you guys revert the version and ran it? sorry very new to all this and tooling

jefvanlieshout commented 1 year ago

I got this fixed on ubuntu. ran the command "sudo chmod -R 777 autogpt/" while in the Auto-GPT folder.

lucasmocellin commented 1 year ago

@Dids @kelteseth how did you guys revert the version and ran it? sorry very new to all this and tooling

I actually changed i docker-compose.yml from: image: significantgravitas/auto-gpt to: image: significantgravitas/auto-gpt:v0.4.0

but then I got a new error: FAILED FILE VALIDATION The file prompt_settings.yaml wasn't found

I copied the file from the repo, still doesn't work ¯_(ツ)_/¯

homerdiaz commented 1 year ago

Same here. macOsBigSur/Docker/GPT3.5/Auto-gpt 4.1

2023-06-23 20:38:34,651 ERROR logs:_log:143 Error parsing JSON response with literal_eval invalid syntax (, line 1) 2023-06-23 20:38:34,656 ERROR logs:_log:143 JSON Validation Error: 'thoughts' is a required property

Failed validating 'required' in schema: {'$schema': 'http://json-schema.org/draft-07/schema#', 'additionalProperties': False, 'properties': {'command': {'additionalProperties': False, 'properties': {'args': {'type': 'object'}, 'name': {'type': 'string'}}, 'required': ['name', 'args'], 'type': 'object'}, 'thoughts': {'additionalProperties': False, 'properties': {'criticism': {'description': 'constructive ' 'self-criticism', 'type': 'string'}, 'plan': {'description': '- ' 'short ' 'bulleted\n' '- ' 'list ' 'that ' 'conveys\n' '- ' 'long-term ' 'plan', 'type': 'string'}, 'reasoning': {'type': 'string'}, 'speak': {'description': 'thoughts ' 'summary ' 'to ' 'say ' 'to ' 'user', 'type': 'string'}, 'text': {'description': 'thoughts', 'type': 'string'}}, 'required': ['text', 'reasoning', 'plan', 'criticism', 'speak'], 'type': 'object'}}, 'required': ['thoughts', 'command'], 'type': 'object'}

On instance: {}

Macavity77 commented 1 year ago

looks like reverting should work

CryptoRedneck commented 1 year ago

@Dids @kelteseth how did you guys revert the version and ran it? sorry very new to all this and tooling

I actually changed i docker-compose.yml from: image: significantgravitas/auto-gpt to: image: significantgravitas/auto-gpt:v0.4.0

but then I got a new error: FAILED FILE VALIDATION The file prompt_settings.yaml wasn't found

I copied the file from the repo, still doesn't work ¯(ツ)/¯

Same issue as @lucasmocellin

I was able to revert to 0.4.0, and then apply this hotfix:

katmai commented 1 year ago

Not sure if this is the same but i am getting a lot of these as well "Error parsing JSON response with literal_eval '{' was never closed (, line 1)"

QvQQ commented 11 months ago

Have the same issue. If anyone in the future find a way to solve it plz kick my ass.

VPetar commented 11 months ago

I'm also lining up here for ass-kicking like @QvQQ

AdamSoufiane commented 10 months ago

same issue after ti was working for a day or two.

ondrejlohnisky commented 10 months ago

I hope it helps someone: version significantgravitas/auto-gpt:v0.3.1 is the one that works for me. simply make a empty directory bot/ or something

create docker-compose.yaml

version: "3.9"
services:
auto-gpt:
image: significantgravitas/auto-gpt:v0.3.1
depends_on:
  redis:
    condition: service_started
env_file:
  - .env
environment:
  MEMORY_BACKEND: ${MEMORY_BACKEND:-redis}
  REDIS_HOST: ${REDIS_HOST:-redis}
profiles: ["exclude-from-up"]
volumes:
  - ./auto_gpt_workspace:/app/autogpt/auto_gpt_workspace
  - ./data:/app/data
  ## allow auto-gpt to write logs to disk
  - ./logs:/app/logs
  ## uncomment following lines if you want to make use of these files
  ## you must have them existing in the same folder as this docker-compose.yml
  #- type: bind
  #  source: ./azure.yaml
  #  target: /app/azure.yaml
  #- type: bind
  #  source: ./ai_settings.yaml
  #  target: /app/ai_settings.yaml
redis:
image: "redis/redis-stack-server:latest"

create .env i use LocalAI to host my local LLM's so set the proper url, for me i had to use SSL I created proxy pass with certbot for that. (It was refusing connection...)
```
OPENAI_API_KEY=sk---anystringhere
OPENAI_API_BASE=https://your_domain:443/v1
```
run sudo docker-compose run --rm auto-gpt --gpt3only in that directory

Hope it helps! 🥇

cfytrok commented 9 months ago

The problem is that gpt sometimes responds in markdown format and starts the response with "``json". To force responses in bare json, you can add the beginning of the response to the request. prompt.append(Message("assistant", """thoughts: {text:"""))` The problem is that the beginning of the response should be merged with the gpt response.

mariusorani commented 9 months ago

The problem is the OPENAI_FUNCTIONS setting. Turn OPENAI_FUNCTIONS=False and the SMART_LLM=gpt-4-0314

stompinghoss commented 8 months ago

Sorry this is a bit long, but I am trying to give a comprehensive explanation and what I tried. I have the exact same error as mentioned at the top of this thread.

I am getting an intermittent error as below. Further down, I’ve explained the different things I’ve tried but nothing seems to work. Any thoughts or ideas greatly appreciated.

I am trying to follow the tutorial at

https://lablab.ai/t/autogpt-tutorial-how-to-use-and-create-agent-for-coding-game

(though I did not follow their setup instructions. That was because initially, from my previous setup (which had been working for a week or so with various experiments, albeit, all in manual mode), I was not having errors. But since I started getting issues, I’ve tried all sorts of setup routes but everything is breaking with the above parsing error, usually after one or two attempts at the prompt mentioned below.

Here is the prompt, initial results and the error:

So at the prompt I enter: AI Agent for coding

And then I get the following output;

NOTE:All files/directories created by this agent can be found inside its workspace at:  /app/auto_gpt_workspace
CodeGeniusGPT  has been created with the following details:
Name:  CodeGeniusGPT
Role:  an advanced AI coding assistant that helps developers and programmers in writing efficient and error-free code by providing expert guidance, suggestions, and solutions for various programming languages and frameworks.
Goals: 
-  Assist in problem-solving, debugging, and troubleshooting to help you overcome coding challenges and improve the quality of your code.
-  Offer personalized code recommendations and best practices to optimize performance, enhance readability, and ensure adherence to coding standards.
-  Provide comprehensive documentation and explanations for complex concepts, algorithms, and design patterns to enhance your understanding and knowledge of coding principles.
-  Stay up-to-date with the latest programming trends, libraries, and frameworks to offer relevant and cutting-edge solutions for your coding needs.
-  Foster a collaborative coding environment by facilitating code reviews, pair programming, and knowledge sharing to promote continuous learning and growth.
  Could not load MemoryItems from file: input length is 0: line 1 column 1 (char 0)
Using memory of type:  JSONFileMemory
Using Browser:  chrome
  Error parsing JSON response with literal_eval invalid syntax (<unknown>, line 2)
Response could not be parsed: Validation of response failed:
  'thoughts' is a required property

Failed validating 'required' in schema:
    {'$schema': 'http://json-schema.org/draft-07/schema#',
     'additionalProperties': False,
     'properties': {'command': {'additionalProperties': False,
                                'properties': {'args': {'type': 'object'},
                                               'name': {'type': 'string'}},
                                'required': ['name', 'args'],
                                'type': 'object'},
                    'thoughts': {'additionalProperties': False,
                                 'properties': {'criticism': {'description': 'constructive '
                                                                             'self-criticism',
                                                              'type': 'string'},
                                                'plan': {'description': '- '
                                                                        'short '
                                                                        'bulleted\n'
                                                                        '- '
                                                                        'list '
                                                                        'that '
                                                                        'conveys\n'
                                                                        '- '
                                                                        'long-term '
                                                                        'plan',
                                                         'type': 'string'},
                                                'reasoning': {'type': 'string'},
                                                'speak': {'description': 'thoughts '
                                                                         'summary '
                                                                         'to '
                                                                         'say '
                                                                         'to '
                                                                         'user',
                                                          'type': 'string'},
                                                'text': {'description': 'thoughts',
                                                         'type': 'string'}},
                                 'required': ['text',
                                              'reasoning',
                                              'plan',
                                              'criticism',
                                              'speak'],
                                 'type': 'object'}},
     'required': ['thoughts', 'command'],
     'type': 'object'}

On instance:
    {};
  'command' is a required property

Failed validating 'required' in schema:
    {'$schema': 'http://json-schema.org/draft-07/schema#',
     'additionalProperties': False,
     'properties': {'command': {'additionalProperties': False,
                                'properties': {'args': {'type': 'object'},
                                               'name': {'type': 'string'}},
                                'required': ['name', 'args'],
                                'type': 'object'},
                    'thoughts': {'additionalProperties': False,
                                 'properties': {'criticism': {'description': 'constructive '
                                                                             'self-criticism',
                                                              'type': 'string'},
                                                'plan': {'description': '- '
                                                                        'short '
                                                                        'bulleted\n'
                                                                        '- '
                                                                        'list '
                                                                        'that '
                                                                        'conveys\n'
                                                                        '- '
                                                                        'long-term '
                                                                        'plan',
                                                         'type': 'string'},
                                                'reasoning': {'type': 'string'},
                                                'speak': {'description': 'thoughts '
                                                                         'summary '
                                                                         'to '
                                                                         'say '
                                                                         'to '
                                                                         'user',
                                                          'type': 'string'},
                                                'text': {'description': 'thoughts',
                                                         'type': 'string'}},
                                 'required': ['text',
                                              'reasoning',
                                              'plan',
                                              'criticism',
                                              'speak'],
                                 'type': 'object'}},
     'required': ['thoughts', 'command'],
     'type': 'object'}

On instance:
    {}  
CODEGENIUSGPT THOUGHTS: 
REASONING:  None
CRITICISM:  None
NO ACTION SELECTED:  The Agent failed to select an action.
  Enter 'y' to authorise command, 'y -N' to run N continuous commands, 'n' to exit program, or enter feedback for CodeGeniusGPT...
  Asking user via keyboard...
Input:

Since encountering this, I’ve wiped out my install and tried a few things, as below.

I have an openai API paid plan and key. To check it’s being detected in the .env file, if I take the key out, it’ll ask me to enter it at the command line. I then use the instructions at https://docs.agpt.co/setup/ There are several routes through. I do find them unclear. So here is what I tried (note I did also try instructions you can download on autogpt.net but same results):

First attempt:

Set up with Docker:

mkdir AutoGPT
cd AutoGPT

Create and save the following into docker-compose.yml (I have docker installed)

version: "3.9"
services:
  auto-gpt:
    image: significantgravitas/auto-gpt
    env_file:
      - .env
    profiles: ["exclude-from-up"]
    volumes:
      - ./auto_gpt_workspace:/app/auto_gpt_workspace
      - ./data:/app/data
      ## allow auto-gpt to write logs to disk
      - ./logs:/app/logs
      ## uncomment following lines if you want to make use of these files
      ## you must have them existing in the same folder as this docker-compose.yml
      #- type: bind
      #  source: ./azure.yaml
      #  target: /app/azure.yaml
      #- type: bind
      #  source: ./ai_settings.yaml
      #  target: /app/ai_settings.yaml

Save a .env file into the same location with the following changes from the .env.template

OPENAI_API_KEY=<my key>
OPENAI_ORGANIZATION=<my org>
SMART_LLM=gpt-4
FAST_LLM=gpt-3.5-turbo

Note, I’ve tried this with JSON file default and got similar issues.

REDIS_HOST=localhost
## REDIS_PORT - Redis port (Default: 6379)
REDIS_PORT=6379

## REDIS_PASSWORD - Redis password (Default: "")
REDIS_PASSWORD=<my password>

## MEMORY_BACKEND - Memory backend type
# MEMORY_BACKEND=json_file

## MEMORY_INDEX - Value used in the Memory backend for scoping, naming, or indexing (Default: auto-gpt)
# MEMORY_INDEX=auto-gpt

## GOOGLE_API_KEY - Google API key (Default: None)
GOOGLE_API_KEY=<my key>

## GOOGLE_CUSTOM_SEARCH_ENGINE_ID - Google custom search engine ID (Default: None)
GOOGLE_CUSTOM_SEARCH_ENGINE_ID=<my id>

ELEVENLABS_API_KEY=<my key>

then

docker pull significantgravitas/auto-gpt Then onto the run with docker section

docker compose version
Docker Compose version v2.22.0-desktop.2

So I think that’s ok.

docker compose build auto-gpt (actually, thinking about it, that is maybe redundant...)

Just gives [+] Building 0.0s (0/0)

Then do

docker compose run --rm auto-gpt

And that will seem to startup:

[+] Building 0.0s (0/0)                                                                                                                                                  docker:desktop-linux
[+] Building 0.0s (0/0)                                                                                                                                                  docker:desktop-linux
  plugins_config.yaml does not exist, creating base config.
NEWS:  Welcome to Auto-GPT!
NEWS:  
NEWS:  
Welcome to Auto-GPT!  run with '--help' for more information.
Create an AI-Assistant:  input '--manual' to enter manual mode.
  Asking user via keyboard...
I want Auto-GPT to:

So at the prompt I enter: AI Agent for coding

Then I get a long error above.

I then try changing the .env by commenting out all the REDIS related lines. I gave it the same prompt. Result is the same error. So that route seems to be a failure.

Running it multiple times, I found the error is intermittent.

So then wiped all that and started through the git route from the agpt docs:

git clone https://github.com/Significant-Gravitas/AutoGPT.git

Gives the error

warning: the following paths have collided (e.g. case-sensitive paths
on a case-insensitive filesystem) and only one from the same
colliding group is in the working tree:

  'arena/TestAgent.json'
  'arena/testAgent.json'

Tried a few things to get to the stable branch but probably did it wrong so wanted to start fresh with what I thought more likely to work;

git clone -b stable --single-branch https://github.com/Significant-Gravitas/AutoGPT.git Gives:

Cloning into 'AutoGPT'...
remote: Enumerating objects: 12692, done.
remote: Total 12692 (delta 0), reused 0 (delta 0), pack-reused 12692
Receiving objects: 100% (12692/12692), 4.98 MiB | 5.93 MiB/s, done.
Resolving deltas: 100% (8551/8551), done.

Seemed to work. Put the .env back in.

docker compose build auto-gpt Seemed to work:

docker compose run --rm auto-gpt Gives:

[+] Building 0.0s (0/0)                                                                                                                                                  docker:desktop-linux
[+] Building 0.0s (0/0)                                                                                                                                                  docker:desktop-linux
  plugins_config.yaml does not exist, creating base config.
NEWS:  Welcome to Auto-GPT!
NEWS:  
NEWS:  
Welcome to Auto-GPT!  run with '--help' for more information.
Create an AI-Assistant:  input '--manual' to enter manual mode.
  Asking user via keyboard...
I want Auto-GPT to: AI Agent for coding
NOTE:All files/directories created by this agent can be found inside its workspace at:  /app/auto_gpt_workspace
CodeGeniusGPT  has been created with the following details:
Name:  CodeGeniusGPT
Role:  an AI coding assistant that helps developers and programmers in writing efficient and error-free code by providing expert guidance, suggestions, and solutions for various programming languages and frameworks.
Goals: 
-  Assist in problem-solving, debugging, and troubleshooting to help you overcome coding challenges and improve the quality of your code.
-  Offer personalized code recommendations and best practices to optimize performance, enhance readability, and ensure adherence to coding standards.
-  Provide comprehensive documentation and explanations for complex concepts, algorithms, and design patterns to enhance your understanding and knowledge of coding principles.
-  Stay up-to-date with the latest programming languages, frameworks, and tools to offer relevant and cutting-edge solutions for your coding needs.
-  Foster a collaborative coding environment by facilitating code reviews, pair programming, and knowledge sharing to promote continuous learning and growth.
  Could not load MemoryItems from file: input length is 0: line 1 column 1 (char 0)
Using memory of type:  JSONFileMemory
Using Browser:  chrome
| Thinking...

Then I intermittently get the same error. Tried doing ./run.sh but same results, i.e. intermittent error.

As per someone’s suggestion above, I tried: SMART_LLM=gpt-4-0314 But I just get an error saying I don’t have access to it. Per another suggestion, I also tried: OPENAI_FUNCTIONS=False But just got the same parsing error.

Any ideas greatly appreciated.

Achulhu commented 8 months ago

Sorry this is a bit long, but I am trying to give a comprehensive explanation and what I tried. I have the exact same error as mentioned at the top of this thread.

I am getting an intermittent error as below. Further down, I’ve explained the different things I’ve tried but nothing seems to work. Any thoughts or ideas greatly appreciated.

I am trying to follow the tutorial at

https://lablab.ai/t/autogpt-tutorial-how-to-use-and-create-agent-for-coding-game

(though I did not follow their setup instructions. That was because initially, from my previous setup (which had been working for a week or so with various experiments, albeit, all in manual mode), I was not having errors. But since I started getting issues, I’ve tried all sorts of setup routes but everything is breaking with the above parsing error, usually after one or two attempts at the prompt mentioned below.

Here is the prompt, initial results and the error:

So at the prompt I enter: AI Agent for coding

And then I get the following output;

NOTE:All files/directories created by this agent can be found inside its workspace at:  /app/auto_gpt_workspace
CodeGeniusGPT  has been created with the following details:
Name:  CodeGeniusGPT
Role:  an advanced AI coding assistant that helps developers and programmers in writing efficient and error-free code by providing expert guidance, suggestions, and solutions for various programming languages and frameworks.
Goals: 
-  Assist in problem-solving, debugging, and troubleshooting to help you overcome coding challenges and improve the quality of your code.
-  Offer personalized code recommendations and best practices to optimize performance, enhance readability, and ensure adherence to coding standards.
-  Provide comprehensive documentation and explanations for complex concepts, algorithms, and design patterns to enhance your understanding and knowledge of coding principles.
-  Stay up-to-date with the latest programming trends, libraries, and frameworks to offer relevant and cutting-edge solutions for your coding needs.
-  Foster a collaborative coding environment by facilitating code reviews, pair programming, and knowledge sharing to promote continuous learning and growth.
  Could not load MemoryItems from file: input length is 0: line 1 column 1 (char 0)
Using memory of type:  JSONFileMemory
Using Browser:  chrome
  Error parsing JSON response with literal_eval invalid syntax (<unknown>, line 2)
Response could not be parsed: Validation of response failed:
  'thoughts' is a required property

Failed validating 'required' in schema:
    {'$schema': 'http://json-schema.org/draft-07/schema#',
     'additionalProperties': False,
     'properties': {'command': {'additionalProperties': False,
                                'properties': {'args': {'type': 'object'},
                                               'name': {'type': 'string'}},
                                'required': ['name', 'args'],
                                'type': 'object'},
                    'thoughts': {'additionalProperties': False,
                                 'properties': {'criticism': {'description': 'constructive '
                                                                             'self-criticism',
                                                              'type': 'string'},
                                                'plan': {'description': '- '
                                                                        'short '
                                                                        'bulleted\n'
                                                                        '- '
                                                                        'list '
                                                                        'that '
                                                                        'conveys\n'
                                                                        '- '
                                                                        'long-term '
                                                                        'plan',
                                                         'type': 'string'},
                                                'reasoning': {'type': 'string'},
                                                'speak': {'description': 'thoughts '
                                                                         'summary '
                                                                         'to '
                                                                         'say '
                                                                         'to '
                                                                         'user',
                                                          'type': 'string'},
                                                'text': {'description': 'thoughts',
                                                         'type': 'string'}},
                                 'required': ['text',
                                              'reasoning',
                                              'plan',
                                              'criticism',
                                              'speak'],
                                 'type': 'object'}},
     'required': ['thoughts', 'command'],
     'type': 'object'}

On instance:
    {};
  'command' is a required property

Failed validating 'required' in schema:
    {'$schema': 'http://json-schema.org/draft-07/schema#',
     'additionalProperties': False,
     'properties': {'command': {'additionalProperties': False,
                                'properties': {'args': {'type': 'object'},
                                               'name': {'type': 'string'}},
                                'required': ['name', 'args'],
                                'type': 'object'},
                    'thoughts': {'additionalProperties': False,
                                 'properties': {'criticism': {'description': 'constructive '
                                                                             'self-criticism',
                                                              'type': 'string'},
                                                'plan': {'description': '- '
                                                                        'short '
                                                                        'bulleted\n'
                                                                        '- '
                                                                        'list '
                                                                        'that '
                                                                        'conveys\n'
                                                                        '- '
                                                                        'long-term '
                                                                        'plan',
                                                         'type': 'string'},
                                                'reasoning': {'type': 'string'},
                                                'speak': {'description': 'thoughts '
                                                                         'summary '
                                                                         'to '
                                                                         'say '
                                                                         'to '
                                                                         'user',
                                                          'type': 'string'},
                                                'text': {'description': 'thoughts',
                                                         'type': 'string'}},
                                 'required': ['text',
                                              'reasoning',
                                              'plan',
                                              'criticism',
                                              'speak'],
                                 'type': 'object'}},
     'required': ['thoughts', 'command'],
     'type': 'object'}

On instance:
    {}  
CODEGENIUSGPT THOUGHTS: 
REASONING:  None
CRITICISM:  None
NO ACTION SELECTED:  The Agent failed to select an action.
  Enter 'y' to authorise command, 'y -N' to run N continuous commands, 'n' to exit program, or enter feedback for CodeGeniusGPT...
  Asking user via keyboard...
Input:

Since encountering this, I’ve wiped out my install and tried a few things, as below.

I have an openai API paid plan and key. To check it’s being detected in the .env file, if I take the key out, it’ll ask me to enter it at the command line. I then use the instructions at https://docs.agpt.co/setup/ There are several routes through. I do find them unclear. So here is what I tried (note I did also try instructions you can download on autogpt.net but same results):

First attempt:

Set up with Docker:

mkdir AutoGPT
cd AutoGPT

Create and save the following into docker-compose.yml (I have docker installed)

version: "3.9"
services:
  auto-gpt:
    image: significantgravitas/auto-gpt
    env_file:
      - .env
    profiles: ["exclude-from-up"]
    volumes:
      - ./auto_gpt_workspace:/app/auto_gpt_workspace
      - ./data:/app/data
      ## allow auto-gpt to write logs to disk
      - ./logs:/app/logs
      ## uncomment following lines if you want to make use of these files
      ## you must have them existing in the same folder as this docker-compose.yml
      #- type: bind
      #  source: ./azure.yaml
      #  target: /app/azure.yaml
      #- type: bind
      #  source: ./ai_settings.yaml
      #  target: /app/ai_settings.yaml

Save a .env file into the same location with the following changes from the .env.template

OPENAI_API_KEY=<my key>
OPENAI_ORGANIZATION=<my org>
SMART_LLM=gpt-4
FAST_LLM=gpt-3.5-turbo

Note, I’ve tried this with JSON file default and got similar issues.

REDIS_HOST=localhost
## REDIS_PORT - Redis port (Default: 6379)
REDIS_PORT=6379

## REDIS_PASSWORD - Redis password (Default: "")
REDIS_PASSWORD=<my password>

## MEMORY_BACKEND - Memory backend type
# MEMORY_BACKEND=json_file

## MEMORY_INDEX - Value used in the Memory backend for scoping, naming, or indexing (Default: auto-gpt)
# MEMORY_INDEX=auto-gpt

## GOOGLE_API_KEY - Google API key (Default: None)
GOOGLE_API_KEY=<my key>

## GOOGLE_CUSTOM_SEARCH_ENGINE_ID - Google custom search engine ID (Default: None)
GOOGLE_CUSTOM_SEARCH_ENGINE_ID=<my id>

ELEVENLABS_API_KEY=<my key>

then

docker pull significantgravitas/auto-gpt Then onto the run with docker section

docker compose version
Docker Compose version v2.22.0-desktop.2

So I think that’s ok.

docker compose build auto-gpt (actually, thinking about it, that is maybe redundant...)

Just gives [+] Building 0.0s (0/0)

Then do

docker compose run --rm auto-gpt

And that will seem to startup:

[+] Building 0.0s (0/0)                                                                                                                                                  docker:desktop-linux
[+] Building 0.0s (0/0)                                                                                                                                                  docker:desktop-linux
  plugins_config.yaml does not exist, creating base config.
NEWS:  Welcome to Auto-GPT!
NEWS:  
NEWS:  
Welcome to Auto-GPT!  run with '--help' for more information.
Create an AI-Assistant:  input '--manual' to enter manual mode.
  Asking user via keyboard...
I want Auto-GPT to:

So at the prompt I enter: AI Agent for coding

Then I get a long error above.

I then try changing the .env by commenting out all the REDIS related lines. I gave it the same prompt. Result is the same error. So that route seems to be a failure.

Running it multiple times, I found the error is intermittent.

So then wiped all that and started through the git route from the agpt docs:

git clone https://github.com/Significant-Gravitas/AutoGPT.git

Gives the error

warning: the following paths have collided (e.g. case-sensitive paths
on a case-insensitive filesystem) and only one from the same
colliding group is in the working tree:

  'arena/TestAgent.json'
  'arena/testAgent.json'

Tried a few things to get to the stable branch but probably did it wrong so wanted to start fresh with what I thought more likely to work;

git clone -b stable --single-branch https://github.com/Significant-Gravitas/AutoGPT.git Gives:

Cloning into 'AutoGPT'...
remote: Enumerating objects: 12692, done.
remote: Total 12692 (delta 0), reused 0 (delta 0), pack-reused 12692
Receiving objects: 100% (12692/12692), 4.98 MiB | 5.93 MiB/s, done.
Resolving deltas: 100% (8551/8551), done.

Seemed to work. Put the .env back in.

docker compose build auto-gpt Seemed to work:

docker compose run --rm auto-gpt Gives:

[+] Building 0.0s (0/0)                                                                                                                                                  docker:desktop-linux
[+] Building 0.0s (0/0)                                                                                                                                                  docker:desktop-linux
  plugins_config.yaml does not exist, creating base config.
NEWS:  Welcome to Auto-GPT!
NEWS:  
NEWS:  
Welcome to Auto-GPT!  run with '--help' for more information.
Create an AI-Assistant:  input '--manual' to enter manual mode.
  Asking user via keyboard...
I want Auto-GPT to: AI Agent for coding
NOTE:All files/directories created by this agent can be found inside its workspace at:  /app/auto_gpt_workspace
CodeGeniusGPT  has been created with the following details:
Name:  CodeGeniusGPT
Role:  an AI coding assistant that helps developers and programmers in writing efficient and error-free code by providing expert guidance, suggestions, and solutions for various programming languages and frameworks.
Goals: 
-  Assist in problem-solving, debugging, and troubleshooting to help you overcome coding challenges and improve the quality of your code.
-  Offer personalized code recommendations and best practices to optimize performance, enhance readability, and ensure adherence to coding standards.
-  Provide comprehensive documentation and explanations for complex concepts, algorithms, and design patterns to enhance your understanding and knowledge of coding principles.
-  Stay up-to-date with the latest programming languages, frameworks, and tools to offer relevant and cutting-edge solutions for your coding needs.
-  Foster a collaborative coding environment by facilitating code reviews, pair programming, and knowledge sharing to promote continuous learning and growth.
  Could not load MemoryItems from file: input length is 0: line 1 column 1 (char 0)
Using memory of type:  JSONFileMemory
Using Browser:  chrome
| Thinking...

Then I intermittently get the same error. Tried doing ./run.sh but same results, i.e. intermittent error.

As per someone’s suggestion above, I tried: SMART_LLM=gpt-4-0314 But I just get an error saying I don’t have access to it. Per another suggestion, I also tried: OPENAI_FUNCTIONS=False But just got the same parsing error.

Any ideas greatly appreciated.

Tried many of the same steps and suggestions, ended up here as well. For the amount of people using AutoGPT successfully and the amount encountering the same error there is clearly a missing link somewhere. I can't event get --help to successfully load all the way.

mohcineelharras commented 8 months ago

I hope it helps someone: version significantgravitas/auto-gpt:v0.3.1 is the one that works for me. simply make a empty directory bot/ or something

create docker-compose.yaml
version: "3.9"
services:
  auto-gpt:
    image: significantgravitas/auto-gpt:v0.3.1
    depends_on:
      redis:
        condition: service_started
    env_file:
      - .env
    environment:
      MEMORY_BACKEND: ${MEMORY_BACKEND:-redis}
      REDIS_HOST: ${REDIS_HOST:-redis}
    profiles: ["exclude-from-up"]
    volumes:
      - ./auto_gpt_workspace:/app/autogpt/auto_gpt_workspace
      - ./data:/app/data
      ## allow auto-gpt to write logs to disk
      - ./logs:/app/logs
      ## uncomment following lines if you want to make use of these files
      ## you must have them existing in the same folder as this docker-compose.yml
      #- type: bind
      #  source: ./azure.yaml
      #  target: /app/azure.yaml
      #- type: bind
      #  source: ./ai_settings.yaml
      #  target: /app/ai_settings.yaml
  redis:
    image: "redis/redis-stack-server:latest"
create .env i use LocalAI to host my local LLM's so set the proper url, for me i had to use SSL I created proxy pass with certbot for that. (It was refusing connection...)
OPENAI_API_KEY=sk---anystringhere
OPENAI_API_BASE=https://your_domain:443/v1
run sudo docker-compose run --rm auto-gpt --gpt3only in that directory

Hope it helps! 🥇

when following exactly what you said I am able to run on LM_STUDIO but still have a little problem

Using memory of type: RedisMemory Using Browser: chrome Apparently json was fixed. The JSON object is invalid.
THOUGHTS: None REASONING: None CRITICISM: None NEXT ACTION: COMMAND = Error: ARGUMENTS = Missing 'command' object in JSON Enter 'y' to authorise command, 'y -N' to run N continuous commands, 's' to run self-feedback commands, 'n' to exit program, or enter feedback for ... Asking user via keyboard... Input:-y SYSTEM: Human feedback: -y Apparently json was fixed. The JSON object is invalid.
THOUGHTS: None REASONING: None CRITICISM: None NEXT ACTION: COMMAND = Error: ARGUMENTS = Missing 'command' object in JSON Enter 'y' to authorise command, 'y -N' to run N continuous commands, 's' to run self-feedback commands, 'n' to exit program, or enter feedback for ... Asking user via keyboard...

mohcineelharras commented 8 months ago

I hope it helps someone: version significantgravitas/auto-gpt:v0.3.1 is the one that works for me. simply make a empty directory bot/ or something

create docker-compose.yaml
version: "3.9"
services:
  auto-gpt:
    image: significantgravitas/auto-gpt:v0.3.1
    depends_on:
      redis:
        condition: service_started
    env_file:
      - .env
    environment:
      MEMORY_BACKEND: ${MEMORY_BACKEND:-redis}
      REDIS_HOST: ${REDIS_HOST:-redis}
    profiles: ["exclude-from-up"]
    volumes:
      - ./auto_gpt_workspace:/app/autogpt/auto_gpt_workspace
      - ./data:/app/data
      ## allow auto-gpt to write logs to disk
      - ./logs:/app/logs
      ## uncomment following lines if you want to make use of these files
      ## you must have them existing in the same folder as this docker-compose.yml
      #- type: bind
      #  source: ./azure.yaml
      #  target: /app/azure.yaml
      #- type: bind
      #  source: ./ai_settings.yaml
      #  target: /app/ai_settings.yaml
  redis:
    image: "redis/redis-stack-server:latest"
create .env i use LocalAI to host my local LLM's so set the proper url, for me i had to use SSL I created proxy pass with certbot for that. (It was refusing connection...)
OPENAI_API_KEY=sk---anystringhere
OPENAI_API_BASE=https://your_domain:443/v1
run sudo docker-compose run --rm auto-gpt --gpt3only in that directory

Hope it helps! 🥇

Yo man, I am trying localai too now could you tell if you have gpu acceleration ? which embedding model did you choose ? which llm did you work with ?

jamesk9526 commented 8 months ago

I cannot Get this to work- same thing.

here is an exsample --> All packages are installed. WARNING: You do not have access to gpt-3.5-turbo. Setting fast_llm to gpt-3.5-turbo. WARNING: You do not have access to gpt-4-0314. Setting smart_llm to gpt-3.5-turbo. NEWS: Welcome to Auto-GPT! NEWS: NEWS: Welcome back! Would you like me to return to being BMIGPT? Asking user via keyboard... Continue with the last settings? Name: BMIGPT Role: A knowledgeable and empathetic assistant who helps users track their body mass index (BMI) and provides personal Goals: ["Accurately calculate the user's BMI based on their height, weight, and age.", 'Provide clear and concise expla API Budget: infinite Continue (y/n): y NOTE:All files/directories created by this agent can be found inside its workspace at: C:\Users\James Knox\Downloads\AutoGPT-stable\auto_gpt_workspace BMIGPT has been created with the following details: Name: BMIGPT Role: A knowledgeable and empathetic assistant who helps users track their body mass index (BMI) and provides personalized recommendations for a healthy lifestyle. Goals:

Accurately calculate the user's BMI based on their height, weight, and age.
Provide clear and concise explanations of BMI categories and their associated health risks.
Offer actionable tips and recommendations for maintaining a healthy weight within each BMI category.
Encourage users to set and track progress towards their health goals through regular check-ins and motivational support. Using memory of type: JSONFileMemory Using Browser: chrome Failed to update API costs: KeyError: 'E:\Freedom GPT\TheBloke\Llama-2-7B-Chat-GGML\llama-2-7b-chat.ggmlv3.q3_K_L.bin' Error parsing JSON response with literal_eval unterminated string literal (detected at line 1) (, line 1) Response could not be parsed: Validation of response failed: 'thoughts' is a required property

Failed validating 'required' in schema: {'$schema': 'http://json-schema.org/draft-07/schema#', 'additionalProperties': False, 'properties': {'command': {'additionalProperties': False, 'properties': {'args': {'type': 'object'}, 'name': {'type': 'string'}}, 'required': ['name', 'args'], 'type': 'object'}, 'thoughts': {'additionalProperties': False, 'properties': {'criticism': {'description': 'constructive ' 'self-criticism', 'type': 'string'}, 'plan': {'description': '- ' 'short ' 'bulleted\n' '- ' 'list ' 'that ' 'conveys\n' '- ' 'long-term ' 'plan', 'type': 'string'}, 'reasoning': {'type': 'string'}, 'speak': {'description': 'thoughts ' 'summary ' 'to ' 'say ' 'to ' 'user', 'type': 'string'}, 'text': {'description': 'thoughts', 'type': 'string'}}, 'required': ['text', 'reasoning', 'plan', 'criticism', 'speak'], 'type': 'object'}}, 'required': ['thoughts', 'command'], 'type': 'object'}

On instance: {}; 'command' is a required property

Failed validating 'required' in schema: {'$schema': 'http://json-schema.org/draft-07/schema#', 'additionalProperties': False, 'properties': {'command': {'additionalProperties': False, 'properties': {'args': {'type': 'object'}, 'name': {'type': 'string'}}, 'required': ['name', 'args'], 'type': 'object'}, 'thoughts': {'additionalProperties': False, 'properties': {'criticism': {'description': 'constructive ' 'self-criticism', 'type': 'string'}, 'plan': {'description': '- ' 'short ' 'bulleted\n' '- ' 'list ' 'that ' 'conveys\n' '- ' 'long-term ' 'plan', 'type': 'string'}, 'reasoning': {'type': 'string'}, 'speak': {'description': 'thoughts ' 'summary ' 'to ' 'say ' 'to ' 'user', 'type': 'string'}, 'text': {'description': 'thoughts', 'type': 'string'}}, 'required': ['text', 'reasoning', 'plan', 'criticism', 'speak'], 'type': 'object'}}, 'required': ['thoughts', 'command'], 'type': 'object'}

On instance: {} BMIGPT THOUGHTS: REASONING: None CRITICISM: None NO ACTION SELECTED: The Agent failed to select an action. Enter 'y' to authorise command, 'y -N' to run N continuous commands, 'n' to exit program, or enter feedback for BMIGPT... Asking user via keyboard... Input:

jamesk9526 commented 8 months ago

I figured This Out --> Failed validating 'required' in schema: {'$schema': 'http://json-schema.org/draft-07/schema#', 'additionalProperties': False, 'properties': {'command': {'additionalProperties': False, 'properties': {'args': {'type': 'object'}, 'name': {'type': 'string'}}, 'required': ['name', 'args'], 'type': 'object'}, 'thoughts': {'additionalProperties': False, 'properties': {'criticism': {'description': 'constructive ' 'self-criticism', 'type': 'string'}, 'plan': {'description': '- ' 'short ' 'bulleted\n' '- ' 'list ' 'that ' 'conveys\n' '- ' 'long-term ' 'plan', 'type': 'string'}, 'reasoning': {'type': 'string'}, 'speak': {'description': 'thoughts ' 'summary ' 'to ' 'say ' 'to ' 'user', 'type': 'string'}, 'text': {'description': 'thoughts', 'type': 'string'}}, 'required': ['text', 'reasoning', 'plan', 'criticism', 'speak'], 'type': 'object'}}, 'required': ['thoughts', 'command'], 'type': 'object'}

On instance: {}

If you are using a local server or whatever your server settings for your model is. Sure that in the Prompt formatting it is not inserting new lines.

It worked for me give it a try!

jshingler commented 7 months ago

The problem is the OPENAI_FUNCTIONS setting. Turn OPENAI_FUNCTIONS=False and the SMART_LLM=gpt-4-0314

Seems to help

rodrigolopezguerra commented 7 months ago

https://github.com/Significant-Gravitas/AutoGPT/issues/6320

rodrigolopezguerra commented 7 months ago

The problem is the OPENAI_FUNCTIONS setting. Turn OPENAI_FUNCTIONS=False and the SMART_LLM=gpt-4-0314

Maybe this is working because this is realated to : https://platform.openai.com/docs/guides/function-calling We are aware of an issue with non-ASCII outputs in gpt-3.5-turbo-1106 and gpt-4-1106-preview, and are working on implementing a fix. When these models generate a function call and the arguments include non-ASCII characters, the API may return Unicode escape sequences instead of the Unicode character directly. For example, arguments may look like {"location": "D\u00fcsseldorf"} instead of {"location": "Düsseldorf"}. Most applications should not be affected by this, as JSON parsers in languages like Python and Javascript will parse these strings into the correct objects. To stay updated on this topic, please subscribe to [this community forum thread](https://community.openai.com/t/gpt-4-1106-preview-messes-up-function-call-parameters-encoding/478500).

github-actions[bot] commented 4 months ago

This issue has automatically been marked as stale because it has not had any activity in the last 50 days. You can unstale it by commenting or removing the label. Otherwise, this issue will be closed in 10 days.

github-actions[bot] commented 4 months ago

This issue was closed automatically because it has been stale for 10 days with no activity.

Significant-Gravitas / AutoGPT