GPT-Professor

Tool used to create a simple CLI around ChatGPT, in order to help any kind of teaching activity

Installation

You need to install Node.js After that, you can install the tool using npm i -g gpt-professor Make sure everything is working:

$ gpt-professor --version
0.0.8

Tools

Measure percentage of generated content

• Name: measure-generated-content
• Description: This tool will scan a folder (or a file) and will compute the percentage of AI generated content. Relies on zerogpt api.
• Arg: The path of the file or folder to scan
• Options:
  • -c, --concurrency: how many concurrent requests
• Example:

gpt-professor measure-generated-content /tmp/my-folder-to-scan

/tmp/my-folder-to-scan/file_test/.keep => 0.00% of potential AI generated content for 0 detected words
/tmp/my-folder-to-scan/file_test/rapport.pdf => 33.96% of potential AI generated content for 8503 detected words
/tmp/my-folder-to-scan/file_test/truc.txt => 0.00% of potential AI generated content for 187 detected words
/tmp/my-folder-to-scan/file_test/huc.pdf => 2.18% of potential AI generated content for 10010 detected words
/tmp/my-folder-to-scan/file_test/int.pdf => 12.99% of potential AI generated content for 12783 detected words
/tmp/my-folder-to-scan/file_test/lou.pdf => 9.41% of potential AI generated content for 6607 detected words

Autograde a repository with ChatGPT 4

• Name: autograde

• Description: This tool will send the repository content to the AI and will ask it to autograde the code based on criteria you have defined

• Options:

  • -c, --config: the autograde configuration path
  • -r, --repo: the repository filepath to autograde
  • -o, --output: where to store the autograde result (by default its output in the console)
  • -s, --minify: does the script try to shrink files before sending them to the AI (to reduce token and billing)

• Required env var: OPENAI_API_KEY (see https://help.openai.com/en/articles/4936850-where-do-i-find-my-secret-api-key)

• Optional env var: OPENAI_BASE_URL to set the endpoint to call (by default, its openai api production)

• Example: autograde.json (config file)

  {
      "files": {
          "include_patterns": [ "**/*.js", "**/*.vue" ],
          "exclude_patterns": [ "**/node_modules/**", "babel.config.js" ]
      },
      "criteria": {
          "use_vue_js": {
              "condition": "Make sure the project is using vuejs"
          },
          "use_computed_property": {
              "condition": "Make sure there is at least 1 computed property"
          },
          "use_at_least_2_components": {
              "condition": "Make sure there is at least 2 components defined and used"
          },
          "vue_router_is_used": {
              "condition": "Make sure it is using vue-router in App.vue with a least a router-link component"
          },
          "flutter": {
              "condition": "Make sure the project is using flutter"
          },
          "is_palindrome": {
              "condition": "Make sure there is a function used to check if a word is a palindrome"
          }
      }
  }

  Command:

  gpt-professor autograde -c file_test/autograde.json -r file_test/tp6 -o file_test/professorgpt.json
  
  Will send files to AI: [file_test/tp6/src/App.vue,file_test/tp6/src/main.js,file_test/tp6/src/router.js,file_test/tp6/src/store.js,file_test/tp6/src/components/TodoDetail.vue,file_test/tp6/src/components/home.vue,file_test/tp6/src/components/task.vue]
  
  Result:
  {"use_vue_js": {"status": "PASS"}, 
  "use_computed_property": {"status": "PASS"},
  "use_at_least_2_components": {"status": "PASS"},
  "vue_router_is_used": {"status": "PASS"},
  "flutter": {"status": "FAIL"},
  "is_palindrome": {"status": "FAIL"}}

Autograde a repository with Gemini

• Name: autograde-gemini

• Description: This tool will send the repository content to the AI and will ask it to autograde the code based on criteria you have defined

• Options:

  • -c, --config: the autograde configuration path
  • -r, --repo: the repository filepath to autograde
  • -o, --output: where to store the autograde result (by default its output in the console)
  • -s, --minify: does the script try to shrink files before sending them to the AI (to reduce token and billing)

• Required env var: GEMINI_API_KEY

• Example: autograde.json (config file)

  {
      "files": {
          "include_patterns": [ "**/*.js", "**/*.vue" ],
          "exclude_patterns": [ "**/node_modules/**", "babel.config.js" ]
      },
      "criteria": {
          "use_vue_js": {
              "condition": "Make sure the project is using vuejs"
          },
          "use_computed_property": {
              "condition": "Make sure there is at least 1 computed property"
          },
          "use_at_least_2_components": {
              "condition": "Make sure there is at least 2 components defined and used"
          },
          "vue_router_is_used": {
              "condition": "Make sure it is using vue-router in App.vue with a least a router-link component"
          },
          "flutter": {
              "condition": "Make sure the project is using flutter"
          },
          "is_palindrome": {
              "condition": "Make sure there is a function used to check if a word is a palindrome"
          }
      }
  }

  Command:

  gpt-professor autograde-gemini -c file_test/autograde.json -r file_test/tp6 -o file_test/professorgpt.json
  
  Will send files to AI: [file_test/tp6/src/App.vue,file_test/tp6/src/main.js,file_test/tp6/src/router.js,file_test/tp6/src/store.js,file_test/tp6/src/components/TodoDetail.vue,file_test/tp6/src/components/home.vue,file_test/tp6/src/components/task.vue]
  
  Result:
  {"use_vue_js": {"status": "PASS"}, 
  "use_computed_property": {"status": "PASS"},
  "use_at_least_2_components": {"status": "PASS"},
  "vue_router_is_used": {"status": "PASS"},
  "flutter": {"status": "FAIL"},
  "is_palindrome": {"status": "FAIL"}}

Give a feedback on a repository with ChatGPT4

• Name: feedback

• Description: Automatically give a feedback to a Github repository

• Options:

  • -c, --config: the feedback configuration path
  • -r, --repo: the repository url (https or ssh) to feedback
  • -s, --minify: does the script try to shrink files before sending them to the AI (to reduce token and billing)

• Required env var: OPENAI_API_KEY (see https://help.openai.com/en/articles/4936850-where-do-i-find-my-secret-api-key)

• Required env var: GIT_PAT (Github Personal Access Token, with issue write, and repo read permissions)

• Optional env var: OPENAI_BASE_URL to set the endpoint to call (by default, its openai api production)

• Example: feedback.json (config file)

  {
      "files": {
          "include_patterns": [ "**/*.js", "**/*.vue" ],
          "exclude_patterns": [ "**/node_modules/**", "babel.config.js" ]
      },
      "criteria": {
          "use_vue_js": {
              "name": "Framework VUE.JS is used",
              "condition": "Make sure the project is using vuejs"
          },
          "use_computed_property": {
              "name": "Computed properties should be used",
              "condition": "Make sure there is at least 1 computed property"
          }
      }
  }

  Command:

  gpt-professor feedback -c file_test/feedback.json -r https://github.com/jochy/gpt-professor --minify --debug
  
  Will send files to AI: [file_test/tp6/src/App.vue,file_test/tp6/src/main.js,file_test/tp6/src/router.js,file_test/tp6/src/store.js,file_test/tp6/src/components/TodoDetail.vue,file_test/tp6/src/components/home.vue,file_test/tp6/src/components/task.vue]
  Feedback issue created