Skip to content

Commit

Permalink
Update chatgpt tool to v2 (#1497)
Browse files Browse the repository at this point in the history 
* Use input.ext instead of element_identifier ext, add error for unsupported file

* return better error on wrong api key

* Fix error handling for missing assistant output

* add payment information in readme files

* Refactor to use type instead of ext, rename other types of files to txt ext.

* bump version to 2

* change question param to prompt

* Refactor chatgpt.xml to use 'prompt' instead of 'question' parameter

* fix typo, and using json format for context_files

* change question to prompt in help section and readme

* better readme and help section on explaining prompt

* change help of the prompt

* Apply suggestions from code review

Co-authored-by: Björn Grüning <[email protected]>

* remove duplicate file ext

---------

Co-authored-by: Björn Grüning <[email protected]>
  • Loading branch information
@arash77 @bgruening
arash77 and bgruening authored Sep 11, 2024
1 parent 36fa40d commit c21d9a2
Show file tree
Hide file tree
Showing 3 changed files with 74 additions and 70 deletions.
20 changes: 13 additions & 7 deletions tools/chatgpt/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,14 +2,16 @@

## What it does

This tool leverages OpenAI's ChatGPT API to generate responses based on user-provided context and questions.
Users can upload context data in various formats and ask questions related to that data.
This tool leverages OpenAI's ChatGPT API to generate responses based on user-provided context and prompt.
Users can upload context data in various formats and ask questions or execute prompts related to that data.
The tool then uploads the data to a OpenAI server and processes them using the selected ChatGPT model, returning an AI-generated response tailored to the context provided.

To utilize this tool, users need to input their OpenAI API key in the user preferences. To obtain an API key, visit https://platform.openai.com/account/api-keys.

Make sure to setup the payment method in your OpenAI account to use the API key in here: https://platform.openai.com/settings/organization/billing/

When you run this tool, your input data is sent to OpenAI's servers using your API-key.
OpenAI's models process the data and generate a response based on the context and question provided.
OpenAI's models process the data and generate a response based on the context and prompt provided.
After receiving the response from the OpenAI server, the tool returns it to Galaxy and puts it in your history.
The files that have been uploaded are then deleted from the OpenAI's server, so they are not stored beyond their necessary use.
If the tool fails to delete your uploaded files automatically, you can manually delete them by visiting https://platform.openai.com/storage/. You might want to check your OpenAI storage from time to time as they also have a quota.
Expand All @@ -19,17 +21,21 @@ If the tool fails to delete your uploaded files automatically, you can manually
**Input**

1. **Upload Context Data**: Users can upload up to 500 files in formats such as DOC, DOCX, HTML, JSON, PDF, TXT, JPG, JPEG, PNG, WEBP, or GIF.
This context data serves as the background information for the question you wish to ask.
This context data serves as the input for the prompt you wish to execute.

2. **Provide a Prompt**: Once the context data is added, users can provide a prompt for a task ChatGPT should execute.
The more specific the prompt, the more tailored the response will be.

[General thoughts on prompting with GPT-4](https://help.openai.com/en/articles/4936848-how-do-i-create-a-good-prompt-for-an-ai-model-like-gpt-4)

2. **Ask a Question**: Once the context data is added, users can pose a question related to the content.
The more specific the question, the more tailored the response will be.
[Open Ai's prompt example page for more information](https://platform.openai.com/docs/examples)

3. **Select a Model**: Choose the ChatGPT model that best fits your needs.
Information about different models and their pricing can be found at https://platform.openai.com/docs/models and https://openai.com/api/pricing.


**Output**

The output is a response generated by ChatGPT, crafted based on the provided context data and the question posed.
The output is a response generated by ChatGPT, crafted based on the provided context data and the prompt posed.
This response is saved in the `output.txt` file.

68 changes: 31 additions & 37 deletions tools/chatgpt/chatgpt.py
View file
Original file line number Diff line number Diff line change
@@ -1,9 +1,10 @@
import json
import os
import sys

from openai import OpenAI
from openai import AuthenticationError, OpenAI

context_files = sys.argv[1].split(",")
context_files = json.loads(sys.argv[1])
question = sys.argv[2]
model = sys.argv[3]
with open(sys.argv[4], "r") as f:
Expand All @@ -14,51 +15,39 @@

client = OpenAI(api_key=openai_api_key)

file_search_sup_ext = [
"c",
"cs",
"cpp",
"doc",
"docx",
"html",
"java",
"json",
"md",
"pdf",
"php",
"pptx",
"py",
"rb",
"tex",
"txt",
"css",
"js",
"sh",
"ts",
]

vision_sup_ext = ["jpg", "jpeg", "png", "webp", "gif"]

file_search_file_streams = []
image_files = []

for path in context_files:
ext = path.split(".")[-1].lower()
if ext in vision_sup_ext:
for path, type in context_files:
if type == "image":
if os.path.getsize(path) > 20 * 1024 * 1024:
print(f"File {path} exceeds the 20MB limit and will not be processed.")
sys.exit(1)
file = client.files.create(file=open(path, "rb"), purpose="vision")
promt = {"type": "image_file", "image_file": {"file_id": file.id}}
image_files.append(promt)
elif ext in file_search_sup_ext:
else:
file_search_file_streams.append(open(path, "rb"))

assistant = client.beta.assistants.create(
instructions="You will receive questions about files from file searches and image files. For file search queries, identify and retrieve the relevant files based on the question. For image file queries, analyze the image content and provide relevant information or insights based on the image data.",
model=model,
tools=[{"type": "file_search"}] if file_search_file_streams else [],
)
try:
assistant = client.beta.assistants.create(
instructions=(
"You will receive questions about files from file searches "
"and image files. For file search queries, identify and "
"retrieve the relevant files based on the question. "
"For image file queries, analyze the image content and "
"provide relevant information or insights based on the image data."
),
model=model,
tools=[{"type": "file_search"}] if file_search_file_streams else [],
)
except AuthenticationError as e:
print(f"Authentication error: {e.message}")
sys.exit(1)
except Exception as e:
print(f"An error occurred: {str(e)}")
sys.exit(1)

if file_search_file_streams:
vector_store = client.beta.vector_stores.create()
file_batch = client.beta.vector_stores.file_batches.upload_and_poll(
Expand Down Expand Up @@ -88,7 +77,12 @@
assistant_messages = list(
client.beta.threads.messages.list(thread_id=thread.id, run_id=run.id)
)

if not assistant_messages:
print(
"No output was generated!\nPlease ensure that your OpenAI account has sufficient credits.\n"
"You can check your balance here: https://platform.openai.com/settings/organization/billing"
)
sys.exit(1)
message_content = assistant_messages[0].content[0].text.value
print("Output has been saved!")
with open("output.txt", "w") as f:
Expand Down
56 changes: 30 additions & 26 deletions tools/chatgpt/chatgpt.xml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,29 +2,32 @@
<description>Integrating OpenAI's ChatGPT into Galaxy</description>
<macros>
<token name="@TOOL_VERSION@">2024</token>
<token name="@VERSION_SUFFIX@">1</token>
<token name="@VERSION_SUFFIX@">2</token>
</macros>
<requirements>
<requirement type="package" version="3.12">python</requirement>
<requirement type="package" version="1.35.13">openai</requirement>
</requirements>
<command detect_errors="exit_code"><![CDATA[
#import os
#import re
#set $LINK_LIST = []
#for $input in $context:
#set file_name, ext = os.path.splitext($input.element_identifier)
#if ext == ''
#set ext = '.'+$input.ext
#end if
#set LINK = re.sub('[^\w\-]', '_', str($file_name))+$ext
ln -s '$input' '$LINK' &&
$LINK_LIST.append($LINK)
#end for
#import json
#import os
#import re
#set LINK_LIST = []
#for $input in $context
#set file_name = os.path.splitext($input.element_identifier)[0]
## list of supported filetypes in OpenAI. If Galaxy has a filetype that is not in this list, just use the generic `txt`.
#set ext = $input.ext if $input.ext in ['c', 'cpp', 'css', 'csv', 'docx', 'gif', 'html', 'java', 'jpeg', 'jpg', 'js', 'json', 'md', 'pdf', 'php', 'pkl', 'png', 'pptx', 'py', 'rb', 'tar', 'tex', 'ts', 'txt', 'webp', 'xlsx', 'xml', 'zip'] else 'txt'
#set LINK = re.sub('[^\w\-]', '_', $file_name)+'.'+$ext
ln -s '$input' '$LINK' &&
## OpenAI has some special handling of "images", so let's annotate this here and pass it to the script
#set type = 'image' if $input.ext in ['jpg', 'jpeg', 'png', 'webp', 'gif'] else 'text'
$LINK_LIST.append([$LINK, $type])
#end for
#set context_files = json.dumps($LINK_LIST)
python '$__tool_directory__/chatgpt.py'
'$str(",".join($LINK_LIST))'
'$question'
'$context_files'
'$prompt'
'$model'
'$openai_api_key_file'
]]></command>
Expand All @@ -40,15 +43,15 @@ $__user__.extra_preferences.get('chatgpt|api_key', "")
<option value="all">All models</option>
</param>
<when value="vision">
<param name="model" type="select" optional="false" label="Model" help="Select the model you want to use">
<param name="model" type="select" optional="false" label="Model" help="Select the model you want to use.">
<option value="gpt-4o-mini" selected="true">Affordable and intelligent small model for fast, lightweight tasks (gpt-4o-mini)</option>
<option value="gpt-4o">High-intelligence flagship model for complex, multi-step tasks (gpt-4o)</option>
<option value="gpt-4-turbo">The previous set of high-intelligence model with vision capabilities (gpt-4-turbo)</option>
</param>
<param name="context" type="data" multiple="true" optional="false" format="doc,docx,html,json,pdf,txt,jpg,jpeg,png,webp,gif" label="Context" max="500" help="This data will be uploaded to OpenAI's servers for processing."/>
</when>
<when value="all">
<param name="model" type="select" optional="false" label="Model" help="Select the model you want to use">
<param name="model" type="select" optional="false" label="Model" help="Select the model you want to use.">
<option value="gpt-4o-mini" selected="true">Affordable and intelligent small model for fast, lightweight tasks (gpt-4o-mini)</option>
<option value="gpt-4o">High-intelligence flagship model for complex, multi-step tasks (gpt-4o)</option>
<option value="gpt-4-turbo">The previous set of high-intelligence model with vision capabilities (gpt-4-turbo)</option>
Expand All @@ -58,7 +61,7 @@ $__user__.extra_preferences.get('chatgpt|api_key', "")
<param name="context" type="data" multiple="true" optional="false" format="doc,docx,html,json,pdf,txt" label="Context" max="500" help="This data will be uploaded to OpenAI's servers for processing."/>
</when>
</conditional>
<param name="question" type="text" optional="false" label="Question" help="Question about the text provided." area="true">
<param name="prompt" type="text" optional="false" label="Prompt" help="Prompts or tasks you want ChatGPT to perform." area="true">
<validator type="empty_field"/>
</param>
</inputs>
Expand All @@ -68,7 +71,7 @@ $__user__.extra_preferences.get('chatgpt|api_key', "")
<tests>
<test expect_failure="true" expect_exit_code="1">
<param name="context" value="chatgpt_test.txt" ftype="txt"/>
<param name="question" value="What is this?"/>
<param name="prompt" value="What is this?"/>
<param name="model" value="gpt-4o-mini"/>
<assert_stdout>
<has_text text="OpenAI API key is not provided in user preferences!"/>
Expand All @@ -81,14 +84,15 @@ $__user__.extra_preferences.get('chatgpt|api_key', "")
**What it does**
This tool leverages OpenAI's ChatGPT API to generate responses based on user-provided context and questions.
Users can upload context data in various formats and ask questions related to that data.
This tool leverages OpenAI's ChatGPT API to generate responses based on user-provided context and prompt.
Users can upload context data in various formats and ask questions or execute prompts related to that data.
The tool then uploads the data to a OpenAI server and processes them using the selected ChatGPT model, returning an AI-generated response tailored to the context provided.
To utilize this tool, users need to input their OpenAI API key in the user preferences. To obtain an API key, visit API keys page in your OpenAI Dashboard.
Make sure to setup the payment method in your OpenAI account to use the API key.
When you run this tool, your input data is sent to OpenAI's servers using your API-key.
OpenAI's models process the data and generate a response based on the context and question provided.
OpenAI's models process the data and generate a response based on the context and prompt provided.
After receiving the response from the OpenAI server, the tool returns it to Galaxy and puts it in your history.
The files that have been uploaded are then deleted from the OpenAI's server, so they are not stored beyond their necessary use.
If the tool fails to delete your uploaded files automatically, you can manually delete them in your openai account page. You might want to check your OpenAI storage from time to time as they also have a quota.
Expand All @@ -103,18 +107,18 @@ Usage
**Input**
1. **Upload Context Data**: Users can upload up to 500 files in formats such as DOC, DOCX, HTML, JSON, PDF, TXT, JPG, JPEG, PNG, WEBP, or GIF.
This context data serves as the background information for the question you wish to ask.
This context data serves as the input for the prompt you wish to execute.
2. **Ask a Question**: Once the context data is added, users can pose a question related to the content.
The more specific the question, the more tailored the response will be.
2. **Provide a Prompt**: Once the context data is added, users can provide a prompt for a task ChatGPT should execute.
The more specific the prompt, the more tailored the response will be.
3. **Select a Model**: Choose the ChatGPT model that best fits your needs.
Information about different models and their pricing can be found on the OpenAI website.
**Output**
The output is a response generated by ChatGPT, crafted based on the provided context data and the question posed.
The output is a response generated by ChatGPT, crafted based on the provided context data and the prompt posed.
This response is saved in the `output.txt` file.
Expand Down

0 comments on commit c21d9a2

Please sign in to comment.