gpt-batch 0.1.0__py3-none-any.whl

gpt_batch/__init__.py

@@ -0,0 +1,3 @@
+from .batcher import GPTBatcher
+
+__all__ = ['GPTBatcher']

gpt_batch/batcher.py

@@ -0,0 +1,85 @@
+from openai import OpenAI
+from concurrent.futures import ThreadPoolExecutor, wait
+from functools import partial
+from tqdm import tqdm
+
+class GPTBatcher:
+    def __init__(self, api_key, model_name="gpt-3.5-turbo-0125", system_prompt="",temperature=1,num_workers=64,timeout_duration=60,retry_attempts=2):
+        self.client = OpenAI(api_key=api_key)
+        self.model_name = model_name
+        self.system_prompt = system_prompt
+        self.temperature = temperature
+        self.num_workers = num_workers
+        self.timeout_duration = timeout_duration
+        self.retry_attempts = retry_attempts
+        self.miss_index =[]
+
+    def get_attitude(self, ask_text):
+        index, ask_text = ask_text
+
+        completion = self.client.chat.completions.create(
+            model=self.model_name,
+            messages=[
+                {"role": "system", "content": self.system_prompt},
+                {"role": "user", "content": ask_text}
+            ],
+            temperature=self.temperature,
+        )
+        return (index, completion.choices[0].message.content)
+
+    def process_attitude(self, message_list):
+        new_list = []
+        num_workers = self.num_workers
+        timeout_duration = self.timeout_duration
+        retry_attempts=2
+
+        executor = ThreadPoolExecutor(max_workers=num_workers)
+        message_chunks = list(self.chunk_list(message_list, num_workers))
+        for chunk in tqdm(message_chunks, desc="Processing messages"):
+            future_to_message =  {executor.submit(self.get_attitude, message): message for message in chunk}
+            for _ in range(retry_attempts):
+                done, not_done = wait(future_to_message.keys(), timeout=timeout_duration)
+                for future in not_done:
+                    future.cancel()
+                new_list.extend(future.result() for future in done if future.done())
+                if len(not_done) == 0:
+                    break
+                future_to_message = {executor.submit(self.get_attitude, (future_to_message[future], msg), temperature): future_to_message[future] for future, msg in not_done}
+        executor.shutdown(wait=False)
+        return new_list
+
+    def complete_attitude_list(self,attitude_list, max_length):
+        completed_list = []
+        current_index = 0
+        for item in attitude_list:
+            index, value = item
+            # Fill in missing indices
+            while current_index < index:
+                completed_list.append((current_index, None))
+                current_index += 1
+            # Add the current element from the list
+            completed_list.append(item)
+            current_index = index + 1
+        while current_index < max_length:
+            print("Filling in missing index", current_index)
+            self.miss_index.append(current_index)
+            completed_list.append((current_index, None))
+            current_index += 1
+        return completed_list
+
+    def chunk_list(self, lst, n):
+        """Yield successive n-sized chunks from lst."""
+        for i in range(0, len(lst), n):
+            yield lst[i:i + n]
+
+    def handle_message_list(self,message_list):
+        indexed_list = [(index, data) for index, data in enumerate(message_list)]
+        max_length = len(indexed_list)
+        attitude_list = self.process_attitude(indexed_list)
+        attitude_list.sort(key=lambda x: x[0])
+        attitude_list = self.complete_attitude_list(attitude_list, max_length)
+        attitude_list = [x[1] for x in attitude_list]
+        return attitude_list
+
+    # Add other necessary methods similar to the above, refactored to fit within this class structure.
+

gpt_batch-0.1.0.dist-info/METADATA

@@ -0,0 +1,30 @@
+Metadata-Version: 2.1
+Name: gpt-batch
+Version: 0.1.0
+Summary: A package for batch processing with OpenAI API.
+Home-page: https://github.com/fengsxy/gpt_batch
+Author: Ted Yu
+Author-email: liddlerain@gmail.com
+Description-Content-Type: text/markdown
+Requires-Dist: openai
+Requires-Dist: tqdm
+
+# GPT Batcher
+
+A simple tool to batch process messages using OpenAI's GPT models.
+
+## Installation
+
+Clone this repository and run:
+
+## Usage
+
+Here's how to use the `GPTBatcher`:
+
+```python
+from gpt_batch.batcher import GPTBatcher
+
+batcher = GPTBatcher(key='your_key_here', model_name='gpt-3.5-turbo-1106')
+result = batcher.handle_list(['your', 'list', 'of', 'messages'])
+print(result)
+

gpt_batch-0.1.0.dist-info/RECORD

@@ -0,0 +1,8 @@
+gpt_batch/__init__.py,sha256=pPeIKAbsqVso-jQOnhXWABziotoABCeDR0QrwyzPGv0,61
+gpt_batch/batcher.py,sha256=kzL0OT8dpWUEOiKMs8BptqVGBrlLVcv4UWrWfEEihjg,3720
+tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+tests/test_batcher.py,sha256=uq4GUAtlpofAFRMAUJOCSp5V-qBXpbVR26e_9k5h7-I,1117
+gpt_batch-0.1.0.dist-info/METADATA,sha256=Ns-Dz7Z4cIZFfmp9DpAFpF6zLeaAFa3DWtsNWPv0Peg,707
+gpt_batch-0.1.0.dist-info/WHEEL,sha256=yQN5g4mg4AybRjkgi-9yy4iQEFibGQmlz78Pik5Or-A,92
+gpt_batch-0.1.0.dist-info/top_level.txt,sha256=FtvJB_L9W_S6jL4G8Em_YWphG1wdKAF20BHUrf4B0yM,16
+gpt_batch-0.1.0.dist-info/RECORD,,

gpt_batch-0.1.0.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: bdist_wheel (0.41.2)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

gpt_batch-0.1.0.dist-info/top_level.txt

@@ -0,0 +1,2 @@
+gpt_batch
+tests

tests/test_batcher.py

@@ -0,0 +1,23 @@
+import pytest
+from gpt_batch import GPTBatcher
+import os
+
+def test_handle_message_list():
+    # Initialize the GPTBatcher with hypothetical valid credentials
+    #api_key = #get from system environment
+    api_key = os.getenv('TEST_KEY')
+    if not api_key:
+        raise ValueError("API key must be set in the environment variables")
+    batcher = GPTBatcher(api_key=api_key, model_name='gpt-3.5-turbo-1106', system_prompt="Your task is to discuss privacy questions and provide persuasive answers with supporting reasons.")
+    message_list = ["I think privacy is important", "I don't think privacy is important"]
+
+    # Call the method under test
+    results = batcher.handle_message_list(message_list)
+
+    # Assertions to verify the length of the results and the structure of each item
+    assert len(results) == 2, "There should be two results, one for each message"
+    assert all(len(result) >= 2 for result in results), "Each result should be at least two elements"
+
+# Optionally, you can add a test configuration if you have specific needs
+if __name__ == "__main__":
+    pytest.main()