commit e4781586464f38f7b3ef1a663b2a50cca6cd2b56
parent bb8c6a7ed9b5b5c450c61408f432980be8e6b120
Author: vin <git@vineetk.net>
Date: Sat, 19 Jul 2025 01:51:36 -0400
respect ratelimit and change output extension to .gpt
Diffstat:
| M | main.py | | | 78 | ++++++++++++++++++++++++++++++++++++++---------------------------------------- |
1 file changed, 38 insertions(+), 40 deletions(-)
diff --git a/main.py b/main.py
@@ -1,10 +1,16 @@
import os
import re
import sys
-from openai import OpenAI
+import backoff
+import openai
from datetime import datetime, timedelta
from tqdm import tqdm
+client = openai.OpenAI(
+ #base_url="http://localhost:8080/v1",
+ base_url="https://api.openai.com/v1",
+)
+
SYSTEM_PROMPT_TEMPLATE = """
You are an expert podcast content analyzer. Your task is to identify and extract pre-recorded dynamic advertising segments from podcast transcripts. These ads typically have a distinct tone shift, often using more direct, persuasive language, and frequently include calls to action, product mentions, or specific brand names.
@@ -202,28 +208,29 @@ def chunk_transcript(
return chunks
-def call_openai_api(client: OpenAI, chunk_text: str, system_prompt: str) -> str:
+@backoff.on_exception(backoff.expo, openai.RateLimitError)
+def completions_with_backoff(**kwargs):
+ return client.chat.completions.create(**kwargs)
+
+def call_openai_api(chunk_text: str, system_prompt: str) -> str:
"""
Makes an API call to OpenAI with the given system prompt and transcript chunk.
"""
- try:
- response = client.chat.completions.create(
- #model="Gemma-3-4B_32K",
- #model="google/gemini-2.0-flash-exp:free",
- model="gpt-4.1",
- messages=[
- {"role": "system", "content": system_prompt},
- {"role": "user", "content": f"Input Transcript:\n{chunk_text}"}
- ],
- temperature=0.0,
- )
- #print(response)
- return response.choices[0].message.content.strip()
- except Exception as e:
- print(f"Error calling OpenAI API: {e}")
- return ""
-
-def process_file(input_file: str, output_file: str, client: OpenAI):
+ response = completions_with_backoff(
+ #model="Gemma-3-4B_32K",
+ #model="google/gemini-2.0-flash-exp:free",
+ model="gpt-4.1",
+ messages=[
+ {"role": "system", "content": system_prompt},
+ {"role": "user", "content": f"Input Transcript:\n{chunk_text}"}
+ ],
+ temperature=0.0,
+ max_tokens=1024, # realistically it'd be under like 512
+ )
+ #print(response)
+ return response.choices[0].message.content.strip()
+
+def process_file(input_file: str, output_file: str):
with open(input_file, "r", encoding="utf-8") as f:
full_transcript_content = f.read()
@@ -247,7 +254,11 @@ def process_file(input_file: str, output_file: str, client: OpenAI):
# Add the "Your Turn" header before the chunk content
user_prompt_content = f"**Your Turn - Input Transcript:**\n{chunk_text_for_api}"
- ad_timestamps_str = call_openai_api(client, user_prompt_content, SYSTEM_PROMPT_TEMPLATE)
+ try:
+ ad_timestamps_str = call_openai_api(user_prompt_content, SYSTEM_PROMPT_TEMPLATE)
+ except Exception as e:
+ print(f"Error calling OpenAI API: {e}")
+ continue
if ad_timestamps_str:
# Split by newlines in case the model returns multiple segments
@@ -258,29 +269,16 @@ def process_file(input_file: str, output_file: str, client: OpenAI):
f.write("\n".join(all_ad_timestamps))
def main():
- # --- Configuration ---
- api_key = os.getenv("OPENAI_KEY")
- if not api_key:
- print("OPENAI_KEY empty")
- sys.exit(1)
-
- client = OpenAI(
- #api_key="none",
- api_key=api_key,
- #base_url="http://localhost:8080/v1",
- base_url="https://api.openai.com/v1",
- )
-
paths = sys.argv[1:]
- for i, p in enumerate(paths):
- out = "".join(p.split(".")[:-1]) + "_gpt.txt"
- if os.path.isfile(out)
- paths.remove(i)
+ for p in paths:
+ out = "".join(p.split(".")[:-1]) + ".gpt"
+ if os.path.isfile(out):
+ paths.remove(p)
print(paths)
for p in tqdm(paths):
- out = "".join(p.split(".")[:-1]) + "_gpt.txt"
- process_file(p, out, client)
+ out = "".join(p.split(".")[:-1]) + ".gpt"
+ process_file(p, out)
if __name__ == "__main__":
main()
