refactor: use client.chat.completions everywhere

somehow, client.completions seems very fragile

app/core/podcast/eleventify.py

@@ -49,25 +49,26 @@ def create_episode_summary(content: str, episode: str) -> str:
 
     prompt = f"Given the transcript of today's episode below, write a very brief summary to use as a description for the media file. Your summary should be a single paragraph, not exceeding 2 sentences.\n\n```\n{content}\n```"
 
-    # model = "lmsys/vicuna-13b-v1.5-16k"
-    # model = "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO"
-    # model = "openchat/openchat-3.5-1210"
-    # model = "mistralai/Mixtral-8x7B-Instruct-v0.1"
-    model = "mistralai/Mixtral-8x7B-v0.1"
+    model = "meta-llama/Meta-Llama-3-70B-Instruct-Turbo"
     temperature = 0.75
     max_tokens = 512
 
-    response = client.completions.create(
-        prompt=prompt,
+    completion = client.chat.completions.create(
         model=model,
+        messages=[
+            {
+                "role": "user",
+                "content": prompt,
+            },
+        ],
         temperature=temperature,
         max_tokens=max_tokens,
     )
-    logger.info(response)
+    logger.info(completion)
 
     fallback = f"This is episode {episode} of the Zed News Podcast."
 
-    if result := response.choices[0].text.strip():
+    if result := completion.choices[0].message.content.strip():
         result = result.replace("```", "")  # Remove triple backticks
         first_line = result.splitlines()[0].lower()
         unwanted = ["summary:", "here's", "here is", "sure"]

app/core/social/post.py

@@ -13,17 +13,17 @@
 import pathlib
 import random
 import sys
-import numpy as np
 from http import HTTPStatus
 
 import facebook
+import numpy as np
 import PIL
 import requests
 from dotenv import load_dotenv
 from moviepy.editor import CompositeVideoClip, ImageClip, TextClip, VideoFileClip
-from together import Together
 from moviepy.video.compositing.concatenate import concatenate_videoclips
 from moviepy.video.tools.segmenting import findObjects
+from together import Together
 
 from app.core.utilities import (
     ASSETS_DIR,  # noqa: F401
@@ -258,24 +258,32 @@ def create_facebook_post(content: str, url: str) -> str:
     https://docs.together.ai/reference/complete
     """
 
-    prompt = f"You are a social media marketing guru. Your task is to immediately produce a short Facebook teaser post of today's podcast episode whose details are below. Use bullet points, emojis and hashtags as appropriate. Don't cover every news item, just the most interesting ones. Your post will accompany a video, whose URL is {url}. Please note that this video is just basically the audio with some looping animated background, therefore, your post shouldn't ask people to 'watch' the video per se, but rather to, 'check it out' or 'tune in to'. Do not use markdown.\n\n```\n{content}\n```"
+    system_prompt = f"You are a social media marketing expert. Your task is to produce a short Facebook teaser post for today's podcast episode based on the provided podcast episode details. The post should highlight only the most interesting news items, using bullet points, emojis, and hashtags where appropriate. Keep in mind that your post will accompany a video, whose URL is {url}, and which is primarily an audio recording with a looping animated background. Therefore, avoid asking people to 'watch' the video. Instead, encourage them to 'check it out' or 'tune in.'"
+
+    user_prompt = f"Produce a Facebook teaser post based on the following podcast episode details.\n\n\n{content}\n"
 
-    # model = "lmsys/vicuna-13b-v1.5-16k"
-    # model = "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO"
-    # model = "openchat/openchat-3.5-1210"
-    model = "mistralai/Mixtral-8x7B-v0.1"
+    model = "meta-llama/Meta-Llama-3-70B-Instruct-Turbo"
     temperature = 0.75
     max_tokens = 1024
 
-    response = client.completions.create(
-        prompt=prompt,
+    completion = client.chat.completions.create(
         model=model,
+        messages=[
+            {
+                "role": "system",
+                "content": system_prompt,
+            },
+            {
+                "role": "user",
+                "content": user_prompt,
+            },
+        ],
         temperature=temperature,
         max_tokens=max_tokens,
     )
-    logger.info(response)
+    logger.info(completion)
 
-    if result := response.choices[0].text.strip():
+    if result := completion.choices[0].message.content.strip():
         result = result.replace("```", "")  # Remove triple backticks
         first_line = result.splitlines()[0].lower()
         unwanted = ["facebook post:", "post:", "here's your", "here is"]
@@ -305,21 +313,32 @@ def create_episode_summary(content: str) -> str:
     host = f"Host: {podcast_host}"
     separator = "------------------------------------"
 
-    prompt = f"You are a social media marketing guru. Your task is to write a very brief summary to use as a description for a facebook video post, given the transcript of today's episode below. Use bullet points, emojis and hashtags as appropriate. Do not use markdown. Please note that this video is just basically the audio with some looping animated background, therefore, your post shouldn't ask people to 'watch' the video per se, but rather to, 'check it out' or 'tune in to'. At the end, mention that more details can be obtained from {podcast_url}.\n\n```{separator}\n\n{title}\n{date}\n{host}\n\n{separator}\n\n{content}\n```"
+    system_prompt = f"You are a social media marketing guru. Your task is to write a very brief summary to use as a description for a Facebook video post, given the transcript of today's episode below. Use bullet points, emojis, and hashtags as appropriate. Do not use markdown. Please note that this video is just basically the audio with some looping animated background, so your post shouldn't ask people to 'watch' the video, but rather to 'check it out' or 'tune in to'. At the end, mention that more details can be obtained from {podcast_url}."
+
+    user_prompt = f"{separator}\n\n{title}\n{date}\n{host}\n\n{separator}\n\n{content}\n"
 
-    model = "mistralai/Mixtral-8x7B-v0.1"
+    model = "meta-llama/Meta-Llama-3-70B-Instruct-Turbo"
     temperature = 0.75
     max_tokens = 1024
 
-    response = client.completions.create(
-        prompt=prompt,
+    completion = client.chat.completions.create(
         model=model,
+        messages=[
+            {
+                "role": "system",
+                "content": system_prompt,
+            },
+            {
+                "role": "user",
+                "content": user_prompt,
+            },
+        ],
         temperature=temperature,
         max_tokens=max_tokens,
     )
-    logger.info(response)
+    logger.info(completion)
 
-    if result := response.choices[0].text.strip():
+    if result := completion.choices[0].message.content.strip():
         result = result.replace("```", "")  # Remove triple backticks
         first_line = result.splitlines()[0].lower()
         unwanted = ["summary:", "here's", "here is", "sure"]

app/core/summarization/backends/together.py

@@ -1,6 +1,6 @@
 import logging
-import time
 import sys
+import time
 
 from together import Together, error
 
@@ -16,8 +16,11 @@ def summarize(content: str, title: str) -> str:
     https://docs.together.ai/reference/complete
     """
 
-    prompt = f"You are a distinguished news editor and content publisher, your task is to summarize the following news entry. The summary should accurately reflect the main message and arguments presented in the original news entry, while also being concise and easy to understand. Your summary should not exceed two sentences.\n\n ```{content}```:"
-    model = "mistralai/Mixtral-8x7B-v0.1"
+    system_prompt = "You are a distinguished news editor and content publisher. Your task is to summarize the provided news entry. The summary should accurately reflect the main message and arguments presented in the original news entry, while also being concise and easy to understand."
+
+    user_prompt = f"Summarize the following news entry in not more than two sentences.\n\n ```{content}```"
+
+    model = "meta-llama/Meta-Llama-3-70B-Instruct-Turbo"
     temperature = 0.7
     max_tokens = 192
 
@@ -26,16 +29,25 @@ def summarize(content: str, title: str) -> str:
 
     while retries < max_retries:
         try:
-            response = client.completions.create(
-                prompt=prompt,
+            completion = client.chat.completions.create(
                 model=model,
+                messages=[
+                    {
+                        "role": "system",
+                        "content": system_prompt,
+                    },
+                    {
+                        "role": "user",
+                        "content": user_prompt,
+                    },
+                ],
                 temperature=temperature,
                 max_tokens=max_tokens,
             )
             time.sleep(1.5)
-            logging.info(response)
+            logging.info(completion)
 
-            if result := response.choices[0].text.strip():
+            if result := completion.choices[0].message.content.strip():
                 result = result.replace("```", "")  # Remove triple backticks
                 first_line = result.splitlines()[0].lower()
                 unwanted = ["summary:", "here's", "here is", "sure"]
@@ -62,8 +74,11 @@ def brief_summary(content: str, title: str) -> str:
     https://docs.together.ai/reference/complete
     """
 
-    prompt = f"You are a distinguished news editor and content publisher, your task is to summarize the following news entry in one sentence.\n\n ```{content}```:"
-    model = "mistralai/Mixtral-8x7B-v0.1"
+    system_prompt = "You are a distinguished news editor and content publisher. Your task is to summarize the provided news entry. The summary should accurately reflect the main message and arguments presented in the original news entry, while also being concise and easy to understand."
+
+    user_prompt = f"Summarize the following news entry in one sentence.\n\n ```{content}```"
+
+    model = "meta-llama/Meta-Llama-3-70B-Instruct-Turbo"
     temperature = 0.7
     max_tokens = 96
 
@@ -72,16 +87,34 @@ def brief_summary(content: str, title: str) -> str:
 
     while retries < max_retries:
         try:
-            response = client.completions.create(
-                prompt=prompt,
+            completion = client.chat.completions.create(
                 model=model,
+                messages=[
+                    {
+                        "role": "system",
+                        "content": system_prompt,
+                    },
+                    {
+                        "role": "user",
+                        "content": user_prompt,
+                    },
+                ],
                 temperature=temperature,
                 max_tokens=max_tokens,
             )
             time.sleep(1.5)
-            logging.info(response)
+            logging.info(completion)
 
-            return response.choices[0].text
+            if result := completion.choices[0].message.content.strip():
+                result = result.replace("```", "")  # Remove triple backticks
+                first_line = result.splitlines()[0].lower()
+                unwanted = ["summary:", "here's", "here is", "sure"]
+
+                if any(string in first_line for string in unwanted):
+                    # Remove the first line from result
+                    result = "\n".join(result.split("\n")[1:])
+
+                return result.replace("\n", "")  # Remove newlines
         except error.ServiceUnavailableError:
             retries += 1
             logging.error(f"Service unavailable. Retrying {retries}/{max_retries} in 10 seconds...")