feat: use the docs as input for the AI

Dockerfile

@@ -18,6 +18,9 @@ COPY LICENSE .
 
 ENV BOT_TOKEN=""
 ENV GOOGLE_API_KEY=""
+ENV DISCOURSE_API_KEY=""
+ENV DISCOURSE_BASE_URL=""
+ENV DISCOURSE_USERNAME=""
 
 # Where the database will be stored
 ENV BOT_DATA_DIR=""

aw.py

@@ -5,6 +5,7 @@ import discord
 from discord.ext import commands
 
 from database import initialize_db
+from bot.ai.handle_request import DiscourseSummarizer
 
 GUILD_ID = 1110531063161299074
 BOT_LOG = 1112049391482703873
@@ -19,6 +20,8 @@ load_dotenv(override=True)
 
 initialize_db()
 
+bot.ai_helper = DiscourseSummarizer()
+
 
 @bot.event
 async def on_ready():

bot/ai/handle_request.py

@@ -6,12 +6,125 @@ from google import genai
 API_KEY = os.getenv("GOOGLE_API_KEY")
 
 
-async def forward_to_google_api(prompt, image_object=None):
+class DiscourseSummarizer:
+    def __init__(self):
+        self.model = "gemini-2.0-flash"
+        self.display_name = "alterware"
+        self.cache = None
+        self.ttl = "21600s"
+        self.discourse_data = None
+
+        if not API_KEY:
+            print("Google API key is not set. Please contact the administrator.")
+            return
+
+        self.client = genai.Client(api_key=API_KEY)
+
+    def set_discourse_data(self, topic_data):
+        """
+        Sets the discourse data for the summarizer.
+
+        Args:
+            topic_data (str): The combined text of discourse posts.
+        """
+        self.discourse_data = topic_data
+
+    def summarize_discourse_topic(self, topic_data, system_instruction=None):
+        """
+        Creates a cache for the discourse topic data.
+
+        Args:
+            topic_data (str): The combined text of discourse posts.
+            system_instruction (str, optional): Custom system instruction for the model.
+        """
+        self.cache = self.client.caches.create(
+            model=self.model,
+            config=types.CreateCachedContentConfig(
+                display_name=self.display_name,
+                system_instruction=system_instruction
+                or (
+                    "You are a Discord chat bot named 'AlterWare' who helps users. You should limit your answers to be less than 2000 characters."
+                ),
+                contents=[topic_data],
+                ttl=self.ttl,
+            ),
+        )
+        print(f"Cached content created: {self.cache.name}")
+
+    def update_cache(self):
+        """
+        Updates the cache TTL.
+        """
+        if not self.cache:
+            raise RuntimeError(
+                "Cache has not been created. Run summarize_discourse_topic first."
+            )
+
+        self.client.caches.update(
+            name=self.cache.name, config=types.UpdateCachedContentConfig(ttl="21600s")
+        )
+        print("Cache updated.")
+
+    def ask(self, prompt):
+        """
+        Generates a response using the cached content.
+
+        Args:
+            prompt (str): The user prompt.
+
+        Returns:
+            str: The generated response.
+        """
+        if not self.cache:
+            raise RuntimeError(
+                "Cache has not been created. Run summarize_discourse_topic first."
+            )
+
+        response = self.client.models.generate_content(
+            model=self.model,
+            contents=prompt,
+            config=types.GenerateContentConfig(
+                max_output_tokens=400,
+                system_instruction="You are a Discord chat bot named 'AlterWare' who helps users. You should limit your answers to be less than 2000 characters.",
+                cached_content=self.cache.name,
+            ),
+        )
+        return response.text
+
+    def ask_without_cache(self, prompt):
+        """
+        Generates a response without using cached content, including discourse data.
+
+        Args:
+            prompt (str): The user prompt.
+
+        Returns:
+            str: The generated response.
+        """
+        if not self.discourse_data:
+            raise RuntimeError(
+                "Discourse data has not been set. Use set_discourse_data first."
+            )
+
+        prompt.insert(0, self.discourse_data)
+        response = self.client.models.generate_content(
+            model=self.model,
+            contents=prompt,
+            config=types.GenerateContentConfig(
+                max_output_tokens=400,
+                system_instruction="You are a Discord chat bot named 'AlterWare' who helps users. You should limit your answers to be less than 2000 characters.",
+            ),
+        )
+        return response.text
+
+
+async def forward_to_google_api(prompt, bot, image_object=None):
     """
     Forwards the message content and optional image object to a Google API.
 
     Args:
         prompt (discord.Message): The message object to forward.
+        bot (discord.Client): The Discord bot instance.
         image_object (tuple, optional): A tuple containing the image URL and its MIME type (e.g., ("url", "image/jpeg")).
     """
     if not API_KEY:
@@ -21,8 +134,6 @@ async def forward_to_google_api(prompt, image_object=None):
         )
         return
 
-    client = genai.Client(api_key=API_KEY)
-
     input = [prompt.content]
     if image_object:
         try:
@@ -34,16 +145,9 @@ async def forward_to_google_api(prompt, image_object=None):
             await prompt.reply(f"Failed to fetch the image", mention_author=True)
             return
 
-    response = client.models.generate_content(
+    response = bot.ai_helper.ask_without_cache(input)
-        model="gemini-2.0-flash",
-        contents=input,
-        config=types.GenerateContentConfig(
-            max_output_tokens=400,
-            system_instruction="You are a Discord chat bot named 'AlterWare' who helps users. You should limit your answers to be less than 2000 characters.",
-        ),
-    )
 
     await prompt.reply(
-        response.text,
+        response,
         mention_author=True,
     )

bot/discourse/__init__.py

@@ -0,0 +1 @@
+from .handle_request import fetch_cooked_posts, get_topics_by_tag, get_topics_by_id

bot/discourse/handle_request.py

@@ -0,0 +1,103 @@
+import requests
+import os
+
+from bs4 import BeautifulSoup
+
+DISCOURSE_BASE_URL = os.getenv("DISCOURSE_BASE_URL")
+API_KEY = os.getenv("DISCOURSE_API_KEY")
+API_USERNAME = os.getenv("DISCOURSE_API_USERNAME")
+
+headers = {"Api-Key": API_KEY, "Api-Username": API_USERNAME}
+
+
+def get_topics_by_id(topic_id):
+    """
+    Fetches a topic by its ID and returns the topic data.
+
+    Args:
+        topic_id (int): The ID of the topic to fetch.
+
+    Returns:
+        dict or None: The topic data if successful, otherwise None.
+    """
+    response = requests.get(f"{DISCOURSE_BASE_URL}/t/{topic_id}.json", headers=headers)
+    if response.status_code == 200:
+        return response.json()
+    else:
+        print(
+            f"Error fetching topic {topic_id}: {response.status_code} - {response.text}"
+        )
+        return None
+
+
+def get_topics_by_tag(tag_name):
+    """
+    Fetches all topics with a specific tag and retrieves the cooked string from each post.
+
+    Args:
+        tag_name (str): The name of the tag to filter topics.
+
+    Returns:
+        list: A list of cooked strings from all posts in the topics.
+    """
+    response = requests.get(
+        f"{DISCOURSE_BASE_URL}/tag/{tag_name}.json", headers=headers
+    )
+    if response.status_code == 200:
+        data = response.json()
+        topics = data.get("topic_list", {}).get("topics", [])
+        cooked_strings = []
+
+        for topic in topics:
+            topic_id = topic["id"]
+            topic_data = get_topics_by_id(topic_id)
+            if topic_data:
+                posts = topic_data.get("post_stream", {}).get("posts", [])
+                for post in posts:
+                    cooked_strings.append(post.get("cooked", ""))
+        return cooked_strings
+    else:
+        print(
+            f"Error fetching topics with tag '{tag_name}': {response.status_code} - {response.text}"
+        )
+        return []
+
+
+def fetch_cooked_posts(tag_name):
+    """
+    Fetches cooked strings from posts with a specific tag.
+
+    Args:
+        tag_name (str): The name of the tag to filter topics.
+
+    Returns:
+        list: A list of cooked strings from posts with the specified tag.
+    """
+    return get_topics_by_tag(tag_name)
+
+
+def html_to_text(html_content):
+    """
+    Cleans the provided HTML content and converts it to plain text.
+
+    Args:
+        html_content (str): The HTML content to clean.
+
+    Returns:
+        str: The cleaned plain text.
+    """
+    soup = BeautifulSoup(html_content, "html.parser")
+    return soup.get_text(separator="\n").strip()
+
+
+def combine_posts_text(posts):
+    """
+    Combines the cooked content of all posts into a single plain text block.
+
+    Args:
+        posts (list): A list of posts, each containing a "cooked" HTML string.
+
+    Returns:
+        str: The combined plain text of all posts.
+    """
+    return "\n\n".join([html_to_text(post["cooked"]) for post in posts])

bot/events_handlers/message_events.py

@@ -329,7 +329,7 @@ async def handle_message(message, bot):
                     image_object = (attachment.url, "image/png")
                     break
 
-            await forward_to_google_api(message, image_object)
+            await forward_to_google_api(message, bot, image_object)
             return
 
     # Too many mentions

bot/tasks.py

@@ -5,6 +5,7 @@ import discord
 from discord.ext import tasks, commands
 
 from bot.utils import aware_utcnow, fetch_api_data
+from bot.discourse.handle_request import fetch_cooked_posts, combine_posts_text
 
 from database import migrate_users_with_role
 
@@ -116,6 +117,36 @@ class SteamSaleChecker(commands.Cog):
         await self.bot.wait_until_ready()
 
 
+class DiscourseUpdater(commands.Cog):
+    def __init__(self, bot):
+        self.bot = bot
+        self.update_discourse_data.start()  # Start the task when the cog is loaded
+
+    def cog_unload(self):
+        self.update_discourse_data.cancel()  # Stop the task when the cog is unloaded
+
+    @tasks.loop(hours=6)
+    async def update_discourse_data(self):
+        """
+        Periodically fetches and updates Discourse data for the bot.
+        """
+        tag_name = "docs"
+        print("Fetching Discourse data...")
+        cooked_posts = fetch_cooked_posts(tag_name)
+        if cooked_posts:
+            combined_text = combine_posts_text(
+                [{"cooked": post} for post in cooked_posts]
+            )
+            self.bot.ai_helper.set_discourse_data(combined_text)
+            print("Discourse data updated successfully.")
+        else:
+            print(f"No posts found for tag '{tag_name}'.")
+
+    @update_discourse_data.before_loop
+    async def before_update_discourse_data(self):
+        await self.bot.wait_until_ready()
+
+
 async def setup(bot):
     @tasks.loop(minutes=10)
     async def update_status():
@@ -152,5 +183,6 @@ async def setup(bot):
     heat_death.start()
 
     await bot.add_cog(SteamSaleChecker(bot))
+    await bot.add_cog(DiscourseUpdater(bot))
 
     print("Tasks extension loaded!")

requirements.txt

@@ -4,3 +4,4 @@ audioop-lts
 python-dotenv
 pynacl
 google-genai
+beautifulsoup4