feat: add AI topic to navigation tabs

README.md

@@ -83,6 +83,7 @@ S3_HOST_BUCKET=%(bucket)s.s3.amazonaws.com
 # Open AI (Aliyun)
 OPENAI_API_BASE=https://dashscope.aliyuncs.com/compatible-mode/v1
 OPENAI_API_KEY=********
+OPENAI_MODEL=qwen-turbo
 ```
 
 - Generate a [personal access token](https://github.com/settings/tokens) from GitHub and fill the `GITHUB_ACCESS_TOKEN`

frontend/src/layouts/MainLayout.vue

@@ -85,6 +85,7 @@ export default {
       drawer: false,
       topics: [
         "all",
+        "ai",
         "python",
         "java",
         "cpp",

utils.py

@@ -14,6 +14,7 @@
 from sqlalchemy import Column, String, Integer, DateTime, JSON
 from sqlalchemy.orm import declarative_base
 from openai import OpenAI
+from pydantic import BaseModel
 
 MAX_COMMENT_LENGTH = 512
 
@@ -22,6 +23,9 @@
     api_key=os.getenv("OPENAI_API_KEY"), base_url=os.getenv("OPENAI_API_BASE")
 )
 
+# OpenAI model for chat completions (default: qwen-turbo)
+OPENAI_MODEL = os.getenv("OPENAI_MODEL", "qwen-turbo")
+
 
 class LogFormatter(logging.Formatter):
 
@@ -175,7 +179,7 @@ def tldr(text: str) -> str:
         + f"The README of the repository is: \n\n{text}"
     )
     resp = openai_client.chat.completions.create(
-        model="qwen-turbo",
+        model=OPENAI_MODEL,
         messages=[
             {
                 "role": "user",
@@ -186,6 +190,50 @@ def tldr(text: str) -> str:
     return resp.choices[0].message.content
 
 
+class AIRelevance(BaseModel):
+    """Model for AI relevance detection."""
+
+    is_ai_related: bool
+
+
+def isai(text: str) -> bool:
+    """
+    Determine if a repository is related to AI based on its description.
+
+    Args:
+        text: The repository description or README content.
+
+    Returns:
+        True if the repository is AI-related, False otherwise.
+    """
+    prompt = (
+        "Determine if this GitHub repository is related to Artificial Intelligence (AI), "
+        "Large Language Models (LLMs), Vision Language Model (VLM), World Model, "
+        "Retrieval-Augmented Generation (RAG), Vector Database, Embedding, Agent,"
+        "Vibe Coding, Harness Engineering, or other AI fields. "
+        "Consider libraries, frameworks, models, and tools for AI development.\n\n"
+        f"Repository description/README:\n{text}"
+    )
+
+    resp = openai_client.beta.chat.completions.parse(
+        model=OPENAI_MODEL,
+        messages=[
+            {
+                "role": "user",
+                "content": prompt,
+            }
+        ],
+        response_format=AIRelevance,
+        extra_body={
+            "chat_template_kwargs": {
+                "enable_thinking": False,
+            }
+        },
+    )
+
+    return resp.choices[0].message.parsed.is_ai_related
+
+
 def get_repo_info(github_client: Github, full_name: str) -> Optional[Dict]:
     """
     Get GitHub repository information.
@@ -204,6 +252,14 @@ def get_repo_info(github_client: Github, full_name: str) -> Optional[Dict]:
     if description is None:
         description = tldr(repo.get_readme().decoded_content.decode("utf-8"))
         print("QWEN:", description)
+
+    # Check if repository is AI-related and add "ai" category
+    if isai(description):
+        if categories is None:
+            categories = ["ai"]
+        elif "ai" not in categories:
+            categories.append("ai")
+
     description_embedding = embedding(description)
     item = {
         "ItemId": full_name.replace("/", ":").lower(),

x.py

@@ -226,15 +226,16 @@ def upgrade_embedding():
     cursor = ""
     while True:
         items, cursor = gorse_client.get_items(1000, cursor)
-        if cursor == "":
-            break
         for item in tqdm(items):
             if len(item["Comment"]) > 0:
                 item["Labels"]["embedding"] = embedding(item["Comment"])
                 gorse_client.update_item(
                     item["ItemId"],
                     labels=item["Labels"],
                 )
+
+        if cursor == "":
+            break
 
 
 def write_dump(f, data: message.Message):
@@ -317,5 +318,46 @@ def dump_playground(database: str, username: Optional[str], password: Optional[s
             f"Dump complete: {num_users} users, {num_items} items, {num_feedback} feedback.")
 
 
+
+@command.command()
+def upgrade_ai():
+    """Upgrade items with AI category detection."""
+    cursor = ""
+    updated_count = 0
+    while True:
+        items, cursor = gorse_client.get_items(1000, cursor)
+        for item in tqdm(items):
+            item_id = item["ItemId"]
+            categories = item.get("Categories") or []
+
+            # Skip if already has "ai" category
+            if "ai" in categories:
+                continue
+
+            # Get description from comment
+            description = item.get("Comment", "")
+            if not description:
+                continue
+
+            # Check if AI-related
+            try:
+                if isai(description):
+                    categories.append("ai")
+                    gorse_client.update_item(
+                        item_id,
+                        categories=categories,
+                    )
+                    updated_count += 1
+                    print(f"UPDATE {item_id} -> ai")
+            except Exception as e:
+                print(f"FAIL {item_id}: {e}")
+                continue
+
+        if cursor == "":
+            break
+
+    print(f"Upgrade complete: {updated_count} items updated with 'ai' category.")
+
+
 if __name__ == "__main__":
     command()