feat: add framework-agnostic weaviate tool

agentstack/tools/weaviate/__init__.py

@@ -0,0 +1,108 @@
+import os
+import json
+import weaviate
+from typing import Optional
+from weaviate.classes.config import Configure
+from weaviate.classes.init import Auth
+
+# Required environment variables
+url = os.getenv("WEAVIATE_URL")
+api_key = os.getenv("WEAVIATE_API_KEY")
+openai_key = os.getenv("WEAVIATE_OPENAI_API_KEY") or os.getenv("OPENAI_API_KEY")
+
+if not url:
+    raise Exception((
+        "Weaviate URL has not been provided.\n"
+        "Did you set the WEAVIATE_URL in your project's .env file?"
+    ))
+
+if not api_key:
+    raise Exception((
+        "Weaviate API key has not been provided.\n"
+        "Did you set the WEAVIATE_API_KEY in your project's .env file?"
+    ))
+
+if not openai_key:
+    raise Exception((
+        "OpenAI API key has not been provided.\n"
+        "Did you set either WEAVIATE_OPENAI_API_KEY or OPENAI_API_KEY in your project's .env file?"
+    ))
+
+def search_collection(
+    collection_name: str,
+    query: str,
+    limit: int = 3,
+    model: str = "nomic-embed-text"
+) -> str:
+    """Search a Weaviate collection using near-text queries.
+
+    Args:
+        collection_name: Name of the collection to search
+        query: The search query
+        limit: Maximum number of results (default: 3)
+        model: Text embedding model to use (default: nomic-embed-text)
+
+    Returns:
+        str: JSON string containing search results
+    """
+    headers = {"X-OpenAI-Api-Key": openai_key}
+    vectorizer = Configure.Vectorizer.text2vec_openai(model=model)
+
+    client = weaviate.connect_to_weaviate_cloud(
+        cluster_url=url,
+        auth_credentials=Auth.api_key(api_key),
+        headers=headers
+    )
+
+    try:
+        collection = client.collections.get(collection_name)
+        if not collection:
+            raise ValueError(f"Collection {collection_name} not found")
+
+        response = collection.query.near_text(
+            query=query,
+            limit=limit
+        )
+
+        results = []
+        for obj in response.objects:
+            results.append(obj.properties)
+
+        return json.dumps(results, indent=2)
+    finally:
+        client.close()
+
+def create_collection(
+    collection_name: str,
+    model: str = "nomic-embed-text"
+) -> str:
+    """Create a new Weaviate collection.
+
+    Args:
+        collection_name: Name of the collection to create
+        model: Text embedding model to use (default: nomic-embed-text)
+
+    Returns:
+        str: Success message
+    """
+    headers = {"X-OpenAI-Api-Key": openai_key}
+    vectorizer = Configure.Vectorizer.text2vec_openai(model=model)
+
+    client = weaviate.connect_to_weaviate_cloud(
+        cluster_url=url,
+        auth_credentials=Auth.api_key(api_key),
+        headers=headers
+    )
+
+    try:
+        collection = client.collections.get(collection_name)
+        if collection:
+            return f"Collection {collection_name} already exists"
+
+        client.collections.create(
+            name=collection_name,
+            vectorizer_config=vectorizer
+        )
+        return f"Created collection {collection_name}"
+    finally:
+        client.close()

agentstack/tools/weaviate/config.json

@@ -0,0 +1,19 @@
+{
+  "name": "weaviate",
+  "url": "https://github.com/weaviate/weaviate-python-client",
+  "category": "vector-store",
+  "env": {
+    "WEAVIATE_URL": null,
+    "WEAVIATE_API_KEY": null,
+    "WEAVIATE_OPENAI_API_KEY": null
+  },
+  "dependencies": [
+    "weaviate-client>=3.0.0",
+    "openai>=1.0.0"
+  ],
+  "tools": [
+    "search_collection",
+    "create_collection"
+  ],
+  "cta": "🔗 Create your Weaviate cluster here: https://console.weaviate.cloud/"
+}