vector-vault 7.2.4__tar.gz → 7.2.6__tar.gz

{vector_vault-7.2.4 → vector_vault-7.2.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vector_vault
-Version: 7.2.4
+Version: 7.2.6
 Summary: Quickly create RAG apps, Agents, and Unleash the full power of AI with Vector Vault
 Home-page: https://github.com/John-Rood/VectorVault
 Author: VectorVault.io

{vector_vault-7.2.4 → vector_vault-7.2.6}/setup.py

@@ -2,7 +2,7 @@ from setuptools import setup, find_packages
 
 setup(
     name="vector_vault",
-    version="7.2.4",
+    version="7.2.6",
     packages=find_packages(),
     author="VectorVault.io",
     author_email="john@johnrood.com",

{vector_vault-7.2.4 → vector_vault-7.2.6}/vector_vault.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vector-vault
-Version: 7.2.4
+Version: 7.2.6
 Summary: Quickly create RAG apps, Agents, and Unleash the full power of AI with Vector Vault
 Home-page: https://github.com/John-Rood/VectorVault
 Author: VectorVault.io

{vector_vault-7.2.4 → vector_vault-7.2.6}/vectorvault/ai.py

@@ -395,20 +395,24 @@ class GrokPlatform(LLMPlatform):
 
         # Example model token limits for xAI Grok (adjust as needed)
         self.model_token_limits = {
+            "grok-code-fast": 256000,
             "grok-4": 256000,
-            "grok-3": 128000,
-            "grok-3-fast-latest": 128000,
-            "grok-3-mini": 128000,
-            "grok-3-mini-fast": 128000,
-            "grok-2-vision-latest": 32000,
-            "default": "grok-3-latest",
+            "grok-4-fast-reasoning": 2000000,
+            "grok-4-fast-non-reasoning": 2000000,
+            "grok-3": 131072,
+            "grok-3-mini": 131072,
+            "grok-2-vision-latest": 32768,
+            "default": "grok-4-fast-non-reasoning",
         }
         self.front_model_token_limits = {
-            "grok-4": 128000,
-            "grok-3": 128000,
-            "grok-3-mini": 128000,
-            "grok-2-vision-latest": 16000,
-            "default": "grok-3-latest",
+            "grok-code-fast": 256000,
+            "grok-4": 256000,
+            "grok-4-fast-reasoning": 2000000,
+            "grok-4-fast-non-reasoning": 2000000,
+            "grok-3": 131072,
+            "grok-3-mini": 131072,
+            "grok-2-vision-latest": 32768,
+            "default": "grok-4-fast-non-reasoning",
         }
 
         self.default_model = self.model_token_limits["default"]
@@ -520,7 +524,7 @@ class GrokPlatform(LLMPlatform):
             }
         ]
         """
-        model = model if model else self.default_model
+        model = model if model else "grok-2-vision-latest"
         if not image_path and not image_url:
             raise ValueError("Must specify either image_path or image_url.")
 

{vector_vault-7.2.4 → vector_vault-7.2.6}/vectorvault/vault.py

@@ -852,7 +852,7 @@ class Vault:
     def load_vectors(self, vault = None):
         start_time = time.time()
         vault = vault if vault else self.vault 
-        t = T(target=self.load_mapping())
+        t = T(target=self.load_mapping, args=(vault,))
         t.start()
         temp_file_path = self.cloud_manager.download_to_temp_file(name_vecs(vault, self.user, self.api))
         self._vectors = get_vectors(self.dims)
@@ -1003,7 +1003,7 @@ class Vault:
             Internal function that returns vector similar items. Requires input vector, returns similar items
         '''
         try:
-            self.load_vectors()
+            self.load_vectors(vault)
             start_time = time.time()
             if not include_distances:
                 vecs = self.vectors.get_nns_by_vector(vector, n)
@@ -1181,6 +1181,38 @@ class Vault:
         return self.get_items_by_vector(vector, n, include_distances = include_distances, vault = vault if vault else self.vault)
 
 
+    def get_similar_from_vaults(self, text: str, n: int = 4, vaults: List[str] = None):
+        '''
+            Searches for the n most similar items in each of the provided vaults, combines the results,
+            sorts by distance, and returns the closest n items overall (with distances).
+
+            Args:
+                text: The query text to embed and search with
+                n: The number of nearest items to retrieve per vault, and the number of items to return overall
+                vaults: A non-empty list of vault names to search
+
+            Returns:
+                A list of up to n items, each including a 'distance' field
+        '''
+        if not vaults or not isinstance(vaults, list):
+            raise ValueError("A non-empty list of vaults must be provided.")
+
+        T(target=self.cloud_manager.build_update, args=(n,)).start()
+        vector = self.process_batch([text], never_stop=False, loop_timeout=180)[0]
+
+        combined_results = []
+        for v in vaults:
+            try:
+                results = self.get_items_by_vector(vector, n, include_distances=True, vault=v)
+                combined_results.extend(results)
+            except Exception:
+                # Skip vaults that cannot be loaded or queried
+                continue
+
+        combined_sorted = sorted(combined_results, key=lambda x: x.get('distance', float('inf')))
+        return combined_sorted[:n]
+
+
     def add_item(self, text: str, meta: dict = None, name: str = None):
         """
             If your text length is greater than 15000 characters, you should use Vault.split_text(your_text) to 
@@ -1317,6 +1349,7 @@ class Vault:
             timeout: int = 300,
             image_path: str = None,
             image_url: str = None,
+            vaults: List[str] = None,
             ):
         '''
             Chat get response from OpenAI's ChatGPT. 
@@ -1404,7 +1437,10 @@ class Vault:
                             else:
                                 search_input = segment + history if history_search else segment
                                 
-                            context = self.get_similar(search_input, n=n_context)
+                            if vaults and isinstance(vaults, list) and all(isinstance(v, str) for v in vaults):
+                                context = self.get_similar_from_vaults(search_input, n=n_context, vaults=vaults)
+                            else:
+                                context = self.get_similar(search_input, n=n_context)
                             input_ = str(context) if include_context_meta else ''
                             for text in context:
                                 input_ += text['data']
@@ -1454,6 +1490,7 @@ class Vault:
             timeout: int = 300,
             image_path: str = None,
             image_url: str = None,
+            vaults: List[str] = None,
             ):
         '''
             Always use this get_chat_stream() wrapped by either print_stream(), or cloud_stream()
@@ -1547,7 +1584,10 @@ class Vault:
                             else:
                                 search_input = segment + history if history_search else segment
                             
-                            context = self.get_similar(search_input, n=n_context)
+                            if vaults and isinstance(vaults, list) and all(isinstance(v, str) for v in vaults):
+                                context = self.get_similar_from_vaults(search_input, n=n_context, vaults=vaults)
+                            else:
+                                context = self.get_similar(search_input, n=n_context)
                             input_ = str(context) if include_context_meta else ''
                             for text in context:
                                 input_ += text['data']