1.9.2

.github/pull_request_template.md

@@ -1,18 +1,10 @@
 # Description
 
-Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context. List any dependencies that are required for this change.
-
-Related to issue #(issue)
-
-## Type of change
-
-- [ ] Bug fix (non-breaking change which fixes an issue)
-- [ ] New feature (non-breaking change which adds functionality)
-- [ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)
-- [ ] This change requires a documentation update
+Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context.  
 
 # Checklist:
 
-- [ ] My code follows the style guidelines of this project
+- [ ] I submitted my PR to branch `develop`
 - [ ] I have performed a self-review of my own code
 - [ ] I have commented my code, particularly in hard-to-understand areas
+- [ ] I have run or added tests to cover my contribution

core/cat/env.py

@@ -24,6 +24,7 @@ def get_supported_env_variables():
         "CCAT_CORS_ENABLED": "true",
         "CCAT_CACHE_TYPE": "in_memory",
         "CCAT_CACHE_DIR": "/tmp",
+        "CCAT_QDRANT_CLIENT_TIMEOUT": None,
     }
 
 

core/cat/looking_glass/cheshire_cat.py

@@ -7,7 +7,7 @@
 from langchain_core.runnables import RunnableLambda
 from langchain_core.prompts import ChatPromptTemplate
 from langchain_core.output_parsers.string import StrOutputParser
-from langchain_community.llms import Cohere
+from langchain_cohere import ChatCohere
 from langchain_openai import ChatOpenAI, OpenAI
 from langchain_google_genai import ChatGoogleGenerativeAI
 
@@ -208,7 +208,7 @@ def load_language_embedder(self) -> embedders.EmbedderSettings:
         # For Azure avoid automatic embedder selection
 
         # Cohere
-        elif type(self._llm) in [Cohere]:
+        elif type(self._llm) in [ChatCohere]:
             embedder = embedders.EmbedderCohereConfig.get_embedder_from_config(
                 {
                     "cohere_api_key": self._llm.cohere_api_key,

core/cat/looking_glass/stray_cat.py

@@ -625,7 +625,7 @@ def classify(
 Allowed classes are:
 {labels_list}{examples_list}
 
-"{sentence}" -> """
+Just output the class, nothing else."""
 
         response = self.llm(prompt)
 

core/cat/memory/vector_memory.py

@@ -65,6 +65,9 @@ def connect_to_vector_memory(self) -> None:
             qdrant_https = is_https(qdrant_host)
             qdrant_host = extract_domain_from_url(qdrant_host)
             qdrant_api_key = get_env("CCAT_QDRANT_API_KEY")
+
+            qdrant_client_timeout = get_env("CCAT_QDRANT_CLIENT_TIMEOUT")
+            qdrant_client_timeout = int(qdrant_client_timeout) if qdrant_client_timeout is not None else None
 
             try:
                 s = socket.socket()
@@ -81,6 +84,7 @@ def connect_to_vector_memory(self) -> None:
                 port=qdrant_port,
                 https=qdrant_https,
                 api_key=qdrant_api_key,
+                timeout=qdrant_client_timeout
             )
 
     def delete_collection(self, collection_name: str):

core/cat/memory/vector_memory_collection.py

@@ -59,11 +59,11 @@ def check_embedding_size(self):
             == self.embedder_size
         )
         alias = self.embedder_name + "_" + self.collection_name
-        if (
-            alias
-            == self.client.get_collection_aliases(self.collection_name)
-            .aliases[0]
-            .alias_name
+
+        existing_aliases = self.client.get_collection_aliases(self.collection_name).aliases
+
+        if ( len(existing_aliases) > 0 and
+            alias == existing_aliases[0].alias_name
             and same_size
         ):
             log.debug(f'Collection "{self.collection_name}" has the same embedder')
@@ -94,31 +94,48 @@ def create_db_collection_if_not_exists(self):
 
     # create collection
     def create_collection(self):
-        log.warning(f'Creating collection "{self.collection_name}" ...')
-        self.client.create_collection(
-            collection_name=self.collection_name,
-            vectors_config=VectorParams(
-                size=self.embedder_size, distance=Distance.COSINE
-            ),
-            # hybrid mode: original vector on Disk, quantized vector in RAM
-            optimizers_config=OptimizersConfigDiff(memmap_threshold=20000),
-            quantization_config=ScalarQuantization(
-                scalar=ScalarQuantizationConfig(
-                    type=ScalarType.INT8, quantile=0.95, always_ram=True
-                )
-            ),
-        )
-
-        self.client.update_collection_aliases(
-            change_aliases_operations=[
-                CreateAliasOperation(
-                    create_alias=CreateAlias(
-                        collection_name=self.collection_name,
-                        alias_name=self.embedder_name + "_" + self.collection_name,
+        try:
+            log.warning(f'Creating collection "{self.collection_name}" ...')
+            self.client.create_collection(
+                collection_name=self.collection_name,
+                vectors_config=VectorParams(
+                    size=self.embedder_size, distance=Distance.COSINE
+                ),
+                # hybrid mode: original vector on Disk, quantized vector in RAM
+                optimizers_config=OptimizersConfigDiff(memmap_threshold=20000),
+                quantization_config=ScalarQuantization(
+                    scalar=ScalarQuantizationConfig(
+                        type=ScalarType.INT8, quantile=0.95, always_ram=True
                     )
-                )
-            ]
-        )
+                ),
+            )
+        except Exception as e:
+            log.error(f"Error creating collection {self.collection_name}. Try setting a higher timeout value in CCAT_QDRANT_CLIENT_TIMEOUT: {e}")
+            self.client.delete_collection(self.collection_name)
+            raise
+
+        try:
+            alias_name=self.embedder_name + "_" + self.collection_name
+            log.warning(f'Creating alias {alias_name} for collection "{self.collection_name}" ...')
+
+            self.client.update_collection_aliases(
+                change_aliases_operations=[
+                    CreateAliasOperation(
+                        create_alias=CreateAlias(
+                            collection_name=self.collection_name,
+                            alias_name=alias_name,
+                        )
+                    )
+                ]
+            )
+
+            log.warning(f'Created alias {alias_name} for collection "{self.collection_name}" ...')
+        except Exception as e:
+            log.error(f"Error creating collection alias {alias_name} for collection {self.collection_name}: {e}")
+            self.client.delete_collection(self.collection_name)
+            log.error(f" collection {self.collection_name} deleted")
+            raise
+
 
     # adapted from https://github.com/langchain-ai/langchain/blob/bfc12a4a7644cfc4d832cc4023086a7a5374f46a/libs/langchain/langchain/vectorstores/qdrant.py#L1965
     def _qdrant_filter_from_dict(self, filter: dict) -> Filter:

core/cat/routes/plugins.py

@@ -28,7 +28,18 @@ async def get_available_plugins(
     # index registry plugins by url
     registry_plugins_index = {}
     for p in registry_plugins:
-        plugin_url = p["url"]
+        plugin_url = p.get("plugin_url", None)
+        if plugin_url is None:
+            log.warning(f"Plugin {p.get('name')} has no `plugin_url`. It will be skipped from the registry list.")
+            continue
+        # url = p.get("url", None)
+        # if url and url != plugin_url:
+        #     log.info(f"Plugin {p.get('name')} has `url` {url} different from `plugin_url` {plugin_url}. please check the plugin.")
+        if plugin_url in registry_plugins_index:
+            current = registry_plugins_index[plugin_url]
+            log.warning(f"duplicate plugin_url {plugin_url} found in registry. Plugins {p.get('name')} has same url than {current.get('name')}. Skipping.")
+            continue    
+
         registry_plugins_index[plugin_url] = p
 
     # get active plugins
@@ -53,19 +64,20 @@ async def get_available_plugins(
         manifest["endpoints"] = [{"name": endpoint.name, "tags": endpoint.tags} for endpoint in p.endpoints]
         manifest["forms"] = [{"name": form.name} for form in p.forms]
 
+        # do not show already installed plugins among registry plugins
+        r = registry_plugins_index.pop(manifest["plugin_url"], None)
+
         # filter by query
         plugin_text = [str(field) for field in manifest.values()]
         plugin_text = " ".join(plugin_text).lower()
+
         if (query is None) or (query.lower() in plugin_text):
-            for r in registry_plugins:
-                if r["plugin_url"] == p.manifest["plugin_url"]:
-                    if r["version"] != p.manifest["version"]:
-                        manifest["upgrade"] = r["version"]
+            if r is not None:
+                r_version = r.get("version", None)
+                if r_version is not None and r_version != p.manifest.get("version"):
+                    manifest["upgrade"] = r["version"]
             installed_plugins.append(manifest)
 
-        # do not show already installed plugins among registry plugins
-        registry_plugins_index.pop(manifest["plugin_url"], None)
-
     return {
         "filters": {
             "query": query,
@@ -298,4 +310,4 @@ async def delete_plugin(
     # remove folder, hooks and tools
     ccat.mad_hatter.uninstall_plugin(plugin_id)
 
-    return {"deleted": plugin_id}
+    return {"deleted": plugin_id}

core/cat/routes/websocket/websocket.py

@@ -43,9 +43,9 @@ async def websocket_endpoint(
     except WebSocketDisconnect:
         log.info(f"WebSocket connection closed for user {cat.user_id}")
     finally:
-        
+
         # cat's working memory in this scope has not been updated
-        #cat.load_working_memory_from_cache()
-        
+        cat.load_working_memory_from_cache()
+
         # Remove connection on disconnect
         websocket_manager.remove_connection(cat.user_id)

core/cat/utils.py

@@ -7,7 +7,7 @@
 from typing import Dict, Tuple
 from pydantic import BaseModel, ConfigDict
 
-from langchain.evaluation import StringDistance, load_evaluator, EvaluatorType
+from rapidfuzz.distance import Levenshtein
 from langchain_core.output_parsers import JsonOutputParser
 from langchain_core.prompts import PromptTemplate
 from langchain_core.utils import get_colored_text
@@ -145,7 +145,7 @@ def explicit_error_message(e):
 def deprecation_warning(message: str, skip=3):
     """Log a deprecation warning with caller's information.
         "skip" is the number of stack levels to go back to the caller info."""
-    
+
     caller = get_caller_info(skip, return_short=False)
 
     # Format and log the warning message
@@ -155,24 +155,17 @@ def deprecation_warning(message: str, skip=3):
 
 
 def levenshtein_distance(prediction: str, reference: str) -> int:
-    jaro_evaluator = load_evaluator(
-        EvaluatorType.STRING_DISTANCE, distance=StringDistance.LEVENSHTEIN
-    )
-    result = jaro_evaluator.evaluate_strings(
-        prediction=prediction,
-        reference=reference,
-    )
-    return result["score"]
-
+    res = Levenshtein.normalized_distance(prediction, reference)
+    return res
 
 def parse_json(json_string: str, pydantic_model: BaseModel = None) -> dict:
     # instantiate parser
     parser = JsonOutputParser(pydantic_object=pydantic_model)
 
     # clean to help small LLMs
     replaces = {
-        "\_": "_",
-        "\-": "-",
+        "\\_": "_",
+        "\\-": "-",
         "None": "null",
         "{{": "{",
         "}}": "}",
@@ -185,7 +178,7 @@ def parse_json(json_string: str, pydantic_model: BaseModel = None) -> dict:
 
     # parse
     parsed = parser.parse(json_string[start_index:])
-    
+
     if pydantic_model:
         return pydantic_model(**parsed)
     return parsed
@@ -213,7 +206,7 @@ def match_prompt_variables(
             prompt_template = \
                 prompt_template.replace("{" + m + "}", "")
             log.debug(f"Placeholder '{m}' not found in prompt variables, removed")
-            
+
     return prompt_variables, prompt_template
 
 
@@ -255,7 +248,7 @@ def get_caller_info(skip=2, return_short=True, return_string=True):
     start = 0 + skip
     if len(stack) < start + 1:
         return None
-    
+
     parentframe = stack[start][0]
 
     # module and packagename.
@@ -347,7 +340,7 @@ class BaseModelDict(BaseModel):
     def __getitem__(self, key):
         # deprecate dictionary usage
         deprecation_warning(
-            f'To get `{key}` use dot notation instead of dictionary keys, example:' 
+            f'To get `{key}` use dot notation instead of dictionary keys, example:'
             f'`obj.{key}` instead of `obj["{key}"]`'
         )
 

core/pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "Cheshire-Cat"
 description = "Production ready AI assistant framework"
-version = "1.9.1"
+version = "1.9.2"
 requires-python = ">=3.10"
 license = { file = "LICENSE" }
 authors = [

core/tests/cache/test_core_caches.py

@@ -16,18 +16,23 @@ def create_cache(cache_type):
         assert False
 
 
+
 @pytest.mark.parametrize("cache_type", ["in_memory", "file_system"])
 def test_cache_creation(cache_type):
-
-    cache = create_cache(cache_type)
-
-    if cache_type == "in_memory":
-        assert cache.items == {}
-        assert cache.max_items == 100
-    else:
-        assert cache.cache_dir == "/tmp_cache"
-        assert os.path.exists("/tmp_cache")
-        assert os.listdir("/tmp_cache") == []
+    try:
+        cache = create_cache(cache_type)
+
+        if cache_type == "in_memory":
+            assert cache.items == {}
+            assert cache.max_items == 100
+        else:
+            assert cache.cache_dir == "/tmp_cache"
+            assert os.path.exists("/tmp_cache")
+            assert os.listdir("/tmp_cache") == []
+    finally:
+        import shutil
+        if os.path.exists("/tmp_cache"):
+            shutil.rmtree("/tmp_cache")
 
 
 @pytest.mark.parametrize("cache_type", ["in_memory", "file_system"])
@@ -36,13 +41,13 @@ def test_cache_get_insert(cache_type):
     cache = create_cache(cache_type)
 
     assert cache.get_item("a") is None
-    
-    c1 = CacheItem("a", [])  
+
+    c1 = CacheItem("a", [])
     cache.insert(c1)
     assert cache.get_item("a").value == []
     assert cache.get_value("a") == []
 
-        
+
     c1.value = [0]
     cache.insert(c1) # will be overwritten
     assert cache.get_item("a").value == [0]
@@ -64,7 +69,7 @@ def test_cache_delete(cache_type):
 
     c1 = CacheItem("a", [])
     cache.insert(c1)
-    
+
     cache.delete("a")
     assert cache.get_item("a") is None