fixed but when using Vectara query engine in non-summary mode (#11668)

* fixed but when using Vecrtara query engine in non-summary node - it needs to have output synthesizers after refactor * updated vectara autoRetriever example notebook * bugfix * bugfix in handling error updated documentId has to be 64 characters (instead of longer) * fixes from review * fix per review
run-llama · Mar 7, 2024 · a3491cf · a3491cf
1 parent d0cc330
commit a3491cf
Show file tree

Hide file tree

Showing 2 changed files with 24 additions and 7 deletions.
diff --git a/docs/examples/retrievers/vectara_auto_retriever.ipynb b/docs/examples/retrievers/vectara_auto_retriever.ipynb
@@ -56,7 +56,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!pip install llama-index"
+    "!pip install llama-index llama-index-indices-managed-vectara llama-index-llms-openai"
    ]
   },
   {
@@ -222,6 +222,7 @@
    "source": [
     "import getpass\n",
     "import openai\n",
+    "import os\n",
     "\n",
     "if not os.environ.get(\"OPENAI_API_KEY\", None):\n",
     "    os.environ[\"OPENAI_API_KEY\"] = getpass.getpass(\"OpenAI API Key:\")\n",
@@ -290,11 +291,17 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "llm = OpenAI(model=\"gpt-4-1106-preview\", temperature=0)\n",
+    "from llama_index.indices.managed.vectara import VectaraAutoRetriever\n",
+    "from llama_index.core.indices.service_context import ServiceContext\n",
+    "from llama_index.llms.openai import OpenAI\n",
+    "\n",
+    "llm = OpenAI(model=\"gpt-3.5-turbo\", temperature=0)\n",
     "\n",
     "retriever = VectaraAutoRetriever(\n",
     "    index,\n",
     "    vector_store_info=vector_store_info,\n",
+    "    llm=llm,\n",
+    "    verbose=False,\n",
     ")"
    ]
   },
@@ -395,6 +402,7 @@
     "retriever = VectaraAutoRetriever(\n",
     "    index,\n",
     "    vector_store_info=vector_store_info,\n",
+    "    llm=llm,\n",
     "    filter=\"doc.rating > 8\",\n",
     ")\n",
     "retriever.retrieve(\"movie about toys\")"
@@ -438,6 +446,7 @@
     "retriever = VectaraAutoRetriever(\n",
     "    index,\n",
     "    vector_store_info=vector_store_info,\n",
+    "    llm=llm,\n",
     "    filter=\"doc.rating > 8\",\n",
     "    vectara_query_mode=\"mmr\",\n",
     "    mmr_k=50,\n",

diff --git a/...s/indices/llama-index-indices-managed-vectara/llama_index/indices/managed/vectara/base.py b/...s/indices/llama-index-indices-managed-vectara/llama_index/indices/managed/vectara/base.py
@@ -30,6 +30,9 @@
 from llama_index.core.settings import Settings
 from llama_index.core.storage.storage_context import StorageContext
 
+from llama_index.core.response_synthesizers import ResponseMode
+from llama_index.core import get_response_synthesizer
+
 _logger = logging.getLogger(__name__)
 
 
@@ -193,13 +196,12 @@ def _index_doc(self, doc: dict) -> str:
         )
 
         status_code = response.status_code
-
         result = response.json()
 
         status_str = result["status"]["code"] if "status" in result else None
-        if status_code == 409 or status_str and (status_str == "ALREADY_EXISTS"):
+        if status_code == 409 and status_str and (status_str == "ALREADY_EXISTS"):
             return "E_ALREADY_EXISTS"
-        elif status_code == 200 or status_str and (status_str == "INVALID_ARGUMENT"):
+        elif status_code == 200 and status_str and (status_str == "INVALID_ARGUMENT"):
             return "E_INVALID_ARGUMENT"
         elif status_str and (status_str == "FORBIDDEN"):
             return "E_NO_PERMISSIONS"
@@ -215,7 +217,7 @@ def _insert(
         """Insert a set of documents (each a node)."""
 
         def gen_hash(s: str) -> str:
-            hash_object = blake2b()
+            hash_object = blake2b(digest_size=32)
             hash_object.update(s.encode("utf-8"))
             return hash_object.hexdigest()
 
@@ -360,8 +362,14 @@ def as_query_engine(
             )
 
             retriever = self.as_retriever(**kwargs)
+            response_synthesizer = get_response_synthesizer(
+                response_mode=ResponseMode.COMPACT,
+                llm=llm,
+            )
             return RetrieverQueryEngine.from_args(
-                retriever=retriever, llm=llm, **kwargs
+                retriever=retriever,
+                response_synthesizer=response_synthesizer,
+                **kwargs,
             )
 
     @classmethod