update redis to use newer vector store provider

Author: dminnear-rh

.env

@@ -20,7 +20,6 @@ EMBEDDING_LENGTH=768
 # === Redis ===
 REDIS_URL=redis://localhost:6379
 REDIS_INDEX=docs
-REDIS_SCHEMA=redis_schema.yaml
 
 # === Elasticsearch ===
 ELASTIC_URL=http://localhost:9200

Containerfile

@@ -18,7 +18,6 @@ COPY vector_db ./vector_db
 COPY loaders ./loaders
 COPY embed_documents.py .
 COPY config.py .
-COPY redis_schema.yaml .
 COPY .env .
 
 RUN chown -R 1001:0 .

config.py

@@ -114,8 +114,7 @@ def _init_db_provider(db_type: str) -> DBProvider:
         if db_type == "REDIS":
             url = get("REDIS_URL")
             index = os.getenv("REDIS_INDEX", "docs")
-            schema = os.getenv("REDIS_SCHEMA", "redis_schema.yaml")
-            return RedisProvider(embedding_model, url, index, schema)
+            return RedisProvider(embedding_model, url, index)
 
         elif db_type == "ELASTIC":
             url = get("ELASTIC_URL")

redis_schema.yaml

@@ -1,17 +1,16 @@
 beautifulsoup4
 hf_xet
+langchain
 langchain-community
 langchain-elasticsearch
 langchain-huggingface
 langchain-postgres
 langchain-qdrant
+langchain-redis
 langchain-sqlserver
-langchain
-psycopg[binary]
 pyodbc
 pypdf
 python-dotenv
 qdrant-client
-redis < 6.0.0
 sentence-transformers
 unstructured[md]

requirements.in

@@ -46,6 +46,8 @@ click==8.2.1
     # via
     #   nltk
     #   python-oxmsg
+coloredlogs==15.0.1
+    # via redisvl
 cryptography==45.0.3
     # via
     #   azure-identity
@@ -108,6 +110,8 @@ huggingface-hub==0.32.4
     #   sentence-transformers
     #   tokenizers
     #   transformers
+humanfriendly==10.0
+    # via coloredlogs
 hyperframe==6.1.0
     # via h2
 idna==3.10
@@ -117,7 +121,9 @@ idna==3.10
     #   requests
     #   yarl
 jinja2==3.1.6
-    # via torch
+    # via
+    #   langchain-redis
+    #   torch
 joblib==1.5.1
     # via
     #   nltk
@@ -140,6 +146,7 @@ langchain-core==0.3.63
     #   langchain-huggingface
     #   langchain-postgres
     #   langchain-qdrant
+    #   langchain-redis
     #   langchain-sqlserver
     #   langchain-text-splitters
 langchain-elasticsearch==0.3.2
@@ -150,6 +157,8 @@ langchain-postgres==0.0.14
     # via -r requirements.in
 langchain-qdrant==0.2.0
     # via -r requirements.in
+langchain-redis==0.2.0
+    # via -r requirements.in
 langchain-sqlserver==0.1.2
     # via -r requirements.in
 langchain-text-splitters==0.3.8
@@ -169,6 +178,8 @@ markupsafe==3.0.2
     # via jinja2
 marshmallow==3.26.1
     # via dataclasses-json
+ml-dtypes==0.4.1
+    # via redisvl
 mpmath==1.3.0
     # via sympy
 msal==1.32.3
@@ -194,9 +205,12 @@ numpy==1.26.4
     #   elasticsearch
     #   langchain-community
     #   langchain-postgres
+    #   langchain-redis
     #   langchain-sqlserver
+    #   ml-dtypes
     #   pgvector
     #   qdrant-client
+    #   redisvl
     #   scikit-learn
     #   scipy
     #   transformers
@@ -263,12 +277,8 @@ protobuf==6.31.1
     # via qdrant-client
 psutil==7.0.0
     # via unstructured
-psycopg[binary]==3.2.9
-    # via
-    #   -r requirements.in
-    #   langchain-postgres
-psycopg-binary==3.2.9
-    # via psycopg
+psycopg==3.2.9
+    # via langchain-postgres
 psycopg-pool==3.2.6
     # via langchain-postgres
 pycparser==2.22
@@ -281,6 +291,7 @@ pydantic==2.11.5
     #   langsmith
     #   pydantic-settings
     #   qdrant-client
+    #   redisvl
     #   unstructured-client
 pydantic-core==2.33.2
     # via pydantic
@@ -310,12 +321,17 @@ python-magic==0.4.27
     # via unstructured
 python-oxmsg==0.0.2
     # via unstructured
+python-ulid==3.0.0
+    # via
+    #   langchain-redis
+    #   redisvl
 pyyaml==6.0.2
     # via
     #   huggingface-hub
     #   langchain
     #   langchain-community
     #   langchain-core
+    #   redisvl
     #   transformers
 qdrant-client==1.14.2
     # via
@@ -324,7 +340,9 @@ qdrant-client==1.14.2
 rapidfuzz==3.13.0
     # via unstructured
 redis==5.2.1
-    # via -r requirements.in
+    # via redisvl
+redisvl==0.4.1
+    # via langchain-redis
 regex==2024.11.6
     # via
     #   nltk
@@ -376,10 +394,13 @@ sqlalchemy==2.0.41
     #   langchain-sqlserver
 sympy==1.14.0
     # via torch
+tabulate==0.9.0
+    # via redisvl
 tenacity==9.1.2
     # via
     #   langchain-community
     #   langchain-core
+    #   redisvl
 threadpoolctl==3.6.0
     # via scikit-learn
 tokenizers==0.21.1

requirements.txt

@@ -1,9 +1,8 @@
 import logging
-from typing import List, Optional
+from typing import List
 
-import redis
-from langchain_community.vectorstores.redis import Redis as RedisVectorStore
 from langchain_core.documents import Document
+from langchain_redis import RedisVectorStore
 
 from vector_db.db_provider import DBProvider
 
@@ -14,81 +13,44 @@ class RedisProvider(DBProvider):
     """
     Redis-backed vector DB provider using RediSearch and LangChain's Redis integration.
 
-    This provider connects to a Redis instance, checks if the specified index exists,
-    and either loads from it or creates a new index on first insert. Vectors are stored
-    using the RediSearch module with configurable schema.
-
     Attributes:
-        redis_client (redis.Redis): Raw Redis client for low-level access.
-        db (Optional[RedisVectorStore]): LangChain vector store, lazily created on first add.
+        db (RedisVectorStore): LangChain vector store
 
     Args:
         embedding_model (str): Name of the embedding model to use for text chunks.
         url (str): Redis connection string (e.g., "redis://localhost:6379").
         index (str): RediSearch index name to use for vector storage.
-        schema (str): Path to schema file where the RediSearch index definition is written.
 
     Example:
         >>> from vector_db.redis_provider import RedisProvider
         >>> provider = RedisProvider(
         ...     embedding_model="BAAI/bge-large-en-v1.5",
         ...     url="redis://localhost:6379",
-        ...     index="validated_docs",
-        ...     schema="redis_schema.yaml"
+        ...     index="validated_docs"
         ... )
         >>> provider.add_documents(docs)
     """
 
-    def __init__(self, embedding_model: str, url: str, index: str, schema: str):
+    def __init__(self, embedding_model: str, url: str, index: str):
         """
         Initialize a Redis-backed vector store provider.
 
         Args:
             embedding_model (str): HuggingFace model for embeddings.
             url (str): Redis connection string.
             index (str): Name of the RediSearch index to use.
-            schema (str): Path to write RediSearch schema YAML (used on creation).
         """
         super().__init__(embedding_model)
-        self.url = url
-        self.index = index
-        self.schema = schema
-        self.db: Optional[RedisVectorStore] = None
-
-        try:
-            self.redis_client = redis.from_url(self.url)
-            self.redis_client.ping()
-            logger.info("Connected to Redis instance at %s", self.url)
-        except Exception:
-            logger.exception("Failed to connect to Redis at %s", self.url)
-            raise
 
-        if self._index_exists():
-            logger.info("Using existing Redis index: %s", self.index)
-            self.db = RedisVectorStore.from_existing_index(
-                embedding=self.embeddings,
-                redis_url=self.url,
-                index_name=self.index,
-                schema=self.schema,
-            )
-        else:
-            logger.info(
-                "Redis index %s does not exist. Will create on first add_documents call.",
-                self.index,
-            )
+        self.db = RedisVectorStore(
+            index_name=index, embeddings=self.embeddings, redis_url=url
+        )
 
-    def _index_exists(self) -> bool:
-        """
-        Check whether the Redis index already exists.
-
-        Returns:
-            bool: True if the index exists, False otherwise.
-        """
-        try:
-            self.redis_client.ft(self.index).info()
-            return True
-        except Exception:
-            return False
+        logger.info(
+            "Connected to Redis at %s (index: %s)",
+            url,
+            index,
+        )
 
     def add_documents(self, docs: List[Document]) -> None:
         """
@@ -97,15 +59,4 @@ def add_documents(self, docs: List[Document]) -> None:
         Args:
             docs (List[Document]): LangChain document chunks to embed and store.
         """
-        if self.db is None:
-            logger.info("Creating new Redis index: %s", self.index)
-            self.db = RedisVectorStore.from_documents(
-                documents=docs,
-                embedding=self.embeddings,
-                redis_url=self.url,
-                index_name=self.index,
-            )
-            logger.info("Writing Redis schema to file: %s", self.schema)
-            self.db.write_schema(self.schema)
-        else:
-            self.db.add_documents(docs)
+        self.db.add_documents(docs)