介绍 Redis 向量库，助力增强生成式 AI 开发

pip install redisvl>=0.1.0

docker run -d -p 6379:6379 -p 8001:8001 redis/redis-stack:latest

index:
  name: sec-filings
  prefix: chunk

fields:
  - name: id
    type: tag
    attrs:
      sortable: true
  - name: content
    type: text
    attrs:
      sortable: true
  - name: company
    type: tag
    attrs:
      sortable: true
  - name: timestamp
    type: numeric
    attrs:
      sortable: true
  - name: content_embedding
    type: vector
    attrs:
      dims: 1024
      algorithm: hnsw
      datatype: float32
      distance_metric: cosine

from redisvl.schema import IndexSchema

schema = IndexSchema.from_yaml("schema.yaml")

from redis import Redis
from redisvl.index import SearchIndex

# Establish a connection with Redis
client = Redis.from_url("redis://localhost:6379")

# Link the schema with our Redis client to create the search index
index = SearchIndex(schema, client)

# Create the index in Redis
index.create()

from redisvl.utils.vectorize import CohereTextVectorizer

# Instantiate the Cohere text vectorizer
co = CohereTextVectorizer()

# Generate an embedding for a single query
embedding = co.embed(
    "How much debt is the company in?", input_type="search_query"
)

# Generate embeddings for multiple queries
embeddings = co.embed_many([
    "How much debt is the company in?",
    "What do revenue projections look like?"
], input_type="search_query")

# Example dataset as a list of dictionaries
data = [
    {
        "id": "doc1",
        "content": "raw SEC filing text content",
        "company": "nike",
        "timestamp": 20230101,
        "content_embedding": co.embed(
            "raw SEC filing text content",  input_type="search_document", as_buffer=True
        )
    },
    # More records...
]

# Insert data into the index
keys = index.load(data)

from redisvl.query import VectorQuery

query = "How much debt is the company in?"

query_vector = co.embed(query, input_type="search_query", as_buffer=True)

query = VectorQuery(
    vector=query_vector, 
    vector_field_name="content_embedding",
    num_results=5
)

results = index.query(query)

from redisvl.query.filter import Tag

# Apply a filter for the company name
query.set_filter(Tag("company") == "nike")

# Execute the filtered query
results = index.query(query)

from redisvl.extensions.llmcache import SemanticCache

# Set up the LLM cache
llmcache = SemanticCache(
    name="llmcache",                     # underlying search index name
    redis_url="redis://localhost:6379",  # redis connection url string
    distance_threshold=0.2               # semantic cache distance threshold
)

# Cache the question, answer, and arbitrary metadata
llmcache.store(
    prompt="What is the capital city of France?",
    response="Paris",
    metadata={"city": "Paris", "country": "france"}
)

# Check for a semantically similar result
question = "What actually is the capital of France?"

llmcache.check(prompt=question)[0]['response']

>>> 'Paris'

pip install redisvl>=0.1.0