Get started
Get started with RedisVL
RedisVL is a versatile Python library with an integrated CLI, which is designed to enhance AI applications implemented using Redis. This guide will walk you through the following steps:
- Define an
IndexSchema. - Prepare a sample dataset.
- Create a
SearchIndexobject. - Test
rvlCLI functionality. - Load the sample data.
- Build
VectorQueryobjects and execute searches. - Update a
SearchIndexobject.
Before beginning, be sure of the following:
- You have installed RedisVL and have that environment activated.
- You have a running Redis instance with the search and query capability.
Define an IndexSchema
The IndexSchema maintains crucial index configuration and field definitions to
enable search with Redis. For ease of use, the schema can be constructed from a
Python dictionary or a YAML file.
Example schema creation
Consider a dataset with user information, including job, age, credit_score,
and a three-dimensional user_embedding vector.
You must decide on a Redis index name and key prefix to use for this
dataset. Below are example schema definitions in both YAML and Python dict formats.
YAML definition:
version: '0.1.0'
index:
name: user_simple
prefix: user_simple_docs
fields:
- name: user
type: tag
- name: credit_store
type: tag
- name: job
type: text
- name: age
type: numeric
- name: user_embedding
type: vector
attrs:
algorithm: flat
dims: 3
distance_metric: cosine
datatype: float32
Store this information in a local file, such as schema.yaml, for use with RedisVL.
Python dictionary:
schema = {
"index": {
"name": "user_simple",
"prefix": "user_simple_docs",
},
"fields": [
{"name": "user", "type": "tag"},
{"name": "credit_score", "type": "tag"},
{"name": "job", "type": "text"},
{"name": "age", "type": "numeric"},
{
"name": "user_embedding",
"type": "vector",
"attrs": {
"dims": 3,
"distance_metric": "cosine",
"algorithm": "flat",
"datatype": "float32"
}
}
]
}
Sample dataset preparation
Below, create a mock dataset with user, job, age, credit_score, and
user_embedding fields. The user_embedding vectors are synthetic examples
for demonstration purposes.
For more information on creating real-world embeddings, refer to this article.
import numpy as np
data = [
{
'user': 'john',
'age': 1,
'job': 'engineer',
'credit_score': 'high',
'user_embedding': np.array([0.1, 0.1, 0.5], dtype=np.float32).tobytes()
},
{
'user': 'mary',
'age': 2,
'job': 'doctor',
'credit_score': 'low',
'user_embedding': np.array([0.1, 0.1, 0.5], dtype=np.float32).tobytes()
},
{
'user': 'joe',
'age': 3,
'job': 'dentist',
'credit_score': 'medium',
'user_embedding': np.array([0.9, 0.9, 0.1], dtype=np.float32).tobytes()
}
]
As seen above, the sample user_embedding vectors are converted into bytes using the NumPy Python package.
Create a SearchIndex
With the schema and sample dataset ready, create a SearchIndex:
from redisvl.index import SearchIndex
index = SearchIndex.from_dict(schema)
# or use .from_yaml('schema_file.yaml')
Now we also need to create a Redis connection. There are a few ways to do this:
- Create and manage your own client connection (recommended).
- Provide a simple Redis URL and let RedisVL connect on your behalf.
Bring your own Redis connection instance
This is ideal in scenarios where you have custom settings on the connection instance or if your application will share a connection pool:
from redis import Redis
client = Redis.from_url("redis://localhost:6379")
index.set_client(client)
# optionally provide an async Redis client object to enable async index operations
Let the index manage the connection instance
This is ideal for simple cases:
index.connect("redis://localhost:6379")
# optionally use an async client by passing use_async=True
Create the underlying index
Now that there's a connection to Redis, run the create command.
index.create(overwrite=True)
Note: at this point, the index has no associated data. Data loading follows.
Inspect with the rvl command
Use the rvl CLI command to inspect the newly-created index and its fields:
$ rvl index listall
18:25:34 [RedisVL] INFO Indices:
18:25:34 [RedisVL] INFO 1. user_simple
$ rvl index info -i user_simple
╭──────────────┬────────────────┬──────────────────────┬─────────────────┬────────────╮
│ Index Name │ Storage Type │ Prefixes │ Index Options │ Indexing │
├──────────────┼────────────────┼──────────────────────┼─────────────────┼────────────┤
│ user_simple │ HASH │ ['user_simple_docs'] │ [] │ 0 │
╰──────────────┴────────────────┴──────────────────────┴─────────────────┴────────────╯
Index Fields:
╭────────────────┬────────────────┬─────────┬────────────────┬────────────────╮
│ Name │ Attribute │ Type │ Field Option │ Option Value │
├────────────────┼────────────────┼─────────┼────────────────┼────────────────┤
│ user │ user │ TAG │ SEPARATOR │ , │
│ credit_score │ credit_score │ TAG │ SEPARATOR │ , │
│ job │ job │ TEXT │ WEIGHT │ 1 │
│ age │ age │ NUMERIC │ │ │
│ user_embedding │ user_embedding │ VECTOR │ │ │
╰────────────────┴────────────────┴─────────┴────────────────┴────────────────╯
Load data to SearchIndex
Load the sample dataset to Redis:
keys = index.load(data)
print(keys)
['user:31d4f3c73f1a4c26b41cf0e2b8e0248a',
'user:c9ff740437064b919245e49ef585484d',
'user:6db5f2e09f08438785b73d8048d5350b']
By default, load will create a unique Redis key as a combination of the index key prefix and a UUID. You can also customize the key by providing direct keys or pointing to a specified id_field on load.
Update the index with new data
Update data using the load method:
# Add more data
new_data = [{
'user': 'tyler',
'age': 9,
'job': 'engineer',
'credit_score': 'high',
'user_embedding': np.array([0.1, 0.3, 0.5], dtype=np.float32).tobytes()
}]
keys = index.load(new_data)
print(keys)
['user_simple_docs:ea6e8f2f93d5447c950ccb6843627761']
Create VectorQuery objects
Next, create a vector query object for your newly-populated index. This example will use a simple vector to demonstrate how vector search works. Vectors in production will likely be much larger than three floating point numbers and often require machine learning models (e.g., Huggingface sentence transformers) or an embeddings API (e.g., Cohere and OpenAI). RedisVL provides a set of vectorizers to assist with vector creation.
from redisvl.query import VectorQuery
from jupyterutils import result_print
query = VectorQuery(
vector=[0.1, 0.1, 0.5],
vector_field_name="user_embedding",
return_fields=["user", "age", "job", "credit_score", "vector_distance"],
num_results=3
)
Executing queries
With your VectorQuery object defined, you can execute the query over the SearchIndex using the query method.
results = index.query(query)
result_print(results)
| vector_distance | user | age | job | credit_score |
|---|---|---|---|---|
| 0 | john | 1 | engineer | high |
| 0 | mary | 2 | doctor | low |
| 0.0566299557686 | tyler | 9 | engineer | high |
Using an asynchronous Redis client
The AsyncSearchIndex class, along with an asynchronous Redis Python client, provides for asynchronous queries, index creation, and data loading. This is the
recommended way for working with redisvl in production settings.
from redisvl.index import AsyncSearchIndex
from redis.asyncio import Redis
client = Redis.from_url("redis://localhost:6379")
index = AsyncSearchIndex.from_dict(schema)
index.set_client(client)
# execute the vector query async
results = await index.aquery(query)
result_print(results)
| vector_distance | user | age | job | credit_score |
|---|---|---|---|---|
| 0 | john | 1 | engineer | high |
| 0 | mary | 2 | doctor | low |
| 0.0566299557686 | tyler | 9 | engineer | high |
Update a schema
In some scenarios, it makes sense to update the index schema. With Redis and RedisVL, this is easy because Redis can keep the underlying data in place while you update to the index configuration.
Imagine you want to re-index this data in the following ways:
- Use a
Tagtype for thejobfield instead ofText. - Use an
HNSWvector index for theuser_embeddingfield instead of aflatvector index.
# Modify this schema to have what we want
index.schema.remove_field("job")
index.schema.remove_field("user_embedding")
index.schema.add_fields([
{"name": "job", "type": "tag"},
{
"name": "user_embedding",
"type": "vector",
"attrs": {
"dims": 3,
"distance_metric": "cosine",
"algorithm": "flat",
"datatype": "float32"
}
}
])
# Run the index update but keep underlying data in place
await index.create(overwrite=True, drop=False)
# Execute the vector query
results = await index.aquery(query)
result_print(results)
| vector_distance | user | age | job | credit_score |
|---|---|---|---|---|
| 0 | john | 1 | engineer | high |
| 0 | mary | 2 | doctor | low |
| 0.0566299557686 | tyler | 9 | engineer | high |
Check index stats
Use rvl to check the statistics for the index:
$ rvl stats -i user_simple
Statistics:
╭─────────────────────────────┬─────────────╮
│ Stat Key │ Value │
├─────────────────────────────┼─────────────┤
│ num_docs │ 4 │
│ num_terms │ 0 │
│ max_doc_id │ 4 │
│ num_records │ 20 │
│ percent_indexed │ 1 │
│ hash_indexing_failures │ 0 │
│ number_of_uses │ 2 │
│ bytes_per_record_avg │ 1 │
│ doc_table_size_mb │ 0.00044632 │
│ inverted_sz_mb │ 1.90735e-05 │
│ key_table_size_mb │ 0.000165939 │
│ offset_bits_per_record_avg │ nan │
│ offset_vectors_sz_mb │ 0 │
│ offsets_per_term_avg │ 0 │
│ records_per_doc_avg │ 5 │
│ sortable_values_size_mb │ 0 │
│ total_indexing_time │ 0.246 │
│ total_inverted_index_blocks │ 11 │
│ vector_index_sz_mb │ 0.0201416 │
╰─────────────────────────────┴─────────────╯
Cleanup
# clean up the index
await index.adelete()