## New EmbeddingStore (metadata) `Filter` API
Many embedding stores, such as
[Pinecone](https://docs.pinecone.io/docs/metadata-filtering) and
[Milvus](https://milvus.io/docs/boolean.md) support strict filtering
(think of an SQL "WHERE" clause) during similarity search.
So, if one has an embedding store with movies, for example, one could
search not only for the most semantically similar movies to the given
user query but also apply strict filtering by metadata fields like year,
genre, rating, etc. In this case, the similarity search will be
performed only on those movies that match the filter expression.
Since LangChain4j supports (and abstracts away) many embedding stores,
there needs to be an embedding-store-agnostic way for users to define
the filter expression.
This PR introduces a `Filter` interface, which can represent both simple
(e.g., `type = "documentation"`) and composite (e.g., `type in
("documentation", "tutorial") AND year > 2020`) filter expressions in an
embedding-store-agnostic manner.
`Filter` currently supports the following operations:
- Comparison:
- `IsEqualTo`
- `IsNotEqualTo`
- `IsGreaterThan`
- `IsGreaterThanOrEqualTo`
- `IsLessThan`
- `IsLessThanOrEqualTo`
- `IsIn`
- `IsNotIn`
- Logical:
- `And`
- `Not`
- `Or`
These operations are supported by most embedding stores and serve as a
good starting point. However, the list of operations will expand over
time to include other operations (e.g., `Contains`) supported by
embedding stores.
Currently, the DSL looks like this:
```java
Filter onlyDocs = metadataKey("type").isEqualTo("documentation");
Filter docsAndTutorialsAfter2020 = metadataKey("type").isIn("documentation", "tutorial").and(metadataKey("year").isGreaterThan(2020));
// or
Filter docsAndTutorialsAfter2020 = and(
metadataKey("type").isIn("documentation", "tutorial"),
metadataKey("year").isGreaterThan(2020)
);
```
## Filter expression as a `String`
Filter expression can also be specified as a `String`. This might be
necessary, for example, if the filter expression is generated
dynamically by the application or by the LLM (as in [self
querying](https://python.langchain.com/docs/modules/data_connection/retrievers/self_query/)).
This PR introduces a `FilterParser` interface with a simple `Filter
parse(String)` API, allowing for future support of multiple syntaxes (if
this will be required).
For the out-of-the-box filter syntax, ANSI SQL's `WHERE` clause is
proposed as a suitable candidate for several reasons:
- SQL is well-known among Java developers
- There is extensive tooling available for SQL (e.g., parsers)
- LLMs are pretty good at generating valid SQL, as there are tons of SQL
queries on the internet, which are included in the LLM training
datasets. There are also specialized LLMs that are trained for
text-to-SQL task, such as [SQLCoder](https://huggingface.co/defog).
The downside is that SQL's `WHERE` clause might not support all
operations and data types that could be supported in the future by
various embedding stores. In such case, we could extend it to a superset
of ANSI SQL `WHERE` syntax and/or provide an option to express filters
in the native syntax of the store.
An out-of-the-box implementation of the SQL `FilterParser` is provided
as a `SqlFilterParser` in a separate module
`langchain4j-embedding-store-filter-parser-sql`, using
[JSqlParser](https://github.com/JSQLParser/JSqlParser) under the hood.
`SqlFilterParser` can parse SQL "SELECT" (or just "WHERE" clause)
statement into a `Filter` object:
- `SELECT * FROM fake_table WHERE userId = '123-456'` ->
`metadataKey("userId").isEqualTo("123-456")`
- `userId = '123-456'` -> `metadataKey("userId").isEqualTo("123-456")`
It can also resolve `CURDATE()` and
`CURRENT_DATE`/`CURRENT_TIME`/`CURRENT_TIMESTAMP`:
`SELECT * FROM fake_table WHERE year = EXTRACT(YEAR FROM CURRENT_DATE`
-> `metadataKey("year").isEqualTo(LocalDate.now().getYear())`
## Changes in `Metadata` API
Until now, `Metadata` supported only `String` values. This PR expands
the list of supported value types to `Integer`, `Long`, `Float` and
`Double`. In the future, more types may be added (if needed).
The method `String get(String key)` will be deprecated later in favor
of:
- `String getString(String key)`
- `Integer getInteger(String key)`
- `Long getLong(String key)`
- etc
New overloaded `put(key, value)` methods are introduced to support more
value types:
- `put(String key, int value)`
- `put(String key, long value)`
- etc
## Changes in `EmbeddingStore` API
New method `search` is added that will become the main entry point for
search in the future. All `findRelevant` methods will be deprecated
later.
New `search` method accepts `EmbeddingSearchRequest` and returns
`EmbeddingSearchResult`.
`EmbeddingSearchRequest` contains all search criteria (e.g.
`maxResults`, `minScore`), including new `Filter`.
`EmbeddingSearchResult` contains a list of `EmbeddingMatch`.
```java
EmbeddingSearchResult search(EmbeddingSearchRequest request);
```
## Changes in `EmbeddingStoreContentRetriever` API
`EmbeddingStoreContentRetriever` can now be configured with a static
`filter` as well as dynamic `dynamicMaxResults`, `dynamicMinScore` and
`dynamicFilter` in the builder:
```java
ContentRetriever contentRetriever = EmbeddingStoreContentRetriever.builder()
.embeddingStore(embeddingStore)
.embeddingModel(embeddingModel)
...
.maxResults(3)
// or
.dynamicMaxResults(query -> 3) // You can define maxResults dynamically. The value could, for example, depend on the query or the user associated with the query.
...
.minScore(0.3)
// or
.dynamicMinScore(query -> 0.3)
...
.filter(metadataKey("userId").isEqualTo("123-456")) // Assuming your TextSegments contain Metadata with key "userId"
// or
.dynamicFilter(query -> metadataKey("userId").isEqualTo(query.metadata().chatMemoryId().toString()))
...
.build();
```
So now you can define `maxResults`, `minScore` and `filter` both
statically and dynamically (they can depend on the query, user, etc.).
These values will be propagated to the underlying `EmbeddingStore`.
##
["Self-querying"](https://python.langchain.com/docs/modules/data_connection/retrievers/self_query/)
This PR also introduces `LanguageModelSqlFilterBuilder` in
`langchain4j-embedding-store-filter-parser-sql` module which can be used
with `EmbeddingStoreContentRetriever`'s `dynamicFilter` to automatically
build a `Filter` object from the `Query` using language model and
`SqlFilterParser`.
For example:
```java
TextSegment groundhogDay = TextSegment.from("Groundhog Day", new Metadata().put("genre", "comedy").put("year", 1993));
TextSegment forrestGump = TextSegment.from("Forrest Gump", new Metadata().put("genre", "drama").put("year", 1994));
TextSegment dieHard = TextSegment.from("Die Hard", new Metadata().put("genre", "action").put("year", 1998));
// describe metadata keys as if they were columns in the SQL table
TableDefinition tableDefinition = TableDefinition.builder()
.name("movies")
.addColumn("genre", "VARCHAR", "one of [comedy, drama, action]")
.addColumn("year", "INT")
.build();
LanguageModelSqlFilterBuilder sqlFilterBuilder = new LanguageModelSqlFilterBuilder(model, tableDefinition);
ContentRetriever contentRetriever = EmbeddingStoreContentRetriever.builder()
.embeddingStore(embeddingStore)
.embeddingModel(embeddingModel)
.dynamicFilter(sqlFilterBuilder::build)
.build();
String answer = assistant.answer("Recommend me a good drama from 90s"); // Forrest Gump
```
## Which embedding store integrations will support `Filter`?
In the long run, all (provided the embedding store itself supports it).
In the first iteration, I aim to add support to just a few:
- `InMemoryEmbeddingStore`
- Elasticsearch
- Milvus
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
## Summary by CodeRabbit
- **New Features**
- Introduced filters for checking key's value existence in a collection
for improved data handling.
- **Enhancements**
- Updated `InMemoryEmbeddingStoreTest` to extend a different class for
improved testing coverage and added a new test method.
- **Refactor**
- Made minor formatting adjustments in the assertion block for better
readability.
- **Documentation**
- Updated class hierarchy information for clarity.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
In the Datastax Astra DB saas solution, a new way to integrate with
vector databases has been introduced: using an HTTP APi instead of the
Cassandra Cluster. It is called the DataAPI and use the MongoDB
principles with collections.
The pull request includes the following:
### Update on previous implementations
- Previous implementations of embedding stores have been grouped in a
single `CassandraEmbeddingStore`. It can be instantiated for Astra or
OSS Cassandra based on 2 different constructor builders but everything
else is the same.
- Previous implementations of chat memory stores have been grouped in a
single `CassandraChatMemoryStore`. It can be instantiated for Astra or
OSS Cassandra based on 2 different constructor builders but everything
else is the same.
- Integration test for OSS Cassandra now using test containers (as
Cassandra 5-alpha2 image is out)
- Usage
```java
// Using with Astra (Cassandra AAS in the cloud)
CassandraEmbeddingStore.builderAstra()
.token(token)
.databaseId(dbId)
.databaseRegion(TEST_REGION)
.keyspace(KEYSPACE)
.table(TEST_INDEX)
.dimension(11)
.metric(CassandraSimilarityMetric.COSINE)
.build();
// Using OSS Cassandra
CassandraEmbeddingStore.builder()
.contactPoints(Arrays.asList(contactPoint.getHostName()))
.port(contactPoint.getPort())
.localDataCenter(DATACENTER)
.keyspace(KEYSPACE)
.table(TEST_INDEX)
.dimension(11)
.metric(CassandraSimilarityMetric.COSINE)
.build();
```
-Adding jdk11 in the pom
```
<maven.compiler.source>11</maven.compiler.source>
<maven.compiler.target>11</maven.compiler.target>
```
- introducing `insertMany()`, distributed to all bulk loading
- Extending the variables `EmbeddingStoreIT`
- Using `MessageWindowChatMemory` for the tests.
- Updated subpage 0 to `Overview` and made it the landing page upon
clicking `Tutorials` from the horizontal/navigation/top bar.
- Set up GitHub action to build and publish the docs to the GitHub pages
on release. [Preview](https://amithkoujalgi.github.io/langchain4j/) of
the docs on GitHub pages. This action/workflow can also be triggered
manually.
- Added content to `/integrations/language-models/ollama`
PS:
If GitHub pages has been enabled in the repository, the repo admin
should be able to run the new GitHub workflow/action in the Actions tab
and after the completion of the workflow, the docs should be available
in the link provided by GitHub (or a specified custom domain).
Fixes#241: Added support for Neo4j Vector Index
This commit brings support for Neo4j graph database in general, and uses
the vector index functionality, generally available since version 5.13.
Mostly aligned with the existing WeaviateEmbeddingStoreImpl
implementation and tests.
The tests have some additional Neo4j node assertion to check that the
nodes involved are correctly created.
The module creates indexes, i.e. `"CALL
db.index.vector.createNodeIndex(<indexName>, <label>,
<embeddingProperty>, <dimension>, <distanceType>)"`, if needed, for the
vector search .
The required configurations are:
- the Neo4j index dimension parameter
- the Neo4j Java Driver connection instance
- as an alternative to the Neo4j Java Driver, we can create a
`Neo4jEmbeddingStore.builder().withBasicAuth(<url>, <username>,
<password>)`, which will create a Driver connection instance under the
hood
It is possible to customize, via the builder:
- the index name (with default `langchain-embedding-index`)
- the Neo4j node label (with default `Document`)
- the Neo4j property key which save the embeddings (with default
`embeddingProp`)
- the Neo4j index distanceType parameter
- the metadata prefix (with default `metadata.`)
- the text property key (with default `text`), which store the text
field of the `TextSegment.java`
Created an example PR as well, on `langchain4j-examples` repo:
https://github.com/langchain4j/langchain4j-examples/pull/23
In the GitHub Actions workflow:
- Update actions/checkout to the latest version
- Update actions/setup-java to the latest version (Java 21 already works
but is undocumented, the next version it will be thanks to
https://github.com/actions/setup-java/pull/538😀)
#### Context
Apache Cassandra is a popular open-source database created back in 2008.
This year with
[CEP30](https://cwiki.apache.org/confluence/display/CASSANDRA/CEP-30%3A+Approximate+Nearest+Neighbor%28ANN%29+Vector+Search+via+Storage-Attached+Indexes)
support for vector and similarity searches have been introduced.
Cassandra is very fast in read and write and is used as a cache by many
companies, it as an opportunity to implement the ChatMemoryStore. This
feature is expected for Cassandra 5 at the end of the year but some
docker images are already available.
DataStax AstraDb is a distribution of Apache Cassandra available as Saas
providing a free tier (free forever) of 80 millions queries/month.
[Registration](https://astra.datastax.com). The vector capability is
there production ready.
#### Data Modelling
With the proper data model in Cassandra we can perform both similarity
search, keyword search, metadata search.
```sql
CREATE TABLE sample_vector_table (
row_id text PRIMARY KEY,
attributes_blob text,
body_blob text,
metadata_s map<text, text>,
vector vector<float, 1536>
);
```
#### Implementation Throughts
- The **configuration** to connect to Astra and Cassandra are not
exactly the same so 2 different classes with associated builder are
provided:
[Astra](https://github.com/clun/langchain4j/blob/main/langchain4j/src/main/java/dev/langchain4j/store/embedding/cassandra/AstraDbEmbeddingConfiguration.java)
and [OSS
Cassandra](https://github.com/clun/langchain4j/blob/main/langchain4j/src/main/java/dev/langchain4j/store/embedding/cassandra/CassandraEmbeddingConfiguration.java).
A couple of fields are mutualized but creating a superclass to inherit
from lead to the use of Lombok `@SuperBuilder` and the Javadoc was not
able to found out what to do.
- Instead of passing a large number of arguments like other stores I
prefer to wrap them as a bean. With this trick you can add or remove
attributes, make then optional or mandatory at will. If you need to add
a new attribute in the configuration you do not have to change the
implementation of `XXXStore` and `XXXStoreImpl`
- I create an
[AstractEmbeddedStore<T>](https://github.com/clun/langchain4j/blob/main/langchain4j/src/main/java/dev/langchain4j/store/embedding/AbstractEmbeddingStore.java)
that could very well become the super class for any store. It handles
the different call of the real concrete implementation. (_delegate
pattern_). Some default implementation can be implemented
```java
/**
* Add a list of embeddings to the store.
*
* @param embeddings
* list of embeddings (hold vector)
* @return
* list of ids
*/
@Override
public List<String> addAll(List<Embedding> embeddings) {
Objects.requireNonNull(embeddings, "embeddings must not be null");
return embeddings.stream().map(this::add).collect(Collectors.toList());
}
```
The only method to implement at the Store level is:
```java
/**
* Initialize the concrete implementation.
* @return create implementation class for the store
*/
protected abstract EmbeddingStore<T> loadImplementation()
throws ClassNotFoundException, NoSuchMethodException, InstantiationException,
IllegalAccessException, InvocationTargetException;
```
-
[CassandraEmbeddedStore](https://github.com/clun/langchain4j/blob/main/langchain4j/src/main/java/dev/langchain4j/store/embedding/cassandra/CassandraEmbeddingStore.java#L30)
proposes 2 constructors, one could override the implementation class if
they want (extension point)
#### Tests
- Test classes are provided including some long form examples based on
classed found in `langchain4j-examples` but test are disabled.
- To start a local cassandra use docker and the
[docker-compose](https://github.com/clun/langchain4j/blob/main/langchain4j-cassandra/src/test/resources/docker-compose.yml)
```
docker compose up -d
```
- To run Test with Astra signin with your github account, create a token
(api Key) with role `Organization Administrator` following this
[procedure](https://awesome-astra.github.io/docs/pages/astra/create-token/#c-procedure)
<img width="926" alt="Screenshot 2023-09-06 at 18 14 12"
src="https://github.com/langchain4j/langchain4j/assets/726536/dfd2d9e5-09c9-4504-bfaa-31cfd87704a1">
- Pick the full value of the `token` from the json
<img width="713" alt="Screenshot 2023-09-06 at 18 15 53"
src="https://github.com/langchain4j/langchain4j/assets/726536/1be56234-dd98-4f59-af71-03df42ed6997">
- Create the environment variable `ASTRA_DB_APPLICATION_TOKEN`
```console
export ASTRA_DB_APPLICATION_TOKEN=AstraCS:....<your_token>
```
Some leftovers from an earlier (and now incorrect) CI configuration.
Modules that don't need to comply with the licenses need to deactivate
the relevant plugin on a case-by-case basis.
- all-minilm-l6-v2
- all-minilm-l6-v2-q
- e5-small-v2
- e5-small-v2-q
The idea is to give users an option to embed documents/texts in the same
Java process without any external dependencies.
ONNX Runtime is used to run models inside JVM.
Each model resides in it's own maven module (inside the jar).