Moss integration

docs/docs.json

@@ -186,6 +186,7 @@
                       "integrations/vector-db-integrations/chromadb",
                       "integrations/vector-db-integrations/couchbase",
                       "integrations/vector-db-integrations/milvus",
+                      "integrations/vector-db-integrations/moss",
                       "integrations/vector-db-integrations/pgvector",
                       "integrations/vector-db-integrations/pinecone",
                       "integrations/vector-db-integrations/weaviate"

docs/integrations/vector-db-integrations/moss.mdx

@@ -0,0 +1,117 @@
+---
+title: Moss
+sidebarTitle: Moss
+---
+
+In this section, we present how to connect Moss to MindsDB.
+
+[Moss](https://moss.dev) is a semantic search runtime for Conversational AI agents. It delivers hybrid search with sub-10ms latency.
+
+## Prerequisites
+
+Before proceeding, ensure the following prerequisites are met:
+
+1. Install MindsDB locally via [Docker](/setup/self-hosted/docker) or [Docker Desktop](/setup/self-hosted/docker-desktop).
+2. To connect Moss to MindsDB, install the required dependencies following [this instruction](/setup/self-hosted/docker#install-dependencies).
+3. Create a Moss account and obtain your project credentials from [portal.usemoss.dev](https://portal.usemoss.dev).
+
+## Connection
+
+This handler is implemented using the `moss` Python library.
+
+To connect Moss to MindsDB, use the following statement:
+
+```sql
+CREATE DATABASE moss_db
+WITH ENGINE = 'moss',
+PARAMETERS = {
+    "project_id": "your-project-id",
+    "project_key": "moss_access_key_xxxxx",
+    "alpha": "0.8"
+};
+```
+
+The required parameters are:
+
+- `project_id`: Your Moss project ID, available from [portal.usemoss.dev](https://portal.usemoss.dev).
+- `project_key`: Your Moss project key (secret), available from [portal.usemoss.dev](https://portal.usemoss.dev).
+
+The optional parameters are:
+
+- `alpha`: Hybrid search weight between `0.0` (keyword-only) and `1.0` (semantic-only). Defaults to `0.8`.
+
+## Usage
+
+### Inserting documents
+
+Insert documents to create a new index. The index is built asynchronously — MindsDB waits until it is ready before returning (typically 5–30 seconds).
+
+```sql
+INSERT INTO moss_db.my_index (id, content, metadata)
+VALUES
+    ('doc-1', 'MindsDB unifies AI and data with SQL', '{"category": "mindsdb"}'),
+    ('doc-2', 'Moss delivers sub-10ms hybrid semantic search', '{"category": "moss"}'),
+    ('doc-3', 'RAG pipelines combine retrieval and generation', '{"category": "rag"}');
+```
+
+<Note>
+  The `id` column is optional. If omitted, Moss generates a unique ID
+  automatically using a hash of the document content.
+</Note>
+
+Inserting into an existing index upserts the documents:
+
+```sql
+INSERT INTO moss_db.my_index (id, content, metadata)
+VALUES ('doc-4', 'Vector databases store embeddings for similarity search', '{"category": "vectordb"}');
+```
+
+### Semantic search
+
+Query your index using natural language. Results are ranked by relevance and include a `distance` column (lower = better match):
+
+```sql
+SELECT id, content, distance
+FROM moss_db.my_index
+WHERE content = 'how do I build a RAG pipeline?'
+LIMIT 5;
+```
+
+### Fetch all documents
+
+```sql
+SELECT id, content, metadata
+FROM moss_db.my_index;
+```
+
+### Fetch by ID
+
+```sql
+SELECT id, content
+FROM moss_db.my_index
+WHERE id = 'doc-1';
+```
+
+### Semantic search with metadata filter
+
+Combine a semantic search query with a metadata filter:
+
+```sql
+SELECT id, content, distance
+FROM moss_db.my_index
+WHERE content = 'semantic search performance'
+  AND metadata.category = 'moss'
+LIMIT 3;
+```
+
+### Delete a document
+
+```sql
+DELETE FROM moss_db.my_index WHERE id = 'doc-1';
+```
+
+### Drop an index
+
+```sql
+DROP TABLE moss_db.my_index;
+```

mindsdb/integrations/handlers/moss_handler/README.md

@@ -0,0 +1,118 @@
+---
+title: Moss
+sidebarTitle: Moss
+---
+
+In this section, we present how to connect Moss to MindsDB.
+
+[Moss](https://moss.dev) is a semantic search runtime built for Conversational AI agents. It lets you index documents and run hybrid semantic/keyword queries in under 10ms — fast enough for real-time conversation, compared to 200–300ms with typical retrieval systems.
+
+## Prerequisites
+
+Before proceeding, ensure the following prerequisites are met:
+
+1. Install MindsDB locally via [Docker](/setup/self-hosted/docker) or [Docker Desktop](/setup/self-hosted/docker-desktop).
+2. To connect Moss to MindsDB, install the required dependencies following [this instruction](/setup/self-hosted/docker#install-dependencies).
+3. Create a Moss project and obtain your credentials from the [Moss Portal](https://portal.usemoss.dev).
+
+## Connection
+
+This handler is implemented using the `moss` Python library.
+
+To connect your Moss project to MindsDB, use the following statement:
+
+```sql
+CREATE DATABASE moss_datasource
+WITH ENGINE = 'moss',
+PARAMETERS = {
+    "project_id": "your-project-id",
+    "project_key": "moss_access_key_xxxxx",
+    "alpha": "0.8"
+};
+```
+
+The required parameters are:
+
+* `project_id`: Your Moss project ID, available in the [Moss Portal](https://portal.usemoss.dev).
+* `project_key`: Your Moss project access key.
+
+The optional parameters are:
+
+* `alpha`: Controls the blend between semantic and keyword search. Range is `0.0` to `1.0`, where `0.0` is pure keyword (BM25), `1.0` is pure semantic, and `0.8` is the default.
+
+## Usage
+
+Once connected, you can insert documents into a Moss index. The index is created automatically on the first insert.
+
+```sql
+INSERT INTO moss_datasource.my_index (id, content, metadata)
+VALUES
+    ('doc-1', 'MindsDB unifies AI and data pipelines', '{"category": "product"}'),
+    ('doc-2', 'Connect MindsDB to PostgreSQL, MySQL, and more', '{"category": "integrations"}'),
+    ('doc-3', 'Create AI models using the CREATE MODEL syntax', '{"category": "docs"}');
+```
+
+<Note>
+The `INSERT` statement blocks until Moss finishes building the index, which typically takes 5–30 seconds depending on document count.
+</Note>
+
+To run a semantic search query:
+
+```sql
+SELECT id, content, distance
+FROM moss_datasource.my_index
+WHERE content = 'how do I create an AI model?'
+LIMIT 5;
+```
+
+The `distance` column is `1 - score`, so lower values indicate a closer match.
+
+To fetch all documents without a search query:
+
+```sql
+SELECT id, content, metadata
+FROM moss_datasource.my_index;
+```
+
+To fetch specific documents by ID:
+
+```sql
+SELECT id, content
+FROM moss_datasource.my_index
+WHERE id = 'doc-1';
+```
+
+To filter results by metadata alongside a semantic search:
+
+```sql
+SELECT id, content, distance
+FROM moss_datasource.my_index
+WHERE content = 'connecting to databases'
+  AND metadata.category = 'integrations'
+LIMIT 3;
+```
+
+To delete documents from an index:
+
+```sql
+DELETE FROM moss_datasource.my_index
+WHERE id = 'doc-1';
+```
+
+To drop an index entirely:
+
+```sql
+DROP TABLE moss_datasource.my_index;
+```
+
+## Using Moss in a RAG Pipeline
+
+You can combine Moss with a MindsDB model to build a retrieval-augmented generation (RAG) pipeline entirely in SQL:
+
+```sql
+SELECT r.content, m.answer
+FROM moss_datasource.my_index AS r
+JOIN mindsdb.my_llm AS m
+WHERE r.content = 'what is the refund policy?'
+LIMIT 1;
+```

mindsdb/integrations/handlers/moss_handler/__about__.py

@@ -0,0 +1,9 @@
+__title__ = "MindsDB Moss handler"
+__package_name__ = "mindsdb_moss_handler"
+__version__ = "0.0.1"
+__description__ = "MindsDB handler for Moss semantic search"
+__author__ = "Keshav Arora"
+__github__ = "https://github.com/mindsdb/mindsdb"
+__pypi__ = "https://pypi.org/project/mindsdb/"
+__license__ = "MIT"
+__copyright__ = "Copyright 2024 - mindsdb"

mindsdb/integrations/handlers/moss_handler/__init__.py

@@ -0,0 +1,33 @@
+from mindsdb.integrations.libs.const import HANDLER_SUPPORT_LEVEL, HANDLER_TYPE
+
+from .__about__ import __description__ as description
+from .__about__ import __version__ as version
+from .connection_args import connection_args, connection_args_example
+
+try:
+    from .moss_handler import MossHandler as Handler
+
+    import_error = None
+except Exception as e:
+    Handler = None
+    import_error = e
+
+title = "Moss"
+name = "moss"
+type = HANDLER_TYPE.DATA
+support_level = HANDLER_SUPPORT_LEVEL.COMMUNITY
+icon_path = "icon.png"
+
+__all__ = [
+    "Handler",
+    "version",
+    "name",
+    "type",
+    "title",
+    "description",
+    "support_level",
+    "connection_args",
+    "connection_args_example",
+    "import_error",
+    "icon_path",
+]

mindsdb/integrations/handlers/moss_handler/connection_args.py

@@ -0,0 +1,31 @@
+from collections import OrderedDict
+
+from mindsdb.integrations.libs.const import HANDLER_CONNECTION_ARG_TYPE as ARG_TYPE
+
+connection_args = OrderedDict(
+    project_id={
+        "type": ARG_TYPE.STR,
+        "description": "Moss project ID from the Moss Portal (portal.usemoss.dev)",
+        "required": True,
+    },
+    project_key={
+        "type": ARG_TYPE.PWD,
+        "description": "Moss project key from the Moss Portal",
+        "required": True,
+        "secret": True,
+    },
+    alpha={
+        "type": ARG_TYPE.STR,
+        "description": (
+            "Hybrid search weight between semantic and keyword search. "
+            "0.0 = pure keyword (BM25), 1.0 = pure semantic, 0.8 = default"
+        ),
+        "required": False,
+    },
+)
+
+connection_args_example = OrderedDict(
+    project_id="your-project-id",
+    project_key="moss_access_key_xxxxx",
+    alpha="0.8",
+)