docs: improved info about auto-embeddings in plain mode

Author: sanikolaev

.translation-cache/Creating_a_table/Local_tables/Plain_and_real-time_table_settings.md.json

@@ -471,25 +471,45 @@ Each vector attribute stores an array of floating-point numbers that represent d
 
 ##### Configuring KNN for vector attributes
 
-To enable KNN searches on float vector attributes, you must add a `knn` configuration that specifies the indexing parameters:
+To enable KNN searches on float vector attributes, you must add a `knn` configuration that specifies the indexing parameters. You can configure KNN in two ways:
+
+**1. Manual vector insertion** (you provide pre-computed vectors):
 
 ```ini
 rt_attr_float_vector = image_vector
 rt_attr_float_vector = text_vector
 knn = {"attrs":[{"name":"image_vector","type":"hnsw","dims":768,"hnsw_similarity":"COSINE","hnsw_m":16,"hnsw_ef_construction":200},{"name":"text_vector","type":"hnsw","dims":768,"hnsw_similarity":"COSINE","hnsw_m":16,"hnsw_ef_construction":200}]}
 ```
 
+**2. Auto embeddings** (Manticore generates vectors from text automatically):
+
+```ini
+rt_attr_float_vector = embedding_vector
+rt_field = title
+rt_field = description
+knn = {"attrs":[{"name":"embedding_vector","type":"hnsw","hnsw_similarity":"L2","hnsw_m":16,"hnsw_ef_construction":200,"model_name":"sentence-transformers/all-MiniLM-L6-v2","from":"title"}]}
+```
+
 **Required KNN parameters:**
 - `name`: The name of the vector attribute (must match the `rt_attr_float_vector` name)
 - `type`: Index type, currently only `"hnsw"` is supported
-- `dims`: Number of dimensions in the vectors (must match your embedding model's output)
+- `dims`: Number of dimensions in the vectors. **Required** for manual vector insertion, **must be omitted** when using `model_name` (the model determines dimensions automatically)
 - `hnsw_similarity`: Distance function - `"L2"`, `"IP"` (inner product), or `"COSINE"`
 
 **Optional KNN parameters:**
-- `hnsw_m`: Maximum connections in the graph
-- `hnsw_ef_construction`: Construction time/accuracy trade-off
+- `hnsw_m`: Maximum connections in the graph (default: 16)
+- `hnsw_ef_construction`: Construction time/accuracy trade-off (default: 200)
+
+**Auto-embeddings parameters** (when using `model_name`):
+- `model_name`: The embedding model to use (e.g., `"sentence-transformers/all-MiniLM-L6-v2"`, `"openai/text-embedding-ada-002"`). When specified, `dims` must be omitted as the model determines the dimensions automatically.
+- `from`: Comma-separated list of field names to use for embedding generation, or empty string `""` to use all text/string fields. This parameter is required when `model_name` is specified.
+- `api_key`: API key for API-based models (OpenAI, Voyage, Jina). Only required for API-based embedding services.
+- `cache_path`: Optional path for caching downloaded models (for sentence-transformers models).
+- `use_gpu`: Optional boolean to enable GPU acceleration if available.
+
+**Important:** You cannot specify both `dims` and `model_name` in the same configuration - they are mutually exclusive. Use `dims` for manual vector insertion, or `model_name` for auto-embeddings. Use `dims` for manual vector insertion, or `model_name` for auto-embeddings.
 
-For more details on KNN vector search, see the [KNN documentation](../../Searching/KNN.md).
+For more details on KNN vector search and auto-embeddings, see the [KNN documentation](../../Searching/KNN.md).
 
 #### rt_attr_bool
 

.translation-cache/Searching/KNN.md.json

@@ -39,7 +39,7 @@ Query OK, 0 rows affected (0.01 sec)
 ```
 
 <!-- intro -->
-##### Plain mode (using configuration file):
+##### Plain mode (using configuration file) - Manual vectors:
 
 <!-- request Config -->
 ```ini
@@ -51,6 +51,8 @@ table test_vec {
 }
 ```
 
+**Note:** For auto-embeddings in plain mode, see the example below, which shows how to use `model_name` and `from` parameters in the `knn` configuration.
+
 <!-- end -->
 
 <!-- example knn_insert -->
@@ -110,6 +112,51 @@ CREATE TABLE products_all (
 );
 ```
 
+<!-- intro -->
+##### Plain mode (using configuration file):
+
+<!-- request Config -->
+```ini
+table products {
+    type = rt
+    path = /path/to/products
+    rt_field = title
+    rt_field = description
+    rt_attr_float_vector = embedding_vector
+    knn = {"attrs":[{"name":"embedding_vector","type":"hnsw","hnsw_similarity":"L2","hnsw_m":16,"hnsw_ef_construction":200,"model_name":"sentence-transformers/all-MiniLM-L6-v2","from":"title"}]}
+}
+```
+
+Using OpenAI with API key in plain mode:
+```ini
+table products_openai {
+    type = rt
+    path = /path/to/products_openai
+    rt_field = title
+    rt_field = description
+    rt_attr_float_vector = embedding_vector
+    knn = {"attrs":[{"name":"embedding_vector","type":"hnsw","hnsw_similarity":"L2","hnsw_m":16,"hnsw_ef_construction":200,"model_name":"openai/text-embedding-ada-002","from":"title,description","api_key":"your-api-key-here"}]}
+}
+```
+
+Using all text fields (empty FROM):
+```ini
+table products_all {
+    type = rt
+    path = /path/to/products_all
+    rt_field = title
+    rt_field = description
+    rt_attr_float_vector = embedding_vector
+    knn = {"attrs":[{"name":"embedding_vector","type":"hnsw","hnsw_similarity":"L2","hnsw_m":16,"hnsw_ef_construction":200,"model_name":"sentence-transformers/all-MiniLM-L6-v2","from":""}]}
+}
+```
+
+**Important notes for plain mode:**
+- When using `model_name`, you **must not** specify `dims` - the model automatically determines the vector dimensions. The `dims` and `model_name` parameters are mutually exclusive.
+- When **not** using `model_name` (manual vector insertion), you **must** specify `dims` to indicate the vector dimensions.
+- The `from` parameter specifies which fields to use for embedding generation (comma-separated list, or empty string for all text/string fields). This parameter is required when using `model_name`.
+- For API-based models (OpenAI, Voyage, Jina), include the `api_key` parameter in the knn configuration
+
 <!-- end -->
 
 ##### Inserting data with auto embeddings