@techwavedev/agi-agent-kit 1.1.5 → 1.2.1

package/templates/skills/ec/opensearch/references/ml_neural_search.md

@@ -0,0 +1,576 @@
+# ML/AI & Neural Search Reference
+
+Comprehensive guide for machine learning and semantic search in OpenSearch.
+
+## ML Commons Plugin
+
+### Enable ML Commons
+
+```yaml
+# opensearch.yml
+plugins.ml_commons.only_run_on_ml_node: false
+plugins.ml_commons.model_access_control_enabled: true
+plugins.ml_commons.native_memory_threshold: 90
+plugins.ml_commons.allow_registering_model_via_url: true
+```
+
+### Cluster Settings
+
+```json
+PUT /_cluster/settings
+{
+  "persistent": {
+    "plugins.ml_commons.only_run_on_ml_node": false,
+    "plugins.ml_commons.model_access_control_enabled": true,
+    "plugins.ml_commons.native_memory_threshold": 90
+  }
+}
+```
+
+---
+
+## Model Management
+
+### List Models
+
+```json
+GET /_plugins/_ml/models/_search
+{
+  "query": { "match_all": {} },
+  "size": 100
+}
+```
+
+### Register Model (Local)
+
+```json
+POST /_plugins/_ml/models/_register
+{
+  "name": "sentence-transformers/all-MiniLM-L6-v2",
+  "version": "1.0.1",
+  "model_format": "TORCH_SCRIPT",
+  "model_config": {
+    "model_type": "bert",
+    "embedding_dimension": 384,
+    "framework_type": "sentence_transformers"
+  }
+}
+```
+
+### Register Model (Remote/Connector)
+
+```json
+POST /_plugins/_ml/connectors/_create
+{
+  "name": "Amazon Bedrock Titan Embeddings",
+  "description": "Connector for Titan embedding model",
+  "version": "1.0",
+  "protocol": "aws_sigv4",
+  "credential": {
+    "access_key": "${aws.access_key}",
+    "secret_key": "${aws.secret_key}"
+  },
+  "parameters": {
+    "region": "us-east-1",
+    "service_name": "bedrock"
+  },
+  "actions": [
+    {
+      "action_type": "predict",
+      "method": "POST",
+      "url": "https://bedrock-runtime.us-east-1.amazonaws.com/model/amazon.titan-embed-text-v1/invoke",
+      "headers": {
+        "Content-Type": "application/json"
+      },
+      "request_body": "{ \"inputText\": \"${parameters.inputText}\" }",
+      "pre_process_function": "connector.pre_process.bedrock.embedding",
+      "post_process_function": "connector.post_process.bedrock.embedding"
+    }
+  ]
+}
+```
+
+### Deploy Model
+
+```json
+POST /_plugins/_ml/models/<model_id>/_deploy
+```
+
+### Undeploy Model
+
+```json
+POST /_plugins/_ml/models/<model_id>/_undeploy
+```
+
+### Delete Model
+
+```json
+DELETE /_plugins/_ml/models/<model_id>
+```
+
+---
+
+## Pre-trained Models
+
+### Recommended Models
+
+| Model                                   | Dimensions | Use Case               |
+| --------------------------------------- | ---------- | ---------------------- |
+| `all-MiniLM-L6-v2`                      | 384        | Fast, general purpose  |
+| `all-mpnet-base-v2`                     | 768        | Higher quality, slower |
+| `paraphrase-multilingual-MiniLM-L12-v2` | 384        | Multilingual support   |
+| `msmarco-distilbert-base-tas-b`         | 768        | Passage retrieval      |
+
+### Register Pre-trained Model
+
+```json
+POST /_plugins/_ml/models/_register
+{
+  "name": "huggingface/sentence-transformers/all-MiniLM-L6-v2",
+  "version": "1.0.1",
+  "model_group_id": "<model_group_id>",
+  "model_format": "TORCH_SCRIPT"
+}
+```
+
+---
+
+## Neural Search Pipelines
+
+### Create Ingest Pipeline
+
+```json
+PUT /_ingest/pipeline/neural-search-pipeline
+{
+  "description": "Pipeline for neural search embeddings",
+  "processors": [
+    {
+      "text_embedding": {
+        "model_id": "<model_id>",
+        "field_map": {
+          "title": "title_embedding",
+          "description": "description_embedding"
+        }
+      }
+    }
+  ]
+}
+```
+
+### Multi-field Pipeline
+
+```json
+PUT /_ingest/pipeline/multi-field-embedding
+{
+  "description": "Generate embeddings for multiple fields",
+  "processors": [
+    {
+      "text_embedding": {
+        "model_id": "<model_id>",
+        "field_map": {
+          "title": "title_embedding"
+        }
+      }
+    },
+    {
+      "text_embedding": {
+        "model_id": "<model_id>",
+        "field_map": {
+          "content": "content_embedding"
+        }
+      }
+    }
+  ]
+}
+```
+
+### Search Pipeline (Query-time Embedding)
+
+```json
+PUT /_search/pipeline/neural-search-pipeline
+{
+  "request_processors": [
+    {
+      "neural_query_enricher": {
+        "neural_field_default_id": {
+          "description_embedding": "<model_id>"
+        }
+      }
+    }
+  ]
+}
+```
+
+---
+
+## KNN Index Configuration
+
+### Create KNN Index
+
+```json
+PUT /semantic-products
+{
+  "settings": {
+    "index": {
+      "knn": true,
+      "knn.algo_param.ef_search": 100,
+      "default_pipeline": "neural-search-pipeline"
+    }
+  },
+  "mappings": {
+    "properties": {
+      "title": { "type": "text" },
+      "description": { "type": "text" },
+      "title_embedding": {
+        "type": "knn_vector",
+        "dimension": 384,
+        "method": {
+          "name": "hnsw",
+          "space_type": "l2",
+          "engine": "lucene",
+          "parameters": {
+            "ef_construction": 128,
+            "m": 16
+          }
+        }
+      },
+      "description_embedding": {
+        "type": "knn_vector",
+        "dimension": 384,
+        "method": {
+          "name": "hnsw",
+          "space_type": "cosinesimil",
+          "engine": "nmslib",
+          "parameters": {
+            "ef_construction": 256,
+            "m": 24
+          }
+        }
+      }
+    }
+  }
+}
+```
+
+### KNN Engines
+
+| Engine   | Best For          | Space Types                   |
+| -------- | ----------------- | ----------------------------- |
+| `lucene` | General use, HNSW | l2, cosinesimil               |
+| `nmslib` | High recall       | l2, cosinesimil, innerproduct |
+| `faiss`  | Large scale       | l2, innerproduct              |
+
+### Space Types
+
+| Space Type     | Description        |
+| -------------- | ------------------ |
+| `l2`           | Euclidean distance |
+| `cosinesimil`  | Cosine similarity  |
+| `innerproduct` | Dot product        |
+
+---
+
+## Neural Search Queries
+
+### Basic Neural Query
+
+```json
+POST /semantic-products/_search
+{
+  "query": {
+    "neural": {
+      "description_embedding": {
+        "query_text": "comfortable running shoes for marathon",
+        "model_id": "<model_id>",
+        "k": 10
+      }
+    }
+  }
+}
+```
+
+### Neural with Filter
+
+```json
+POST /semantic-products/_search
+{
+  "query": {
+    "neural": {
+      "description_embedding": {
+        "query_text": "durable hiking boots",
+        "model_id": "<model_id>",
+        "k": 20,
+        "filter": {
+          "bool": {
+            "must": [
+              { "term": { "category": "footwear" } },
+              { "range": { "price": { "lte": 200 } } }
+            ]
+          }
+        }
+      }
+    }
+  }
+}
+```
+
+### KNN Query (Pre-computed Vector)
+
+```json
+POST /semantic-products/_search
+{
+  "query": {
+    "knn": {
+      "description_embedding": {
+        "vector": [0.1, 0.2, 0.3, ...],
+        "k": 10
+      }
+    }
+  }
+}
+```
+
+---
+
+## Hybrid Search
+
+### Neural + Keyword (Simple)
+
+```json
+POST /semantic-products/_search
+{
+  "query": {
+    "hybrid": {
+      "queries": [
+        {
+          "match": {
+            "title": {
+              "query": "running shoes",
+              "boost": 0.3
+            }
+          }
+        },
+        {
+          "neural": {
+            "description_embedding": {
+              "query_text": "running shoes",
+              "model_id": "<model_id>",
+              "k": 50
+            }
+          }
+        }
+      ]
+    }
+  }
+}
+```
+
+### Hybrid with Normalization
+
+```json
+PUT /_search/pipeline/hybrid-pipeline
+{
+  "description": "Hybrid search with score normalization",
+  "phase_results_processors": [
+    {
+      "normalization-processor": {
+        "normalization": {
+          "technique": "min_max"
+        },
+        "combination": {
+          "technique": "arithmetic_mean",
+          "parameters": {
+            "weights": [0.3, 0.7]
+          }
+        }
+      }
+    }
+  ]
+}
+```
+
+```json
+POST /semantic-products/_search?search_pipeline=hybrid-pipeline
+{
+  "query": {
+    "hybrid": {
+      "queries": [
+        { "match": { "title": "running shoes" } },
+        {
+          "neural": {
+            "description_embedding": {
+              "query_text": "running shoes",
+              "model_id": "<model_id>",
+              "k": 50
+            }
+          }
+        }
+      ]
+    }
+  }
+}
+```
+
+### Normalization Techniques
+
+| Technique | Description                  |
+| --------- | ---------------------------- |
+| `min_max` | Scale scores to [0, 1] range |
+| `l2`      | L2 normalization             |
+
+### Combination Techniques
+
+| Technique         | Description                 |
+| ----------------- | --------------------------- |
+| `arithmetic_mean` | Simple average with weights |
+| `geometric_mean`  | Geometric average           |
+| `harmonic_mean`   | Harmonic average            |
+
+---
+
+## Semantic Highlighting (Experimental)
+
+```json
+POST /semantic-products/_search
+{
+  "query": {
+    "neural": {
+      "description_embedding": {
+        "query_text": "comfortable shoes",
+        "model_id": "<model_id>",
+        "k": 10
+      }
+    }
+  },
+  "highlight": {
+    "fields": {
+      "description": {}
+    }
+  }
+}
+```
+
+---
+
+## Reranking
+
+### Cross-Encoder Reranking
+
+```json
+PUT /_search/pipeline/rerank-pipeline
+{
+  "response_processors": [
+    {
+      "rerank": {
+        "ml_opensearch": {
+          "model_id": "<cross_encoder_model_id>"
+        },
+        "context": {
+          "document_fields": ["title", "description"]
+        }
+      }
+    }
+  ]
+}
+```
+
+---
+
+## Performance Tuning
+
+### Index Settings
+
+```json
+PUT /semantic-index/_settings
+{
+  "index": {
+    "knn.algo_param.ef_search": 100,
+    "knn.space_type": "cosinesimil"
+  }
+}
+```
+
+### Reduce Memory Usage
+
+```json
+PUT /semantic-index
+{
+  "mappings": {
+    "properties": {
+      "embedding": {
+        "type": "knn_vector",
+        "dimension": 384,
+        "data_type": "float16"  // Use 16-bit floats
+      }
+    }
+  }
+}
+```
+
+### Batch Indexing
+
+```json
+POST /_bulk
+{"index":{"_index":"semantic-products","_id":"1","pipeline":"neural-search-pipeline"}}
+{"title":"Running Shoes","description":"Lightweight and comfortable"}
+{"index":{"_index":"semantic-products","_id":"2","pipeline":"neural-search-pipeline"}}
+{"title":"Hiking Boots","description":"Durable and waterproof"}
+```
+
+---
+
+## Model Monitoring
+
+### Check Model Status
+
+```json
+GET /_plugins/_ml/models/<model_id>
+```
+
+### Model Stats
+
+```json
+GET /_plugins/_ml/models/<model_id>/_stats
+```
+
+### Profile Inference
+
+```json
+POST /_plugins/_ml/models/<model_id>/_predict
+{
+  "parameters": {
+    "texts": ["Sample text to embed"]
+  }
+}
+```
+
+---
+
+## Troubleshooting
+
+### Model Not Found
+
+```bash
+# Check if model is registered
+GET /_plugins/_ml/models/_search
+
+# Check if model is deployed
+GET /_plugins/_ml/models/<model_id>/_status
+```
+
+### Out of Memory
+
+```yaml
+# Increase native memory threshold
+plugins.ml_commons.native_memory_threshold: 95
+
+# Or use dedicated ML nodes
+plugins.ml_commons.only_run_on_ml_node: true
+```
+
+### Slow Inference
+
+1. Check model deployment status
+2. Use smaller model (e.g., MiniLM instead of mpnet)
+3. Batch requests when possible
+4. Use dedicated ML nodes