elastic.txt

GET /

DELETE movies-v2
DELETE movies_semantic

# Get all movies
POST /movies/_search?size=100

# Full-text search
POST /movies/_search
{
  "query": {
    "query_string": {
      "query": "brooks"
    }
  }
}

# Boolean query - Science-Fiction movies with rating >= 8.5 after 2010
POST /movies/_search
{
  "query": {
    "bool": {
      "must": [
        {
          "match": {
            "genre": "Science-Fiction"
          }
        }
      ],
      "filter": [
        {
          "range": {
            "rating": {
              "gte": 8.5
            }
          }
        },
        {
          "range": {
            "year": {
              "gt": 2010
            }
          }
        }
      ]
    }
  }
}

# Aggregation - Average rating per country
POST /movies/_search
{
  "size": 0,
  "aggs": {
    "per_country": {
      "terms": {
        "field": "country.keyword"
      },
      "aggs": {
        "average_rating": {
          "avg": {
            "field": "rating"
          }
        }
      }
    }
  }
}

# Check mappings
GET /movies/_mappings

# Analyze text
POST /movies/_analyze
{
    "field": "actors",
    "text": "Albert Brooks"
}

POST /movies/_search
{
    "query": {
        "query_string": {
            "query": "brooks"
        }
    }
}

# Term search (exact match)
POST /movies/_search
{
  "query": {
    "term": {
      "country": {
        "value": "United States"
      }
    }
  }
}

POST /movies/_search

# ============================================
# SYNONYMS ANALYZER
# ============================================

PUT /movies-v2
{
  "settings": {
    "analysis": {
      "filter": {
        "country_synonyms": {
          "type": "synonym",
          "synonyms": [
            "USA, United States, US, America",
            "UK, United Kingdom, Britain, England",
            "South Korea, Korea"
          ]
        }
      },
      "analyzer": {
        "country_analyzer": {
          "type": "custom",
          "tokenizer": "keyword",
          "filter": [
            "lowercase",
            "country_synonyms"
          ]
        }
      }
    }
  },
  "mappings": {
    "properties": {
      "country": {
        "type": "text",
        "analyzer": "country_analyzer",
        "fields": {
          "keyword": {
            "type": "keyword"
          }
        }
      }
    }
  }
}


GET movies-v2/_mapping
# Reindex with synonyms
POST _reindex
{
    "source": { "index": "movies" },
    "dest": { "index": "movies-v2" }
}

# Test analyzer
POST /movies-v2/_analyze
{
    "field": "country",
    "text": "United States"
}

# Search with synonym
POST /movies-v2/_search
{
    "query": {
        "match": {
            "country": "America"
        }
    }
}


# ============================================
# SEMANTIC SEARCH SETUP
# ============================================


# Create embedding endpoint
PUT _inference/text_embedding/movies-embeddings
{
    "service": "elasticsearch",
    "service_settings": {
        "num_allocations": 1,
        "num_threads": 1,
        "model_id": ".multilingual-e5-small"
    }
}

# Create pipeline to copy synopsis
PUT _ingest/pipeline/copy_synopsis_to_semantic
{
  "processors": [
    {
      "set": {
        "field": "synopsis_semantic",
        "copy_from": "synopsis"
      }
    }
  ]
}

# Create semantic index
PUT /movies_semantic
{
  "mappings": {
    "properties": {
      "title": {
        "type": "text"
      },
      "synopsis": {
        "type": "text"
      },
      "synopsis_semantic": {
        "type": "semantic_text",
        "inference_id": "movies-embeddings"
      },
      "genre": {
        "type": "keyword"
      },
      "director": {
        "type": "text"
      },
      "year": {
        "type": "integer"
      },
      "rating": {
        "type": "float"
      },
      "country": {
        "type": "keyword"
      }
    }
  }
}
# Reindex with semantic field
POST _reindex
{
  "source": {
    "index": "movies"
  },
  "dest": {
    "index": "movies_semantic",
    "pipeline": "copy_synopsis_to_semantic"
  }
}


# ============================================
# SEMANTIC SEARCH QUERIES
# ============================================

# Check all movies
POST /movies_semantic/_search

# Semantic search - cars
POST /movies_semantic/_search
{
    "query": {
        "semantic": {
            "field": "synopsis_semantic",
            "query": "a movie with cars"
        }
    }
}

# Semantic search - social class conflict
POST /movies_semantic/_search
{
    "query": {
        "semantic": {
            "field": "synopsis_semantic",
            "query": "conflict between rich and poor, social inequality"
        }
    }
}

# Semantic search - questioning reality
POST /movies_semantic/_search
{
    "query": {
        "semantic": {
            "field": "synopsis_semantic",
            "query": "someone discovers their reality is an illusion"
        }
    }
}


# ============================================
# HYBRID SEARCH (RRF)
# ============================================



# Hybrid with filter
POST /movies_semantic/_search
{
  "retriever": {
    "rrf": {
      "retrievers": [
        {
          "standard": {
            "query": {
              "bool": {
                "must": {
                  "match": {
                    "synopsis": "dreams mind"
                  }
                },
                "filter": {
                  "range": {
                    "rating": {
                      "gte": 8
                    }
                  }
                }
              }
            }
          }
        },
        {
          "standard": {
            "query": {
              "bool": {
                "must": {
                  "semantic": {
                    "field": "synopsis_semantic",
                    "query": "exploring dreams and the subconscious"
                  }
                }
              }
            }
          }
        }
      ]
    }
  }
}


# ============================================
# RAG - RETRIEVAL AUGMENTED GENERATION
# ============================================
# Setup LLM endpoint (Gemini)
PUT _inference/completion/movies-chat
{
  "service": "googleaistudio",
  "service_settings": {
    "api_key": "YOUR_API_KEY",
    "model_id": "gemini-2.0-flash"
  }
}

GET  _inference


# Step 1: Ask LLM WITHOUT context (will fail or hallucinate)
POST _inference/completion/movies-chat
{
  "input": "How many Christopher Nolan movies do I have in my database? Which ones?"
}

# Step 2: RETRIEVAL - Search for Nolan movies
POST /movies_semantic/_search
{
  "size": 10,
  "query": {
    "match": {
      "director": "Christopher Nolan"
    }
  }
}

# Step 3: RAG - Give context to LLM
POST _inference/completion/movies-chat
{
  "input": "CONTEXT - Christopher Nolan movies in our database:\n\n1. Inception (2010) - Rating: 8.8/10\nSynopsis: Dom Cobb is a skilled thief who steals secrets from deep within the subconscious during the dream state.\n\n2. Interstellar (2014) - Rating: 8.7/10\nSynopsis: A group of explorers must travel through a wormhole to find a new home for humanity.\n\n3. The Dark Knight (2008) - Rating: 9.0/10\nSynopsis: Batman faces the Joker, a criminal mastermind who wants to plunge Gotham City into anarchy.\n\n4. The Prestige (2006) - Rating: 8.5/10\nSynopsis: Two rival magicians engage in a bitter battle for supremacy in Victorian London.\n\n5. Memento (2000) - Rating: 8.4/10\nSynopsis: A man with short-term memory loss uses notes and tattoos to hunt for his wife's killer.\n\n6. Dunkirk (2017) - Rating: 7.8/10\nSynopsis: Allied soldiers are evacuated during a fierce battle in World War II.\n\n7. Tenet (2020) - Rating: 7.3/10\nSynopsis: A secret agent manipulates time to prevent an attack from the future.\n\n8. Oppenheimer (2023) - Rating: 8.4/10\nSynopsis: The story of J. Robert Oppenheimer and the development of the atomic bomb.\n\n---\nQUESTION: How many Christopher Nolan movies do I have in my database? List them with their ratings and calculate the average rating."
}

POST _inference/completion/movies-chat
{
  "input": "CONTEXT - Christopher Nolan movies in our database:\n\n1. Inception (2010) - Rating: 8.8/10\nSynopsis: Dom Cobb is a skilled thief who steals secrets from deep within the subconscious during the dream state.\n\n2. Interstellar (2014) - Rating: 8.7/10\nSynopsis: A group of explorers must travel through a wormhole to find a new home for humanity.\n\n3. The Dark Knight (2008) - Rating: 9.0/10\nSynopsis: Batman faces the Joker, a criminal mastermind who wants to plunge Gotham City into anarchy.\n\n4. The Prestige (2006) - Rating: 8.5/10\nSynopsis: Two rival magicians engage in a bitter battle for supremacy in Victorian London.\n\n5. Memento (2000) - Rating: 8.4/10\nSynopsis: A man with short-term memory loss uses notes and tattoos to hunt for his wife's killer.\n\n6. Dunkirk (2017) - Rating: 7.8/10\nSynopsis: Allied soldiers are evacuated during a fierce battle in World War II.\n\n7. Tenet (2020) - Rating: 7.3/10\nSynopsis: A secret agent manipulates time to prevent an attack from the future.\n\n8. Oppenheimer (2023) - Rating: 8.4/10\nSynopsis: The story of J. Robert Oppenheimer and the development of the atomic bomb.\n\n---\nQUESTION: How many Christopher Nolan movies do I have in my database? List them with their ratings and calculate the average rating. And give me the name of french version of each movie "
}


POST _query?format=json
{
  "query": """
        FROM movies_semantic
        | WHERE match(director, "Christopher Nolan")
        | LIMIT 10
        | EVAL prompt = CONCAT(
            "Here are Christopher Nolan movies in our database:\n",
            "- ", title, " (", TO_STRING(year), ") - Rating: ", TO_STRING(rating), "\n",
            "Synopsis: ", synopsis, "\n\n",
            "Question: Summarize this movie in few sentences."
          )
        | COMPLETION answer = prompt WITH { "inference_id": "movies-chat" }
        | KEEP title, year, rating, answer
    """
}