Quoted search terms ignore exact phrase matching when combined with other quoted phrases

Description

When performing searches with quoted phrases, Typesense appears to ignore the exact matching requirement when multiple quoted strings are present in the query.

Steps to reproduce

1. Run Typesense locally

docker run --rm -p 8108:8108 typesense/typesense:27.1 --api-key=abcdef --data-dir=/home

2. Create collection

curl -X POST 'http://localhost:8108/collections' \
-H 'X-Typesense-Api-Key: abcdef' \
-H 'Content-Type: application/json' \
-d '{
  "name": "titles",
  "fields": [
    {"name": "title", "type": "string", "index": true}
  ]
}'

3. Insert test data

curl -X POST 'http://localhost:8108/collections/titles/documents/import' \
-H 'X-Typesense-Api-Key: abcdef' \
-H 'Content-Type: application/jsonl' \
-d '{"title": "full stack developer"}
    {"title": "full stack javascript developer"}'

4. Searching "full stack" foobar properly returns 0 results:

  curl --request GET \
  --url 'http://localhost:8108/collections/titles/documents/search?q=%22full%20%20stack%22%20foobar&query_by=title' \
  --header 'X-TYPESENSE-API-KEY: abcdef'

Output:

{
  "facet_counts": [],
  "found": 0,
  "hits": [],
  "out_of": 2,
  "page": 1,
  "request_params": {
    "collection_name": "titles",
    "first_q": "\"full  stack\" foobar",
    "per_page": 10,
    "q": "\"full  stack\" foobar"
  },
  "search_cutoff": false,
  "search_time_ms": 0
}

5. Adding quotes around "foobar" should still return 0 results but returns 2 results:

curl --request GET \
  --url 'http://localhost:8108/collections/titles/documents/search?q=%22full%20%20stack%22%20%22foobar%22&query_by=title' \
  --header 'X-TYPESENSE-API-KEY: abcdef'

Output:

{
  "facet_counts": [],
  "found": 2,
  "hits": [
    {
      "document": { "id": "1", "title": "full stack javascript developer" },
      "highlight": {
        "title": {
          "matched_tokens": ["full", "stack"],
          "snippet": "<mark>full</mark> <mark>stack</mark> javascript developer"
        }
      },
      "highlights": [
        {
          "field": "title",
          "matched_tokens": ["full", "stack"],
          "snippet": "<mark>full</mark> <mark>stack</mark> javascript developer"
        }
      ],
      "text_match": 100015,
      "text_match_info": {
        "best_field_score": "48",
        "best_field_weight": 213,
        "fields_matched": 7,
        "num_tokens_dropped": 1,
        "score": "100015",
        "tokens_matched": 0,
        "typo_prefix_score": 255
      }
    },
    {
      "document": { "id": "0", "title": "full stack developer" },
      "highlight": {
        "title": {
          "matched_tokens": ["full", "stack"],
          "snippet": "<mark>full</mark> <mark>stack</mark> developer"
        }
      },
      "highlights": [
        {
          "field": "title",
          "matched_tokens": ["full", "stack"],
          "snippet": "<mark>full</mark> <mark>stack</mark> developer"
        }
      ],
      "text_match": 100015,
      "text_match_info": {
        "best_field_score": "48",
        "best_field_weight": 213,
        "fields_matched": 7,
        "num_tokens_dropped": 1,
        "score": "100015",
        "tokens_matched": 0,
        "typo_prefix_score": 255
      }
    }
  ],
  "out_of": 2,
  "page": 1,
  "request_params": {
    "collection_name": "titles",
    "first_q": "\"full  stack\" \"foobar\"",
    "per_page": 10,
    "q": "\"full  stack\" \"foobar\""
  },
  "search_cutoff": false,
  "search_time_ms": 0
}

6. Two sets of quoted strings, "full stack" "javascript developer", properly returns only 1 result:

curl --request GET \
  --url 'http://localhost:8108/collections/titles/documents/search?q=%22full%20stack%22%20%22javascript%20developer%22&query_by=title' \
  --header 'X-TYPESENSE-API-KEY: abcdef'

Output:

{
  "facet_counts": [],
  "found": 1,
  "hits": [
    {
      "document": { "id": "1", "title": "full stack javascript developer" },
      "highlight": {
        "title": {
          "matched_tokens": ["full", "stack", "javascript", "developer"],
          "snippet": "<mark>full</mark> <mark>stack</mark> <mark>javascript</mark> <mark>developer</mark>"
        }
      },
      "highlights": [
        {
          "field": "title",
          "matched_tokens": ["full", "stack", "javascript", "developer"],
          "snippet": "<mark>full</mark> <mark>stack</mark> <mark>javascript</mark> <mark>developer</mark>"
        }
      ],
      "text_match": 100015,
      "text_match_info": {
        "best_field_score": "48",
        "best_field_weight": 213,
        "fields_matched": 7,
        "num_tokens_dropped": 1,
        "score": "100015",
        "tokens_matched": 0,
        "typo_prefix_score": 255
      }
    }
  ],
  "out_of": 2,
  "page": 1,
  "request_params": {
    "collection_name": "titles",
    "first_q": "\"full stack\" \"javascript developer\"",
    "per_page": 10,
    "q": "\"full stack\" \"javascript developer\""
  },
  "search_cutoff": false,
  "search_time_ms": 0
}

7. Adding a word not in the data set in the second quoted string, "full stack" "javascript foobar developer", wrongly returns 2 results:

curl --request GET \
  --url 'http://localhost:8108/collections/titles/documents/search?q=%22full%20stack%22%20%22javascript%20foobar%20developer%22&query_by=title' \
  --header 'X-TYPESENSE-API-KEY: abcdef'

Output:

{
  "facet_counts": [],
  "found": 2,
  "hits": [
    {
      "document": { "id": "1", "title": "full stack javascript developer" },
      "highlight": {
        "title": {
          "matched_tokens": ["full", "stack", "javascript", "developer"],
          "snippet": "<mark>full</mark> <mark>stack</mark> <mark>javascript</mark> <mark>developer</mark>"
        }
      },
      "highlights": [
        {
          "field": "title",
          "matched_tokens": ["full", "stack", "javascript", "developer"],
          "snippet": "<mark>full</mark> <mark>stack</mark> <mark>javascript</mark> <mark>developer</mark>"
        }
      ],
      "text_match": 100015,
      "text_match_info": {
        "best_field_score": "48",
        "best_field_weight": 213,
        "fields_matched": 7,
        "num_tokens_dropped": 1,
        "score": "100015",
        "tokens_matched": 0,
        "typo_prefix_score": 255
      }
    },
    {
      "document": { "id": "0", "title": "full stack developer" },
      "highlight": {
        "title": {
          "matched_tokens": ["full", "stack", "developer"],
          "snippet": "<mark>full</mark> <mark>stack</mark> <mark>developer</mark>"
        }
      },
      "highlights": [
        {
          "field": "title",
          "matched_tokens": ["full", "stack", "developer"],
          "snippet": "<mark>full</mark> <mark>stack</mark> <mark>developer</mark>"
        }
      ],
      "text_match": 100015,
      "text_match_info": {
        "best_field_score": "48",
        "best_field_weight": 213,
        "fields_matched": 7,
        "num_tokens_dropped": 1,
        "score": "100015",
        "tokens_matched": 0,
        "typo_prefix_score": 255
      }
    }
  ],
  "out_of": 2,
  "page": 1,
  "request_params": {
    "collection_name": "titles",
    "first_q": "\"full stack\" \"javascript foobar developer\"",
    "per_page": 10,
    "q": "\"full stack\" \"javascript foobar developer\""
  },
  "search_cutoff": false,
  "search_time_ms": 1
}

Expected Behavior

1. Quoted strings should be matched exactly as whole phrases
2. Multiple quoted strings in a query should all need to match for a document to be returned

Actual Behavior

Adding a search term not found in the dataset into a second quoted string in the query will return results as if the second quoted string was not part of the query.

Metadata

Typesense Version: 27.1

OS: Docker

Reproduce the issue with the Docker Typesense 27.1 setup and the collection/search requests shown in the report. Trace how the documents search endpoint handles multiple quoted phrases, then verify that every quoted phrase is matched exactly and that queries containing an absent term return no results.