SQLite file viewer

About this tool

Inspect a small standalone SQLite 3 database in your browser without uploading its contents or changing the original file. Open a local .sqlite, .sqlite3 or .db file, review stored table, index, view and trigger definitions, and browse rows from supported ordinary tables. The interface generates quoted SELECT queries with LIMIT/OFFSET and NOT INDEXED rather than accepting arbitrary SQL. Actual storage-class labels keep integers, real numbers, text, binary bytes and null distinct. Export a selected inclusive row range as typed CSV or a JSON envelope. The pinned sql.js engine and WASM are bundled as same-origin assets and start only when you open a database. Data and query results are not persisted. For queries over CSV/JSON sources, use the separate local SQL workbench.

Executable examples from synthetic SQLite files

{
  "file": "example.sqlite",
  "table": "records",
  "format": "json",
  "start": 1,
  "end": 2,
  "safe": true,
  "excerptColumns": [
    "id",
    "note",
    "payload"
  ]
}

{
  "columnsExcerpt": [
    "id",
    "note",
    "payload"
  ],
  "rowsExcerpt": [
    [
      {
        "type": "integer",
        "value": "1"
      },
      {
        "type": "null",
        "value": null
      },
      {
        "type": "blob",
        "value": "00FF"
      }
    ],
    [
      {
        "type": "integer",
        "value": "2"
      },
      {
        "type": "text",
        "value": ""
      },
      {
        "type": "blob",
        "value": ""
      }
    ]
  ]
}

{
  "file": "utf16le.sqlite",
  "table": "words",
  "format": "json",
  "start": 1,
  "end": 3,
  "safe": true,
  "excerptColumns": [
    "value"
  ]
}

{
  "columnsExcerpt": [
    "value"
  ],
  "rowsExcerpt": [
    [
      {
        "type": "text",
        "value": "猫😀"
      }
    ],
    [
      {
        "type": "text",
        "value": "é"
      }
    ],
    [
      {
        "type": "text",
        "value": "a\u0000b"
      }
    ]
  ]
}

{
  "file": "example.sqlite",
  "table": "records",
  "format": "csv",
  "start": 3,
  "end": 3,
  "safe": true,
  "excerptColumns": [
    "code",
    "name",
    "note"
  ]
}

{
  "csvRecordsExcerpt": [
    [
      "code",
      "name",
      "note"
    ],
    [
      "T:003",
      "T:猫😀",
      "T:=SUM(1,2)"
    ]
  ]
}

{
  "file": "example.sqlite",
  "table": "empty_table",
  "format": "json",
  "start": 0,
  "end": 0,
  "safe": true,
  "excerptColumns": [
    "id",
    "value"
  ]
}

{
  "columnsExcerpt": [
    "id",
    "value"
  ],
  "rowsExcerpt": []
}

{
  "file": "schema-edge.sqlite",
  "table": "strict_values",
  "format": "json",
  "start": 1,
  "end": 3,
  "safe": true,
  "excerptColumns": [
    "value"
  ]
}

{
  "columnsExcerpt": [
    "value"
  ],
  "rowsExcerpt": [
    [
      {
        "type": "text",
        "value": "001"
      }
    ],
    [
      {
        "type": "real",
        "value": 1.5
      }
    ],
    [
      {
        "type": "blob",
        "value": "00FF"
      }
    ]
  ]
}

{
  "file": "schema-edge.sqlite",
  "table": "without rowid",
  "format": "json",
  "start": 1,
  "end": 2,
  "safe": true,
  "excerptColumns": [
    "a",
    "b"
  ]
}

{
  "columnsExcerpt": [
    "a",
    "b"
  ],
  "rowsExcerpt": [
    [
      {
        "type": "text",
        "value": "x"
      },
      {
        "type": "integer",
        "value": "1"
      }
    ],
    [
      {
        "type": "text",
        "value": "x"
      },
      {
        "type": "integer",
        "value": "2"
      }
    ]
  ]
}

{
  "file": "schema-edge.sqlite",
  "table": "generated"
}

{
  "error": "unsupported"
}

{
  "file": "virtual-fts5.sqlite",
  "table": "docs"
}

{
  "error": "unsupported"
}

{
  "file": "wal.sqlite"
}

{
  "error": "wal"
}

{
  "file": "truncated.sqlite"
}

{
  "error": "format"
}

Common SQLite inspection mistakes

• Opening a live application’s main file while ignoring its WAL or recovery journal, rather than exporting a consistent standalone copy.
• Assuming a readable schema or one successful page proves that every database page is intact.
• Interpreting a schema-only object as an empty table, or expecting a virtual-table extension to be loaded.
• Treating a declared column type or NOT NULL flag as a complete description of all actual storage classes and constraints.
• Reading 100,000+ as an exact count or assuming that row positions are sorted by a visible ID.
• Assuming an export contains only the first 12 displayed columns or the current preview page.
• Removing CSV type tags without decoding them, losing the null/text/blob distinction or reintroducing spreadsheet risks.
• Expecting an old result or pending download to survive replacing the file, cancellation, timeout or a language change.

Limits and notes

• —The input profile is one standalone, unencrypted SQLite 3 database of at most 16 MiB, with a consistent supported header, zero reserved page bytes, UTF-8 or UTF-16 LE/BE text encoding, at most 200 schema objects and 128 columns per ordinary table. Names are limited to 1,024 UTF-8 bytes, a stored SQL definition or default/type text to 16 KiB, and the complete schema summary to 512 KiB. WAL-mode main files, WAL/SHM sidecars, SQL dump text, remote URLs, encryption and extension-specific file formats are not accepted. Export a consistent copy from the source application instead of copying a live database alone.
• —Only ordinary supported tables are browsable. Views, triggers, internal SQLite objects and tables with hidden/generated columns are display-only; their expressions are not executed for browsing. Any virtual table disables all row access for the whole database, including ordinary-looking extension shadow tables. Table and index metadata remain subject to the supported profile. The interface generates its own SELECT statements with quoted identifiers, LIMIT/OFFSET and NOT INDEXED on an in-memory copy with query_only enabled. There is no custom SQL editor, WHERE filter, join, ORDER BY, schema modification, extension loading, repair or database-file export.
• —The table counter reads at most 100,001 rows and exposes only the first 100,000. A + marker is a lower-bound warning, not an exact full-table count. Preview pages contain 25 rows and the first 12 columns, with names shortened at 160 characters and cell values at 500. Row positions follow the immutable session’s SQLite scan order, not a portable sorted or insertion order. Each text or BLOB cell is limited to 256 KiB, while internal row materialization and each serialized export must fit 6 MiB. Successful exports contain the full values and every column in the chosen range, up to 1,000 rows. An operation fails on a limit rather than silently exporting a shortened value. Opening and sampling do not run a full integrity check; corruption in unread pages may remain undetected.
• —Every INTEGER is an exact decimal string, including small values. REAL remains a number except +Infinity, -Infinity and -0, which use explicit strings; this does not make SQLite arithmetic exact. TEXT retains its content, including embedded NUL, and BLOB uses uppercase hexadecimal. NULL, empty TEXT, empty BLOB and numeric-looking TEXT have separate type tags. JSON uses ordered columns and row arrays of {type,value} cells, plus table and range metadata. It is an inspection format, not a database backup. Some control characters may not be visually apparent in a preview; inspect the JSON escapes when exact text matters.
• —CSV uses UTF-8, comma separators and CRLF records. Every data value has a two-character tag: N: for NULL, I: for integer text, R: for real text, T: for text and B: for uppercase hexadecimal bytes. N:, T: and B: are distinct. Strip exactly one tag to obtain a non-null payload; literal text containing tags is preserved behind T:. Formula protection changes risky headers only, adding an apostrophe, and is on by default. Disabling it keeps all data tags but can expose header formulas; removing tags later can reintroduce formula and type-coercion risks. Each request has a 10-second browser-scheduled deadline. A 64 MiB SQLite heap limit is not a cap on JavaScript, WASM or total browser memory. Cancel, timeout, failure, changing the file or language, Reset and unmount invalidate work and release the session. No user file content is uploaded, stored or sent to a third-party CDN; first-party application assets still need to load, and downloads can contain sensitive values.

Frequently asked questions

Related tools