Add Amazon Redshift connector (#2)

[jadstrike]

What this adds

Amazon Redshift connector, as requested in the README's next-targets list (issue #2).

• src/connectors/redshift.py — follows the standard connect / list_tables / fetch_table pattern used by the other connectors. Redshift is wire-compatible with PostgreSQL, so it connects via psycopg2 — zero new dependencies.
• Amazon Redshift option in the Database Connectors section of the Streamlit UI (host, port — default 5439, database, username, password, table).

Tests

Added TestRedshiftConnector with three tests (list_tables, fetch_table, row limit). They monkeypatch psycopg2.connect with a stubbed connection, so the suite runs anywhere — no live cluster or AWS credentials required.

pytest tests/ -v

All existing tests still pass.

README

• Redshift usage snippet under Database Connectors
• Connector status table: Amazon Redshift → Stable (auth: user/pass)
• Removed Redshift from the next-targets table and checked the roadmap box

Note

This branch is independent of #2's DuckDB work (both branch off main). If both are merged, the only overlap is the connector dropdown list in app.py — a trivial one-line conflict.

[harshitboots]

def test_fetch_table_validates_table_name(self, fake_connection): from src.connectors.redshift import fetch_table with pytest.raises(ValueError, match="not found"): fetch_table("host", 5439, "dev", "awsuser", "password", "nonexistent_table")

[jadstrike]

Added this test in 087c44c — it passes along with the rest of the suite.

[harshitboots]

Can do this which will give more liberty to user while connecting to any tables

[harshitboots]

` think we can use following code as its Doesn't work if user wants to query views or temporary tables (though they'd need to be in the whitelist query) as well now

actually it will resolve postgres and mysql issue also.

`def fetch_table(host: str, port: int, database: str, user: str, password: str, table: str, limit: int = 1000) -> pd.DataFrame:
"""Fetch a table from Redshift with SQL injection protection."""
conn = connect(host, port, database, user, password)
cursor = conn.cursor()

try:
    # Validate table exists to prevent SQL injection
    valid_tables = list_tables(host, port, database, user, password)
    if table not in valid_tables:
        raise ValueError(f"Table '{table}' not found in database. Available tables: {', '.join(valid_tables)}")
    
    cursor.execute(f"SELECT * FROM {table} LIMIT {limit}")
    columns = [desc[0] for desc in cursor.description]
    rows = cursor.fetchall()
    return pd.DataFrame(rows, columns=columns)
finally:
    cursor.close()
    conn.close()`

[jadstrike]

Applied in 087c44c. fetch_table now validates the table name against list_tables() before querying and raises a ValueError listing the available tables if it's not found. Cursor/connection cleanup is now handled in a finally block. The same pattern would also apply to the Postgres and MySQL connectors — happy to address those in a follow-up PR.