feat(model): Add new LLMClient and new build tools (#967)

dbgpt/datasource/rdbms/base.py

@@ -1,7 +1,6 @@
 from __future__ import annotations
 import sqlparse
 import regex as re
-import pandas as pd
 from urllib.parse import quote
 from urllib.parse import quote_plus as urlquote
 from typing import Any, Iterable, List, Optional, Dict
@@ -383,6 +382,10 @@ class RDBMSDatabase(BaseConnect):
                 return self.get_simple_fields(table_name)
 
     def run_to_df(self, command: str, fetch: str = "all"):
+        import pandas as pd
+
+        # Pandas has too much dependence and the import time is too long
+        # TODO: Remove the dependency on pandas
         result_lst = self.run(command, fetch)
         colunms = result_lst[0]
         values = result_lst[1:]

dbgpt/datasource/rdbms/conn_clickhouse.py

@@ -1,13 +1,8 @@
 import re
 import sqlparse
-import clickhouse_connect
 from typing import List, Optional, Any, Iterable, Dict
 from sqlalchemy import text
-from urllib.parse import quote
-from sqlalchemy.schema import CreateTable
-from urllib.parse import quote_plus as urlquote
 from dbgpt.datasource.rdbms.base import RDBMSDatabase
-from clickhouse_connect.driver import httputil
 from dbgpt.storage.schema import DBType
 from sqlalchemy import (
     MetaData,
@@ -56,6 +51,11 @@ class ClickhouseConnect(RDBMSDatabase):
         engine_args: Optional[dict] = None,
         **kwargs: Any,
     ) -> RDBMSDatabase:
+        import clickhouse_connect
+        from clickhouse_connect.driver import httputil
+
+        # Lazy import
+
         big_pool_mgr = httputil.get_pool_manager(maxsize=16, num_pools=12)
         client = clickhouse_connect.get_client(
             host=host,

dbgpt/datasource/rdbms/conn_sqlite.py

@@ -37,7 +37,14 @@ class SQLiteConnect(RDBMSDatabase):
         """Get table indexes about specified table."""
         cursor = self.session.execute(text(f"PRAGMA index_list({table_name})"))
         indexes = cursor.fetchall()
-        return [(index[1], index[3]) for index in indexes]
+        result = []
+        for idx in indexes:
+            index_name = idx[1]
+            cursor = self.session.execute(text(f"PRAGMA index_info({index_name})"))
+            index_infos = cursor.fetchall()
+            column_names = [index_info[2] for index_info in index_infos]
+            result.append({"name": index_name, "column_names": column_names})
+        return result
 
     def get_show_create_table(self, table_name):
         """Get table show create table about specified table."""

dbgpt/datasource/rdbms/tests/test_conn_sqlite.py

@@ -47,7 +47,8 @@ def test_run_no_throw(db):
 def test_get_indexes(db):
     db.run("CREATE TABLE test (name TEXT);")
     db.run("CREATE INDEX idx_name ON test(name);")
-    assert db.get_indexes("test") == [("idx_name", "c")]
+    indexes = db.get_indexes("test")
+    assert indexes == [{"name": "idx_name", "column_names": ["name"]}]
 
 
 def test_get_indexes_empty(db):