Update tools/sql_tool.py
Browse files- tools/sql_tool.py +55 -28
tools/sql_tool.py
CHANGED
|
@@ -5,59 +5,86 @@ from typing import Optional
|
|
| 5 |
from utils.config import AppConfig
|
| 6 |
from utils.tracing import Tracer
|
| 7 |
|
|
|
|
| 8 |
class SQLTool:
|
| 9 |
def __init__(self, cfg: AppConfig, tracer: Tracer):
|
| 10 |
self.cfg = cfg
|
| 11 |
self.tracer = tracer
|
| 12 |
self.backend = cfg.sql_backend # "bigquery" or "motherduck"
|
|
|
|
| 13 |
if self.backend == "bigquery":
|
| 14 |
from google.cloud import bigquery
|
| 15 |
from google.oauth2 import service_account
|
|
|
|
| 16 |
key_json = os.getenv("GCP_SERVICE_ACCOUNT_JSON")
|
| 17 |
if not key_json:
|
| 18 |
raise RuntimeError("Missing GCP_SERVICE_ACCOUNT_JSON secret")
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
self.client = bigquery.Client(credentials=creds, project=cfg.gcp_project)
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
|
|
|
| 42 |
else:
|
| 43 |
raise RuntimeError("Unknown SQL backend")
|
| 44 |
|
| 45 |
def _nl_to_sql(self, message: str) -> str:
|
| 46 |
-
|
| 47 |
-
|
|
|
|
|
|
|
|
|
|
| 48 |
m = message.lower()
|
|
|
|
|
|
|
| 49 |
if "avg" in m and " by " in m:
|
| 50 |
-
return
|
| 51 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
if re.match(r"^\s*select ", m):
|
| 53 |
return message
|
| 54 |
-
|
|
|
|
|
|
|
| 55 |
|
| 56 |
def run(self, message: str) -> pd.DataFrame:
|
| 57 |
sql = self._nl_to_sql(message)
|
| 58 |
self.tracer.trace_event("sql_query", {"sql": sql, "backend": self.backend})
|
|
|
|
| 59 |
if self.backend == "bigquery":
|
| 60 |
df = self.client.query(sql).to_dataframe()
|
| 61 |
else:
|
|
|
|
| 62 |
df = self.client.execute(sql).fetch_df()
|
| 63 |
-
|
|
|
|
|
|
| 5 |
from utils.config import AppConfig
|
| 6 |
from utils.tracing import Tracer
|
| 7 |
|
| 8 |
+
|
| 9 |
class SQLTool:
|
| 10 |
def __init__(self, cfg: AppConfig, tracer: Tracer):
|
| 11 |
self.cfg = cfg
|
| 12 |
self.tracer = tracer
|
| 13 |
self.backend = cfg.sql_backend # "bigquery" or "motherduck"
|
| 14 |
+
|
| 15 |
if self.backend == "bigquery":
|
| 16 |
from google.cloud import bigquery
|
| 17 |
from google.oauth2 import service_account
|
| 18 |
+
|
| 19 |
key_json = os.getenv("GCP_SERVICE_ACCOUNT_JSON")
|
| 20 |
if not key_json:
|
| 21 |
raise RuntimeError("Missing GCP_SERVICE_ACCOUNT_JSON secret")
|
| 22 |
+
|
| 23 |
+
# Accept full JSON string from Space Secret
|
| 24 |
+
if key_json.strip().startswith("{"):
|
| 25 |
+
import json
|
| 26 |
+
info = json.loads(key_json)
|
| 27 |
+
else:
|
| 28 |
+
info = {}
|
| 29 |
+
|
| 30 |
+
creds = service_account.Credentials.from_service_account_info(info)
|
| 31 |
self.client = bigquery.Client(credentials=creds, project=cfg.gcp_project)
|
| 32 |
+
|
| 33 |
+
elif self.backend == "motherduck":
|
| 34 |
+
import duckdb
|
| 35 |
+
|
| 36 |
+
token = self.cfg.motherduck_token or os.getenv("MOTHERDUCK_TOKEN")
|
| 37 |
+
db_name = self.cfg.motherduck_db or "default"
|
| 38 |
+
if not token:
|
| 39 |
+
raise RuntimeError("Missing MOTHERDUCK_TOKEN")
|
| 40 |
+
|
| 41 |
+
# Plain DuckDB connection
|
| 42 |
+
self.client = duckdb.connect()
|
| 43 |
+
|
| 44 |
+
# Ensure MotherDuck extension is available and loaded
|
| 45 |
+
self.client.execute("INSTALL motherduck;")
|
| 46 |
+
self.client.execute("LOAD motherduck;")
|
| 47 |
+
|
| 48 |
+
# Attach the remote MotherDuck database and use it
|
| 49 |
+
self.client.execute(f"SET motherduck_token='{token}';")
|
| 50 |
+
self.client.execute(f"ATTACH 'md:/{db_name}' AS md;")
|
| 51 |
+
self.client.execute("USE md;") # subsequent queries run against 'md' by default
|
| 52 |
else:
|
| 53 |
raise RuntimeError("Unknown SQL backend")
|
| 54 |
|
| 55 |
def _nl_to_sql(self, message: str) -> str:
|
| 56 |
+
"""
|
| 57 |
+
Minimal NL2SQL heuristic; replace with your own mapping or LLM prompt.
|
| 58 |
+
Expect users to include table names. Example:
|
| 59 |
+
"avg metric by month from analytics.events"
|
| 60 |
+
"""
|
| 61 |
m = message.lower()
|
| 62 |
+
|
| 63 |
+
# Very basic template example (edit to your tables/columns)
|
| 64 |
if "avg" in m and " by " in m:
|
| 65 |
+
return (
|
| 66 |
+
"-- Example template; edit me\n"
|
| 67 |
+
"SELECT DATE_TRUNC('month', date_col) AS month, "
|
| 68 |
+
"AVG(metric) AS avg_metric "
|
| 69 |
+
"FROM analytics.table "
|
| 70 |
+
"GROUP BY 1 ORDER BY 1;"
|
| 71 |
+
)
|
| 72 |
+
|
| 73 |
+
# Pass-through if the user typed SQL explicitly
|
| 74 |
if re.match(r"^\s*select ", m):
|
| 75 |
return message
|
| 76 |
+
|
| 77 |
+
# Fallback
|
| 78 |
+
return "SELECT * FROM analytics.table LIMIT 100;"
|
| 79 |
|
| 80 |
def run(self, message: str) -> pd.DataFrame:
|
| 81 |
sql = self._nl_to_sql(message)
|
| 82 |
self.tracer.trace_event("sql_query", {"sql": sql, "backend": self.backend})
|
| 83 |
+
|
| 84 |
if self.backend == "bigquery":
|
| 85 |
df = self.client.query(sql).to_dataframe()
|
| 86 |
else:
|
| 87 |
+
# DuckDB (MotherDuck): fetch_df returns a pandas DataFrame
|
| 88 |
df = self.client.execute(sql).fetch_df()
|
| 89 |
+
|
| 90 |
+
return df
|