Release

backend/main.py

@@ -2,7 +2,7 @@
 import uvicorn
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
-from routes import query, azure, system, user_queries, data_documents
+from routes import query, azure, system, user_queries, data_documents, audit
 
 app = FastAPI()
 
@@ -55,6 +55,7 @@ async def health_check():
 app.include_router(system.router, prefix="/system", tags=["System"])
 app.include_router(user_queries.router, prefix="/user", tags=["User Queries"])
 app.include_router(data_documents.router, prefix="/data", tags=["Data Documents"])
+app.include_router(audit.router, prefix="/audit", tags=["Audit"])
 
 if __name__ == "__main__":
     uvicorn.run(app, host="0.0.0.0", port=8000)

backend/routes/audit.py

@@ -0,0 +1,33 @@
+from fastapi import APIRouter, Header, Body, HTTPException
+from pydantic import BaseModel
+from typing import List, Dict, Any, Optional
+from services.audit_service import process_audit_question
+from services.gemini_service import VisualizationConfig
+
+router = APIRouter()
+
+
+class AuditQueryRequest(BaseModel):
+    question: str
+
+
+class AuditQueryResponse(BaseModel):
+    sql_query: str
+    results: List[Dict[str, Any]]
+    summary: str
+    visualization: Optional[VisualizationConfig] = None
+
+
+@router.post("/query", response_model=AuditQueryResponse)
+def query_audit_log(
+    body: AuditQueryRequest = Body(...), authorization: str = Header(...)
+):
+    if not authorization.startswith("Bearer "):
+        raise HTTPException(status_code=401, detail="Invalid token format")
+
+    # We might want to validate the token here even if we don't use it for the pg connection directly yet
+    # user_token = authorization.replace("Bearer ", "")
+    # access_token = exchange_token_obo(user_token)
+
+    response = process_audit_question(body.question)
+    return AuditQueryResponse(**response)

backend/services/audit_service.py

@@ -0,0 +1,73 @@
+from typing import Dict, Any
+from services.pg_connection import get_connection
+from services.gemini_service import generate_audit_sql, summarize_audit_results
+
+
+def execute_audit_query(sql_query: str) -> list:
+    """
+    Executes a read-only SQL query against the audit database.
+    """
+    if not sql_query.lower().strip().startswith("select"):
+        return [{"error": "Only SELECT queries are allowed."}]
+
+    try:
+        conn = get_connection()
+        cur = conn.cursor()
+        cur.execute(sql_query)
+
+        # Get column names
+        columns = [desc[0] for desc in cur.description]
+        results = [dict(zip(columns, row)) for row in cur.fetchall()]
+
+        # Serialize datetime and json objects
+        for row in results:
+            for key, value in row.items():
+                if hasattr(value, "isoformat"):
+                    row[key] = value.isoformat()
+                elif isinstance(value, dict):
+                    # Ensure dicts (like diff_data) are kept as dicts for the frontend
+                    pass
+
+        cur.close()
+        conn.close()
+        return results
+    except Exception as e:
+        print(f"Error executing audit query: {e}")
+        return [{"error": str(e)}]
+
+
+def process_audit_question(question: str) -> Dict[str, Any]:
+    """
+    Orchestrates the process of answering a user's audit question:
+    1. Generate SQL from NL question (via Gemini)
+    2. Execute SQL
+    3. Summarize results (via Gemini)
+    """
+    sql_query = generate_audit_sql(question)
+
+    # If the generator returned an error query or invalid SQL, return it
+    if "Error:" in sql_query:
+        return {
+            "sql_query": sql_query,
+            "results": [],
+            "summary": "Could not generate a valid query for your request.",
+        }
+
+    results = execute_audit_query(sql_query)
+
+    # If execution failed
+    if results and "error" in results[0]:
+        return {
+            "sql_query": sql_query,
+            "results": [],
+            "summary": f"Error executing query: {results[0]['error']}",
+        }
+
+    summary_response = summarize_audit_results(question, sql_query, results)
+
+    return {
+        "sql_query": sql_query,
+        "results": results,
+        "summary": summary_response.summary,
+        "visualization": summary_response.visualization,
+    }

backend/services/gemini_service.py

@@ -1,6 +1,29 @@
 from google import genai
 from google.genai import types
 from models.schemas import GeneratedCode, CollectionContext, DebugSuggestionResponse
+from pydantic import BaseModel, Field
+from typing import Optional, List
+
+
+class VisualizationConfig(BaseModel):
+    available: bool = Field(description="Whether a chart is recommended for this data")
+    type: Optional[str] = Field(
+        description="Type of chart: 'bar', 'line', 'pie', 'scatter'"
+    )
+    x_key: Optional[str] = Field(description="Key for X-axis data")
+    y_key: Optional[str] = Field(description="Key for Y-axis data")
+    title: Optional[str] = Field(description="Title for the chart")
+    data_keys: Optional[List[str]] = Field(
+        description="Keys to include in the chart data points (e.g. ['count', 'date'])"
+    )
+
+
+class AuditSummaryResponse(BaseModel):
+    summary: str = Field(description="Markdown summary of the results")
+    visualization: VisualizationConfig = Field(
+        description="Configuration for data visualization"
+    )
+
 
 PROMPT_TEMPLATE_QUERY = """
 You are an assistant that converts user requests into MongoDB query code.
@@ -132,3 +155,92 @@ def generate_suggestion_from_query_error(query: str, error_message: str) -> str:
         response.text.strip() if hasattr(response, "text") else str(response).strip
     )
     return DebugSuggestionResponse(suggestion=suggestion)
+
+
+PROMPT_TEMPLATE_AUDIT_SQL = """
+You are a PostgreSQL expert. Convert the user's natural language question into a read-only SQL query for the `write_audit_log` table.
+Table Schema:
+- user_email (text): Email of the user who performed the operation.
+- operation (text): 'insert', 'update', or 'delete'.
+- database_name (text): Name of the database (format: account.database).
+- collection_name (text): Name of the collection.
+- document_id (text): ID of the affected document.
+- diff_data (jsonb): JSON containing the changes (for updates, it has 'before' and 'after' fields).
+- timestamp_utc (timestamptz): When the operation occurred.
+
+User Question: "{user_input}"
+
+Rules:
+1. Return ONLY the SQL query. No markdown, no explanations.
+2. The query MUST be a SELECT statement.
+3. Use LIMIT 100 if no limit is specified.
+4. If the user asks for "recent", order by timestamp_utc DESC.
+"""
+
+PROMPT_TEMPLATE_AUDIT_SUMMARY = """
+You are a data analyst. Analyze the following SQL query and its results.
+
+User Question: "{user_input}"
+SQL Query: "{sql_query}"
+Results:
+{results}
+
+Tasks:
+1. Provide a concise markdown summary identifying patterns or answering the specific question.
+2. Determine if the data is suitable for visualization (e.g., time series, counts, comparisons).
+3. If suitable, structure a visualization configuration (type, keys, title).
+   - For time series, prefer 'line' or 'bar'.
+   - For categorical counts, use 'bar' or 'pie'.
+"""
+
+
+def generate_audit_sql(user_input: str) -> str:
+    full_prompt = PROMPT_TEMPLATE_AUDIT_SQL.format(user_input=user_input)
+    client = genai.Client()
+    response = client.models.generate_content(
+        model="gemini-2.5-flash",
+        contents=full_prompt,
+        config=types.GenerateContentConfig(
+            thinking_config=types.ThinkingConfig(thinking_budget=0)
+        ),
+    )
+    sql = extract_python_code(response.text)
+    # Basic safety check
+    if not sql.lower().startswith("select"):
+        return "SELECT 'Error: Generated query was not a SELECT statement' as error;"
+    return sql
+
+
+def summarize_audit_results(
+    user_input: str, sql_query: str, results: list
+) -> AuditSummaryResponse:
+    # Truncate results if too large to avoid token limits
+    results_str = str(results)[:10000]
+    full_prompt = PROMPT_TEMPLATE_AUDIT_SUMMARY.format(
+        user_input=user_input, sql_query=sql_query, results=results_str
+    )
+    client = genai.Client()
+    response = client.models.generate_content(
+        model="gemini-2.5-flash",
+        contents=full_prompt,
+        config=types.GenerateContentConfig(
+            response_mime_type="application/json",
+            response_schema=AuditSummaryResponse,
+            thinking_config=types.ThinkingConfig(thinking_budget=0),
+        ),
+    )
+
+    if hasattr(response, "parsed") and response.parsed:
+        return response.parsed
+
+    import json
+
+    try:
+        data = json.loads(response.text)
+        return AuditSummaryResponse(**data)
+    except Exception as e:
+        print(f"Error parsing Gemini response: {e}")
+        return AuditSummaryResponse(
+            summary="Could not generate summary due to parsing error.",
+            visualization=VisualizationConfig(available=False),
+        )

backend/tests/test_audit_service.py

@@ -0,0 +1,56 @@
+import unittest
+from unittest.mock import patch, MagicMock
+from services.audit_service import process_audit_question
+
+
+class TestAuditService(unittest.TestCase):
+
+    @patch("services.audit_service.generate_audit_sql")
+    @patch("services.audit_service.get_connection")
+    @patch("services.audit_service.summarize_audit_results")
+    def test_process_audit_question_success(
+        self, mock_summarize, mock_get_conn, mock_generate_sql
+    ):
+        # Setup mocks
+        mock_generate_sql.return_value = "SELECT * FROM write_audit_log LIMIT 5"
+
+        mock_conn = MagicMock()
+        mock_cursor = MagicMock()
+        mock_get_conn.return_value = mock_conn
+        mock_conn.cursor.return_value = mock_cursor
+
+        # Mock DB results
+        mock_cursor.description = [("user_email",), ("operation",)]
+        mock_cursor.fetchall.return_value = [("[email protected]", "insert")]
+
+        # Mock Summary Response
+        mock_response = MagicMock()
+        mock_response.summary = "Summary of results."
+        mock_response.visualization = None
+        mock_summarize.return_value = mock_response
+
+        # Execute
+        result = process_audit_question("Show me inserts")
+
+        # Assertions
+        self.assertEqual(result["sql_query"], "SELECT * FROM write_audit_log LIMIT 5")
+        self.assertEqual(len(result["results"]), 1)
+        self.assertEqual(result["results"][0]["user_email"], "[email protected]")
+        self.assertEqual(result["summary"], "Summary of results.")
+
+        mock_generate_sql.assert_called_once()
+        mock_cursor.execute.assert_called_with("SELECT * FROM write_audit_log LIMIT 5")
+        mock_summarize.assert_called_once()
+
+    @patch("services.audit_service.generate_audit_sql")
+    def test_process_audit_question_invalid_sql(self, mock_generate_sql):
+        mock_generate_sql.return_value = "DELETE FROM write_audit_log"
+
+        result = process_audit_question("Delete everything")
+
+        self.assertIn("Error executing query", result["summary"])
+        self.assertIn("Only SELECT", result["summary"])
+
+
+if __name__ == "__main__":
+    unittest.main()