gaard-core 0.2.7__tar.gz → 0.2.8__tar.gz

{gaard_core-0.2.7 → gaard_core-0.2.8}/PKG-INFO

@@ -1,10 +1,10 @@
 Metadata-Version: 2.4
 Name: gaard-core
-Version: 0.2.7
+Version: 0.2.8
 Summary: Core GAARD query pipeline, prompt compiler, policies and SQL validation
 Requires-Python: >=3.11
 Description-Content-Type: text/markdown
-Requires-Dist: gaard-plugin-api<0.3.0,>=0.2.7
+Requires-Dist: gaard-plugin-api<0.3.0,>=0.2.8
 Requires-Dist: pydantic>=2.7.0
 Requires-Dist: sqlglot>=25.0.0
 Provides-Extra: dev

{gaard_core-0.2.7 → gaard_core-0.2.8}/pyproject.toml

@@ -4,12 +4,12 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "gaard-core"
-version = "0.2.7"
+version = "0.2.8"
 description = "Core GAARD query pipeline, prompt compiler, policies and SQL validation"
 readme = "README.md"
 requires-python = ">=3.11"
 dependencies = [
-    "gaard-plugin-api>=0.2.7,<0.3.0",
+    "gaard-plugin-api>=0.2.8,<0.3.0",
     "pydantic>=2.7.0",
     "sqlglot>=25.0.0",
 ]

{gaard_core-0.2.7 → gaard_core-0.2.8}/src/gaard_core/query_pipeline/llm_sql_generator.py

@@ -1,3 +1,4 @@
+import logging
 from typing import Any, Protocol
 
 from gaard_core.prompt_compiler.models import CompiledPrompt, SqlGenerationPromptRequest
@@ -8,6 +9,9 @@ from gaard_llm.openai_compatible.client import OpenAICompatibleClient
 from gaard_llm.providers.models import ChatCompletionRequest, ChatMessage
 
 
+logger = logging.getLogger(__name__)
+
+
 class SqlPromptCompiler(Protocol):
     def compile(self, request: SqlGenerationPromptRequest) -> CompiledPrompt:
         pass
@@ -42,6 +46,20 @@ class LlmSqlGenerator:
             )
         )
 
+        # This is deliberately logged as separate fields so a failed text-to-SQL
+        # request can be reconstructed from the service log without guessing
+        # which dialect or schema context reached the model.
+        logger.info(
+            "SQL generation request: model=%r dialect=%r max_rows=%s question=%r "
+            "system_prompt=%r user_prompt=%r",
+            self.model,
+            self.dialect,
+            self.max_rows,
+            request.question,
+            compiled_prompt.system_prompt,
+            compiled_prompt.user_prompt,
+        )
+
         response = self.client.create_chat_completion(
             ChatCompletionRequest(
                 model=self.model,
@@ -62,12 +80,20 @@ class LlmSqlGenerator:
 
         sql = self._clean_sql(response.content)
 
+        logger.info(
+            "SQL generation response: model=%r raw_response=%r cleaned_sql=%r",
+            self.model,
+            response.content,
+            sql,
+        )
+
         return GeneratedSql(
             sql=sql,
             confidence=0.0,
             assumptions=[
                 "SQL generated by OpenAI-compatible LLM provider.",
             ],
+            prompt_metadata=compiled_prompt.metadata,
         )
 
     def _clean_sql(self, value: str) -> str:

{gaard_core-0.2.7 → gaard_core-0.2.8}/src/gaard_core/query_pipeline/models.py

@@ -62,6 +62,7 @@ class GeneratedSql(BaseModel):
     sql: str
     confidence: float = 0.0
     assumptions: list[str] = Field(default_factory=list)
+    prompt_metadata: dict[str, Any] = Field(default_factory=dict)
 
 
 class QueryResult(BaseModel):

{gaard_core-0.2.7 → gaard_core-0.2.8}/src/gaard_core/query_pipeline/pipeline.py

@@ -1,4 +1,6 @@
+import logging
 import time
+from collections.abc import Callable
 from typing import Protocol
 
 from gaard_core.execution.mock_executor import MockQueryExecutor
@@ -16,6 +18,9 @@ from gaard_core.result_interpreter.mock_interpreter import MockResultInterpreter
 from gaard_core.sql_validator.select_only import SelectOnlySqlValidator
 
 
+logger = logging.getLogger(__name__)
+
+
 class SqlGenerator(Protocol):
     def generate(self, request: QueryRequest) -> GeneratedSql:
         pass
@@ -57,7 +62,11 @@ class QueryPipeline:
         self.result_interpretation_mode = result_interpretation_mode
         self.output_classification_mode = output_classification_mode
 
-    def handle(self, request: QueryRequest) -> QueryResponse:
+    def handle(
+        self,
+        request: QueryRequest,
+        on_stage: Callable[[str], None] | None = None,
+    ) -> QueryResponse:
         started_at = time.perf_counter()
 
         try:
@@ -70,28 +79,49 @@ class QueryPipeline:
                 error_detail=exc.message,
             ) from exc
 
+        logger.info(
+            "SQL pipeline generated query: datasource_id=%r datasource_ids=%r sql=%r "
+            "validator=%s executor=%s",
+            request.datasource_id,
+            request.datasource_ids,
+            generated_sql.sql,
+            type(self.sql_validator).__name__,
+            type(self.executor).__name__,
+        )
         self.sql_validator.validate(generated_sql.sql)
+        logger.info("SQL pipeline validation passed: sql=%r", generated_sql.sql)
 
+        if on_stage is not None:
+            on_stage("waiting_on_data_server")
         result = self.executor.execute(generated_sql.sql)
+        logger.info(
+            "SQL pipeline execution completed: sql=%r columns=%r row_count=%s",
+            generated_sql.sql,
+            result.columns,
+            len(result.rows),
+        )
         if not request.interpret:
             duration_ms = round((time.perf_counter() - started_at) * 1000, 2)
+            metadata = {
+                "duration_ms": duration_ms,
+                "datasource_id": request.datasource_id,
+                "user_id": request.user_id,
+                "confidence": generated_sql.confidence,
+                "assumptions": generated_sql.assumptions,
+                "sql_generation_mode": self.sql_generation_mode,
+                "result_interpretation_mode": "none",
+                "output_classification_mode": "none",
+                "output_classification": OutputClassification.UNKNOWN.value,
+                "raw_sql_output": True,
+            }
+            if generated_sql.prompt_metadata:
+                metadata["sql_generation_prompt_metadata"] = generated_sql.prompt_metadata
             return QueryResponse(
                 question=request.question,
                 answer="",
                 sql=generated_sql.sql,
                 rows=result.rows,
-                metadata={
-                    "duration_ms": duration_ms,
-                    "datasource_id": request.datasource_id,
-                    "user_id": request.user_id,
-                    "confidence": generated_sql.confidence,
-                    "assumptions": generated_sql.assumptions,
-                    "sql_generation_mode": self.sql_generation_mode,
-                    "result_interpretation_mode": "none",
-                    "output_classification_mode": "none",
-                    "output_classification": OutputClassification.UNKNOWN.value,
-                    "raw_sql_output": True,
-                },
+                metadata=metadata,
             )
 
         try:
@@ -135,6 +165,9 @@ class QueryPipeline:
         if classification_error:
             metadata["output_classification_error"] = classification_error
 
+        if generated_sql.prompt_metadata:
+            metadata["sql_generation_prompt_metadata"] = generated_sql.prompt_metadata
+
         return QueryResponse(
             question=request.question,
             answer=answer,

{gaard_core-0.2.7 → gaard_core-0.2.8}/src/gaard_core.egg-info/PKG-INFO

@@ -1,10 +1,10 @@
 Metadata-Version: 2.4
 Name: gaard-core
-Version: 0.2.7
+Version: 0.2.8
 Summary: Core GAARD query pipeline, prompt compiler, policies and SQL validation
 Requires-Python: >=3.11
 Description-Content-Type: text/markdown
-Requires-Dist: gaard-plugin-api<0.3.0,>=0.2.7
+Requires-Dist: gaard-plugin-api<0.3.0,>=0.2.8
 Requires-Dist: pydantic>=2.7.0
 Requires-Dist: sqlglot>=25.0.0
 Provides-Extra: dev

{gaard_core-0.2.7 → gaard_core-0.2.8}/src/gaard_core.egg-info/requires.txt

@@ -1,4 +1,4 @@
-gaard-plugin-api<0.3.0,>=0.2.7
+gaard-plugin-api<0.3.0,>=0.2.8
 pydantic>=2.7.0
 sqlglot>=25.0.0