Merge branch 'wip' of https://github.com/rootcodelabs/RAG-Module into streaming-response-formatting

Author: erangi-ar

src/llm_orchestration_service.py

@@ -34,6 +34,7 @@
     INPUT_GUARDRAIL_VIOLATION_MESSAGES,
     OUTPUT_GUARDRAIL_VIOLATION_MESSAGE,
     OUTPUT_GUARDRAIL_VIOLATION_MESSAGES,
+    QUERY_VALIDATION_FAILED_MESSAGES,
     get_localized_message,
     GUARDRAILS_BLOCKED_PHRASES,
     TEST_DEPLOYMENT_ENVIRONMENT,
@@ -52,6 +53,7 @@
 from src.utils.production_store import get_production_store
 from src.utils.language_detector import detect_language, get_language_name
 from src.utils.prompt_config_loader import PromptConfigurationLoader
+from src.utils.query_validator import validate_query_basic
 from src.guardrails import NeMoRailsAdapter, GuardrailCheckResult
 from src.contextual_retrieval import ContextualRetriever
 from src.llm_orchestrator_config.exceptions import (
@@ -170,7 +172,36 @@ def process_orchestration_request(
             # Using setattr for type safety - adds dynamic attribute to Pydantic model instance
             setattr(request, "_detected_language", detected_language)
 
-            # Initialize all service components
+            # STEP 0.5: Basic Query Validation (before expensive component initialization)
+            validation_result = validate_query_basic(request.message)
+            if not validation_result.is_valid:
+                logger.info(
+                    f"[{request.chatId}] Query validation failed: {validation_result.rejection_reason}"
+                )
+                # Get localized message
+                validation_msg = get_localized_message(
+                    QUERY_VALIDATION_FAILED_MESSAGES, detected_language
+                )
+
+                # Return appropriate response type without initializing components
+                if request.environment == TEST_DEPLOYMENT_ENVIRONMENT:
+                    return TestOrchestrationResponse(
+                        llmServiceActive=True,
+                        questionOutOfLLMScope=False,
+                        inputGuardFailed=False,
+                        content=validation_msg,
+                        chunks=None,
+                    )
+                else:
+                    return OrchestrationResponse(
+                        chatId=request.chatId,
+                        llmServiceActive=True,
+                        questionOutOfLLMScope=False,
+                        inputGuardFailed=False,
+                        content=validation_msg,
+                    )
+
+            # Initialize all service components (only for valid queries)
             components = self._initialize_service_components(request)
 
             # Execute the orchestration pipeline
@@ -299,6 +330,22 @@ async def stream_orchestration_response(
         # Using setattr for type safety - adds dynamic attribute to Pydantic model instance
         setattr(request, "_detected_language", detected_language)
 
+        # Step 0.5: Basic Query Validation (before guardrails)
+        validation_result = validate_query_basic(request.message)
+        if not validation_result.is_valid:
+            logger.info(
+                f"[{request.chatId}] Streaming - Query validation failed: {validation_result.rejection_reason}"
+            )
+            # Get localized message
+            validation_msg = get_localized_message(
+                QUERY_VALIDATION_FAILED_MESSAGES, detected_language
+            )
+
+            # Yield SSE format error + END marker
+            yield self._format_sse(request.chatId, validation_msg)
+            yield self._format_sse(request.chatId, "END")
+            return  # Stop processing
+
         # Use StreamManager for centralized tracking and guaranteed cleanup
         async with stream_manager.managed_stream(
             chat_id=request.chatId, author_id=request.authorId
@@ -953,6 +1000,9 @@ def _execute_orchestration_pipeline(
         timing_dict: Dict[str, float],
     ) -> Union[OrchestrationResponse, TestOrchestrationResponse]:
         """Execute the main orchestration pipeline with all components."""
+        # Note: Query validation now happens in process_orchestration_request()
+        # before component initialization for true early rejection
+
         # Step 1: Input Guardrails Check
         if components["guardrails_adapter"]:
             start_time = time.time()

src/llm_orchestration_service_api.py

@@ -839,6 +839,60 @@ def refresh_prompt_config(http_request: Request) -> Dict[str, Any]:
             },
         ) from e
 
+    try:
+        success = orchestration_service.prompt_config_loader.force_refresh()
+
+        if success:
+            # Get prompt metadata without exposing content (security)
+            custom_instructions = (
+                orchestration_service.prompt_config_loader.get_custom_instructions()
+            )
+            prompt_length = len(custom_instructions)
+
+            # Generate hash for verification purposes (without exposing content)
+            import hashlib
+
+            prompt_hash = hashlib.sha256(custom_instructions.encode()).hexdigest()[:16]
+
+            logger.info(
+                f"Prompt configuration cache refreshed successfully ({prompt_length} chars)"
+            )
+
+            return {
+                "refreshed": True,
+                "message": "Prompt configuration refreshed successfully",
+                "prompt_length": prompt_length,
+                "content_hash": prompt_hash,  # Safe: hash instead of preview
+            }
+        else:
+            # No fresh data loaded - could be fetch failure or truly not found
+            error_id = generate_error_id()
+            logger.warning(
+                f"[{error_id}] Prompt configuration refresh returned empty result"
+            )
+            raise HTTPException(
+                status_code=status.HTTP_404_NOT_FOUND,
+                detail={
+                    "error": "No prompt configuration found in database",
+                    "error_id": error_id,
+                },
+            )
+
+    except HTTPException:
+        # Re-raise HTTP exceptions as-is
+        raise
+    except Exception as e:
+        # Unexpected errors during refresh
+        error_id = generate_error_id()
+        logger.error(f"[{error_id}] Failed to refresh prompt configuration: {e}")
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+            detail={
+                "error": "Failed to refresh prompt configuration",
+                "error_id": error_id,
+            },
+        ) from e
+
 
 if __name__ == "__main__":
     logger.info("Starting LLM Orchestration Service API server on port 8100")

src/llm_orchestrator_config/llm_ochestrator_constants.py

@@ -23,6 +23,12 @@
     "en": "I apologize, but I'm unable to provide a response as it may violate our usage policies.",
 }
 
+# Query validation messages - single generic message for all rejection types
+# (empty queries, special characters only, too short, repetitive characters)
+QUERY_VALIDATION_FAILED_MESSAGES = {
+    "et": "Palun esitage kehtiv küsimus või sõnum, et ma saaksin teid aidata."
+}
+
 # Legacy constants for backward compatibility (English defaults)
 OUT_OF_SCOPE_MESSAGE = OUT_OF_SCOPE_MESSAGES["en"]
 TECHNICAL_ISSUE_MESSAGE = TECHNICAL_ISSUE_MESSAGES["en"]
@@ -106,9 +112,9 @@
 
 
 # Helper function to get localized messages
-def get_localized_message(message_dict: dict, language_code: str = "en") -> str:
+def get_localized_message(message_dict: dict, language_code: str = "et") -> str:
     """
-    Get message in the specified language, fallback to English.
+    Get message in the specified language, fallback to Estonian.
 
     Args:
         message_dict: Dictionary with language codes as keys
@@ -117,7 +123,7 @@ def get_localized_message(message_dict: dict, language_code: str = "en") -> str:
     Returns:
         Localized message string
     """
-    return message_dict.get(language_code, message_dict.get("en", ""))
+    return message_dict.get(language_code, message_dict.get("et", ""))
 
 
 # Service endpoints

src/models/request_models.py

@@ -66,19 +66,20 @@ class OrchestrationRequest(BaseModel):
     def validate_and_sanitize_message(cls, v: str) -> str:
         """Sanitize and validate user message.
 
-        Note: Content safety checks (prompt injection, PII, harmful content)
+        Note: This validator only handles security/format concerns:
+        - XSS/HTML sanitization
+        - Maximum length enforcement
+
+        Query quality validation (empty messages, special chars, etc.) is handled
+        by the business logic layer (query_validator) with localized error messages.
+
+        Content safety checks (prompt injection, PII, harmful content)
         are handled by NeMo Guardrails after this validation layer.
         """
         # Sanitize HTML/XSS and normalize whitespace
         v = InputSanitizer.sanitize_message(v)
 
-        # Check if message is empty after sanitization
-        if not v or len(v.strip()) < 3:
-            raise ValueError(
-                "Message must contain at least 3 characters after sanitization"
-            )
-
-        # Check length after sanitization
+        # Check length after sanitization (resource protection)
         if len(v) > StreamConfig.MAX_MESSAGE_LENGTH:
             raise ValueError(
                 f"Message exceeds maximum length of {StreamConfig.MAX_MESSAGE_LENGTH} characters"

src/utils/query_validator.py

@@ -0,0 +1,112 @@
+"""Basic query validation for empty/meaningless inputs.
+
+This module provides lightweight, rule-based validation to reject syntactically
+invalid queries before they reach expensive LLM-based processing stages.
+
+Validation checks (all syntactic, NO semantic):
+- Empty or whitespace-only messages
+- Messages containing only special characters/punctuation (including unicode)
+- Messages with too few meaningful characters (< 2)
+- Messages with only repetitive characters (e.g., "aaaa", "????")
+- Emoji-only messages
+
+Out of scope for this module:
+- Semantic validation (greetings, chitchat, intent detection)
+- Language quality checks
+- Content policy checks (handled by guardrails)
+
+Design decisions:
+- Numbers are considered valid (e.g., "123" passes validation)
+- Mixed alphanumeric with punctuation is valid (e.g., "ab!" passes)
+- Unicode punctuation is treated same as ASCII punctuation
+- Emojis are not considered meaningful characters
+"""
+
+import re
+from typing import Optional
+from pydantic import BaseModel
+
+
+class QueryValidationResult(BaseModel):
+    """Result of basic query validation.
+
+    Attributes:
+        is_valid: True if query passes all validation checks
+        rejection_reason: Optional reason code if validation fails
+                         (empty, special_chars_only, too_short, repetitive)
+    """
+
+    is_valid: bool
+    rejection_reason: Optional[str] = None
+
+
+def validate_query_basic(query: str) -> QueryValidationResult:
+    """
+    Validate query for basic syntactic issues (NOT semantic).
+
+    This is a fast, rule-based check that runs before expensive operations
+    like guardrails or prompt refinement. It only catches obvious syntactic
+    issues, not semantic problems.
+
+    Args:
+        query: User's input message to validate
+
+    Returns:
+        QueryValidationResult with is_valid flag and optional rejection_reason
+
+    Examples:
+        Valid queries:
+        >>> validate_query_basic("How to apply for benefits?")
+        QueryValidationResult(is_valid=True, rejection_reason=None)
+        >>> validate_query_basic("hi")
+        QueryValidationResult(is_valid=True, rejection_reason=None)
+        >>> validate_query_basic("123")
+        QueryValidationResult(is_valid=True, rejection_reason=None)
+        >>> validate_query_basic("ab!")
+        QueryValidationResult(is_valid=True, rejection_reason=None)
+
+        Invalid queries:
+        >>> validate_query_basic("...")
+        QueryValidationResult(is_valid=False, rejection_reason='special_chars_only')
+        >>> validate_query_basic("")
+        QueryValidationResult(is_valid=False, rejection_reason='empty')
+        >>> validate_query_basic("????")
+        QueryValidationResult(is_valid=False, rejection_reason='repetitive')
+        >>> validate_query_basic("a")
+        QueryValidationResult(is_valid=False, rejection_reason='too_short')
+        >>> validate_query_basic("😀😀😀")
+        QueryValidationResult(is_valid=False, rejection_reason='special_chars_only')
+    """
+    # Trim whitespace
+    query = query.strip()
+
+    # Check 1: Empty query
+    if not query:
+        return QueryValidationResult(is_valid=False, rejection_reason="empty")
+
+    # Check 2: Only special characters/punctuation (including unicode and emojis)
+    # Remove all alphanumeric characters (letters and numbers in any language)
+    # If nothing remains or only punctuation/symbols/emojis, reject
+    alphanumeric_pattern = re.compile(r"[\w]", re.UNICODE)
+    has_alphanumeric = bool(alphanumeric_pattern.search(query))
+
+    if not has_alphanumeric:
+        # No letters or numbers found - only punctuation/symbols/emojis
+        return QueryValidationResult(
+            is_valid=False, rejection_reason="special_chars_only"
+        )
+
+    # Check 3: Too short (< 2 meaningful characters)
+    # Extract alphanumeric characters (letters + numbers, unicode-aware)
+    meaningful_chars = alphanumeric_pattern.findall(query)
+    if len(meaningful_chars) < 2:
+        return QueryValidationResult(is_valid=False, rejection_reason="too_short")
+
+    # Check 4: Only repetitive characters (e.g., "aaaa", "????", "111")
+    # If all meaningful characters are the same (case-insensitive), likely spam
+    unique_chars = {c.lower() for c in meaningful_chars}
+    if len(unique_chars) == 1:
+        return QueryValidationResult(is_valid=False, rejection_reason="repetitive")
+
+    # Passed all checks - query is syntactically valid
+    return QueryValidationResult(is_valid=True)

tests/conftest.py

@@ -0,0 +1,8 @@
+"""Pytest configuration for test discovery and imports."""
+
+import sys
+from pathlib import Path
+
+# Add the project root to Python path so tests can import from src
+project_root = Path(__file__).parent.parent
+sys.path.insert(0, str(project_root))