airbyte-agent-mcp 0.1.30__py3-none-any.whl → 0.1.53__py3-none-any.whl

airbyte_agent_mcp/_vendored/connector_sdk/executor/local_executor.py

@@ -674,16 +674,16 @@ class LocalExecutor:
         return {key: value for key, value in params.items() if key in allowed_params}
 
     def _extract_body(self, allowed_fields: list[str], params: dict[str, Any]) -> dict[str, Any]:
-        """Extract body fields from params.
+        """Extract body fields from params, filtering out None values.
 
         Args:
             allowed_fields: List of allowed body field names
             params: All parameters
 
         Returns:
-            Dictionary of body fields
+            Dictionary of body fields with None values filtered out
         """
-        return {key: value for key, value in params.items() if key in allowed_fields}
+        return {key: value for key, value in params.items() if key in allowed_fields and value is not None}
 
     def _serialize_deep_object_params(self, params: dict[str, Any], deep_object_param_names: list[str]) -> dict[str, Any]:
         """Serialize deepObject parameters to bracket notation format.
@@ -837,15 +837,65 @@ class LocalExecutor:
             Request body dict or None if no body needed
         """
         if endpoint.graphql_body:
-            return self._build_graphql_body(endpoint.graphql_body, params)
+            # Extract defaults from query_params_schema for GraphQL variable interpolation
+            param_defaults = {name: schema.get("default") for name, schema in endpoint.query_params_schema.items() if "default" in schema}
+            return self._build_graphql_body(endpoint.graphql_body, params, param_defaults)
         elif endpoint.body_fields:
             return self._extract_body(endpoint.body_fields, params)
         return None
 
+    def _flatten_form_data(self, data: dict[str, Any], parent_key: str = "") -> dict[str, Any]:
+        """Flatten nested dict/list structures into bracket notation for form encoding.
+
+        Stripe and similar APIs require nested arrays/objects to be encoded using bracket
+        notation when using application/x-www-form-urlencoded content type.
+
+        Args:
+            data: Nested dict with arrays/objects to flatten
+            parent_key: Parent key for nested structures (used in recursion)
+
+        Returns:
+            Flattened dict with bracket notation keys
+
+        Examples:
+            >>> _flatten_form_data({"items": [{"price": "p1", "qty": 1}]})
+            {"items[0][price]": "p1", "items[0][qty]": 1}
+
+            >>> _flatten_form_data({"customer": "cus_123", "metadata": {"key": "value"}})
+            {"customer": "cus_123", "metadata[key]": "value"}
+        """
+        flattened = {}
+
+        for key, value in data.items():
+            new_key = f"{parent_key}[{key}]" if parent_key else key
+
+            if isinstance(value, dict):
+                # Recursively flatten nested dicts
+                flattened.update(self._flatten_form_data(value, new_key))
+            elif isinstance(value, list):
+                # Flatten arrays with indexed bracket notation
+                for i, item in enumerate(value):
+                    indexed_key = f"{new_key}[{i}]"
+                    if isinstance(item, dict):
+                        # Nested dict in array - recurse
+                        flattened.update(self._flatten_form_data(item, indexed_key))
+                    elif isinstance(item, list):
+                        # Nested list in array - recurse
+                        flattened.update(self._flatten_form_data({str(i): item}, new_key))
+                    else:
+                        # Primitive value in array
+                        flattened[indexed_key] = item
+            else:
+                # Primitive value - add directly
+                flattened[new_key] = value
+
+        return flattened
+
     def _determine_request_format(self, endpoint: EndpointDefinition, body: dict[str, Any] | None) -> dict[str, Any]:
         """Determine json/data parameters for HTTP request.
 
         GraphQL always uses JSON, regardless of content_type setting.
+        For form-encoded requests, nested structures are flattened into bracket notation.
 
         Args:
             endpoint: Endpoint definition
@@ -862,7 +912,9 @@ class LocalExecutor:
         if is_graphql or endpoint.content_type.value == "application/json":
             return {"json": body}
         elif endpoint.content_type.value == "application/x-www-form-urlencoded":
-            return {"data": body}
+            # Flatten nested structures for form encoding
+            flattened_body = self._flatten_form_data(body)
+            return {"data": flattened_body}
 
         return {}
 
@@ -903,12 +955,18 @@ class LocalExecutor:
 
         return query
 
-    def _build_graphql_body(self, graphql_config: dict[str, Any], params: dict[str, Any]) -> dict[str, Any]:
+    def _build_graphql_body(
+        self,
+        graphql_config: dict[str, Any],
+        params: dict[str, Any],
+        param_defaults: dict[str, Any] | None = None,
+    ) -> dict[str, Any]:
         """Build GraphQL request body with variable substitution and field selection.
 
         Args:
             graphql_config: GraphQL configuration from x-airbyte-body-type extension
             params: Parameters from execute() call
+            param_defaults: Default values for params from query_params_schema
 
         Returns:
             GraphQL request body: {"query": "...", "variables": {...}}
@@ -922,7 +980,7 @@ class LocalExecutor:
 
         # Substitute variables from params
         if "variables" in graphql_config and graphql_config["variables"]:
-            body["variables"] = self._interpolate_variables(graphql_config["variables"], params)
+            body["variables"] = self._interpolate_variables(graphql_config["variables"], params, param_defaults)
 
         # Add operation name if specified
         if "operationName" in graphql_config:
@@ -981,7 +1039,12 @@ class LocalExecutor:
         fields_str = " ".join(graphql_fields)
         return query.replace("{{ fields }}", fields_str)
 
-    def _interpolate_variables(self, variables: dict[str, Any], params: dict[str, Any]) -> dict[str, Any]:
+    def _interpolate_variables(
+        self,
+        variables: dict[str, Any],
+        params: dict[str, Any],
+        param_defaults: dict[str, Any] | None = None,
+    ) -> dict[str, Any]:
         """Recursively interpolate variables using params.
 
         Preserves types (doesn't stringify everything).
@@ -990,15 +1053,18 @@ class LocalExecutor:
         - Direct replacement: "{{ owner }}" → params["owner"] (preserves type)
         - Nested objects: {"input": {"name": "{{ name }}"}}
         - Arrays: [{"id": "{{ id }}"}]
-        - Unsubstituted placeholders: "{{ states }}" → None (for optional params)
+        - Default values: "{{ per_page }}" → param_defaults["per_page"] if not in params
+        - Unsubstituted placeholders: "{{ states }}" → None (for optional params without defaults)
 
         Args:
             variables: Variables dict with template placeholders
             params: Parameters to substitute
+            param_defaults: Default values for params from query_params_schema
 
         Returns:
             Interpolated variables dict with types preserved
         """
+        defaults = param_defaults or {}
 
         def interpolate_value(value: Any) -> Any:
             if isinstance(value, str):
@@ -1012,8 +1078,15 @@ class LocalExecutor:
                         value = value.replace(placeholder, str(param_value))
 
                 # Check if any unsubstituted placeholders remain
-                # If so, return None (treats as "not provided" for optional params)
                 if re.search(r"\{\{\s*\w+\s*\}\}", value):
+                    # Extract placeholder name and check for default value
+                    match = re.search(r"\{\{\s*(\w+)\s*\}\}", value)
+                    if match:
+                        param_name = match.group(1)
+                        if param_name in defaults:
+                            # Use default value (preserves type)
+                            return defaults[param_name]
+                    # No default found - return None (for optional params)
                     return None
 
                 return value
@@ -1056,7 +1129,7 @@ class LocalExecutor:
         if not action:
             return response_data
 
-        is_array_action = action in (Action.LIST, Action.SEARCH)
+        is_array_action = action in (Action.LIST, Action.API_SEARCH)
 
         try:
             # Parse and apply JSONPath expression
@@ -1151,21 +1224,18 @@ class LocalExecutor:
         if action not in (Action.CREATE, Action.UPDATE):
             return
 
-        # Check if endpoint has body fields defined
-        if not endpoint.body_fields:
+        # Get the request schema to find truly required fields
+        request_schema = endpoint.request_schema
+        if not request_schema:
             return
 
-        # For now, we treat all body_fields as potentially required for CREATE/UPDATE
-        # In a more advanced implementation, we could parse the request schema
-        # to identify truly required fields
-        missing_fields = []
-        for field in endpoint.body_fields:
-            if field not in params:
-                missing_fields.append(field)
+        # Only validate fields explicitly marked as required in the schema
+        required_fields = request_schema.get("required", [])
+        missing_fields = [field for field in required_fields if field not in params]
 
         if missing_fields:
             raise MissingParameterError(
-                f"Missing required body fields for {entity}.{action.value}: {missing_fields}. Provided parameters: {list(params.keys())}"
+                f"Missing required body fields for {entity}.{action.value}: {missing_fields}. " f"Provided parameters: {list(params.keys())}"
             )
 
     async def close(self):
@@ -1189,7 +1259,7 @@ class LocalExecutor:
 
 
 class _StandardOperationHandler:
-    """Handler for standard REST operations (GET, LIST, CREATE, UPDATE, DELETE, SEARCH, AUTHORIZE)."""
+    """Handler for standard REST operations (GET, LIST, CREATE, UPDATE, DELETE, API_SEARCH, AUTHORIZE)."""
 
     def __init__(self, context: _OperationContext):
         self.ctx = context
@@ -1202,7 +1272,7 @@ class _StandardOperationHandler:
             Action.CREATE,
             Action.UPDATE,
             Action.DELETE,
-            Action.SEARCH,
+            Action.API_SEARCH,
             Action.AUTHORIZE,
         }
 

airbyte_agent_mcp/_vendored/connector_sdk/extensions.py

@@ -159,6 +159,38 @@ Example:
     ```
 """
 
+AIRBYTE_STREAM_NAME = "x-airbyte-stream-name"
+"""
+Extension: x-airbyte-stream-name
+Location: Schema object (in components.schemas)
+Type: string
+Required: No
+
+Description:
+    Specifies the Airbyte stream name for cache lookup purposes. This maps the entity
+    to the corresponding Airbyte stream, enabling cache-based data retrieval. When
+    specified, the EntityDefinition.stream_name field will be populated with this value.
+
+    This extension is placed on Schema objects alongside x-airbyte-entity-name, following
+    the same pattern. The stream name is an entity-level property (not operation-level)
+    since an entity maps to exactly one Airbyte stream.
+
+Example:
+    ```yaml
+    components:
+      schemas:
+        Customer:
+          type: object
+          x-airbyte-entity-name: customers
+          x-airbyte-stream-name: customers
+          properties:
+            id:
+              type: string
+            name:
+              type: string
+    ```
+"""
+
 AIRBYTE_TOKEN_PATH = "x-airbyte-token-path"
 """
 Extension: x-airbyte-token-path
@@ -495,8 +527,8 @@ class ActionType(str, Enum):
     DELETE = "delete"
     """Delete a record"""
 
-    SEARCH = "search"
-    """Search for records matching specific query criteria"""
+    API_SEARCH = "api_search"
+    """Search for records matching specific query criteria via API"""
 
     DOWNLOAD = "download"
     """Download file content from a URL specified in the metadata response"""
@@ -514,7 +546,7 @@ class BodyType(str, Enum):
 
 
 # Type alias for use in Pydantic models
-ActionTypeLiteral = Literal["get", "list", "create", "update", "delete", "search", "download"]
+ActionTypeLiteral = Literal["get", "list", "create", "update", "delete", "api_search", "download"]
 
 
 # =============================================================================
@@ -548,6 +580,7 @@ def get_all_extension_names() -> list[str]:
         AIRBYTE_ENTITY,
         AIRBYTE_ACTION,
         AIRBYTE_ENTITY_NAME,
+        AIRBYTE_STREAM_NAME,
         AIRBYTE_TOKEN_PATH,
         AIRBYTE_BODY_TYPE,
         AIRBYTE_PATH_OVERRIDE,
@@ -594,6 +627,12 @@ EXTENSION_REGISTRY = {
         "required": False,
         "description": "Links schema to an entity/stream",
     },
+    AIRBYTE_STREAM_NAME: {
+        "location": "schema",
+        "type": "string",
+        "required": False,
+        "description": "Maps entity to Airbyte stream for cache lookup",
+    },
     AIRBYTE_TOKEN_PATH: {
         "location": "securityScheme",
         "type": "string",

airbyte_agent_mcp/_vendored/connector_sdk/http_client.py

@@ -147,6 +147,9 @@ class HTTPClient:
             self.base_url = self.base_url.replace(f"{{{var_name}}}", var_value)
 
         self.auth_config = auth_config
+        assert (
+            self.auth_config.type is not None
+        ), "auth_config.type cannot be None"  # Should never be None when instantiated via the local executor flow
         self.secrets = secrets
         self.logger = logger or NullLogger()
         self.metrics = HTTPMetrics()
@@ -296,12 +299,12 @@ class HTTPClient:
 
                         # Support both sync and async callbacks
                         callback_result = self.on_token_refresh(callback_data)
-                        if hasattr(callback_result, "__await__"):
+                        if callback_result is not None and hasattr(callback_result, "__await__"):
                             await callback_result
                     except Exception as callback_error:
                         self.logger.log_error(
                             request_id=None,
-                            error=("Token refresh callback failed during initialization: " f"{callback_error!s}"),
+                            error=(f"Token refresh callback failed during initialization: {callback_error!s}"),
                             status_code=None,
                         )
 
@@ -485,7 +488,7 @@ class HTTPClient:
                 elif "application/json" in content_type or not content_type:
                     response_data = await response.json()
                 else:
-                    error_msg = f"Expected JSON response for {method.upper()} {url}, " f"got content-type: {content_type}"
+                    error_msg = f"Expected JSON response for {method.upper()} {url}, got content-type: {content_type}"
                     raise HTTPClientError(error_msg)
 
             except ValueError as e:
@@ -556,6 +559,7 @@ class HTTPClient:
             current_token = self.secrets.get("access_token")
             strategy = AuthStrategyFactory.get_strategy(self.auth_config.type)
 
+            # Try to refresh credentials
             try:
                 result = await strategy.handle_auth_error(
                     status_code=status_code,
@@ -564,53 +568,56 @@ class HTTPClient:
                     config_values=self.config_values,
                     http_client=None,  # Let strategy create its own client
                 )
-
-                if result:
-                    # Notify callback if provided (for persistence)
-                    # Include both tokens AND extracted values for full persistence
-                    if self.on_token_refresh is not None:
-                        try:
-                            # Build callback data with both tokens and extracted values
-                            callback_data = dict(result.tokens)
-                            if result.extracted_values:
-                                callback_data.update(result.extracted_values)
-
-                            # Support both sync and async callbacks
-                            callback_result = self.on_token_refresh(callback_data)
-                            if hasattr(callback_result, "__await__"):
-                                await callback_result
-                        except Exception as callback_error:
-                            self.logger.log_error(
-                                request_id=request_id,
-                                error=f"Token refresh callback failed: {str(callback_error)}",
-                                status_code=status_code,
-                            )
-
-                    # Update secrets with new tokens (in-memory)
-                    self.secrets.update(result.tokens)
-
-                    # Update config_values and re-render base_url with extracted values
-                    if result.extracted_values:
-                        self._apply_token_extract(result.extracted_values)
-
-                    if self.secrets.get("access_token") != current_token:
-                        # Retry with new token - this will go through full retry logic
-                        return await self.request(
-                            method=method,
-                            path=path,
-                            params=params,
-                            json=json,
-                            data=data,
-                            headers=headers,
-                        )
-
             except Exception as refresh_error:
                 self.logger.log_error(
                     request_id=request_id,
                     error=f"Credential refresh failed: {str(refresh_error)}",
                     status_code=status_code,
                 )
+                result = None
+
+            # If refresh succeeded, update tokens and retry
+            if result:
+                # Notify callback if provided (for persistence)
+                # Include both tokens AND extracted values for full persistence
+                if self.on_token_refresh is not None:
+                    try:
+                        # Build callback data with both tokens and extracted values
+                        callback_data = dict(result.tokens)
+                        if result.extracted_values:
+                            callback_data.update(result.extracted_values)
+
+                        # Support both sync and async callbacks
+                        callback_result = self.on_token_refresh(callback_data)
+                        if callback_result is not None and hasattr(callback_result, "__await__"):
+                            await callback_result
+                    except Exception as callback_error:
+                        self.logger.log_error(
+                            request_id=request_id,
+                            error=f"Token refresh callback failed: {str(callback_error)}",
+                            status_code=status_code,
+                        )
+
+                # Update secrets with new tokens (in-memory)
+                self.secrets.update(result.tokens)
+
+                # Update config_values and re-render base_url with extracted values
+                if result.extracted_values:
+                    self._apply_token_extract(result.extracted_values)
 
+                if self.secrets.get("access_token") != current_token:
+                    # Retry with new token - this will go through full retry logic
+                    # Any errors from this retry will propagate to the caller
+                    return await self.request(
+                        method=method,
+                        path=path,
+                        params=params,
+                        json=json,
+                        data=data,
+                        headers=headers,
+                    )
+
+        # Refresh failed or token didn't change, log and let original error propagate
         self.logger.log_error(request_id=request_id, error=str(error), status_code=status_code)
 
     async def request(