airbyte-agent-orb 0.1.0__py3-none-any.whl

airbyte_agent_orb/_vendored/connector_sdk/executor/local_executor.py

@@ -0,0 +1,1773 @@
+"""Local executor for direct HTTP execution of connector operations."""
+
+from __future__ import annotations
+
+import asyncio
+import inspect
+import logging
+import os
+import re
+import time
+from collections.abc import AsyncIterator
+from typing import Any, Protocol
+from urllib.parse import quote
+
+from jinja2 import Environment, StrictUndefined
+from jsonpath_ng import parse as parse_jsonpath
+from opentelemetry import trace
+
+from ..auth_template import apply_auth_mapping
+from ..connector_model_loader import load_connector_model
+from ..constants import (
+    DEFAULT_MAX_CONNECTIONS,
+    DEFAULT_MAX_KEEPALIVE_CONNECTIONS,
+)
+from ..http_client import HTTPClient, TokenRefreshCallback
+from ..logging import NullLogger, RequestLogger
+from ..observability import ObservabilitySession
+from ..schema.extensions import RetryConfig
+from ..secrets import SecretStr
+from ..telemetry import SegmentTracker
+from ..types import (
+    Action,
+    AuthConfig,
+    AuthOption,
+    ConnectorModel,
+    EndpointDefinition,
+    EntityDefinition,
+)
+
+from .models import (
+    ActionNotSupportedError,
+    EntityNotFoundError,
+    ExecutionConfig,
+    ExecutionResult,
+    ExecutorError,
+    InvalidParameterError,
+    MissingParameterError,
+    StandardExecuteResult,
+)
+
+
+class _OperationContext:
+    """Shared context for operation handlers."""
+
+    def __init__(self, executor: LocalExecutor):
+        self.executor = executor
+        self.http_client = executor.http_client
+        self.tracker = executor.tracker
+        self.session = executor.session
+        self.logger = executor.logger
+        self.entity_index = executor._entity_index
+        self.operation_index = executor._operation_index
+        # Bind helper methods
+        self.build_path = executor._build_path
+        self.extract_query_params = executor._extract_query_params
+        self.extract_body = executor._extract_body
+        self.extract_header_params = executor._extract_header_params
+        self.build_request_body = executor._build_request_body
+        self.determine_request_format = executor._determine_request_format
+        self.validate_required_body_fields = executor._validate_required_body_fields
+        self.extract_records = executor._extract_records
+
+
+class _OperationHandler(Protocol):
+    """Protocol for operation handlers."""
+
+    def can_handle(self, action: Action) -> bool:
+        """Check if this handler can handle the given action."""
+        ...
+
+    async def execute_operation(
+        self,
+        entity: str,
+        action: Action,
+        params: dict[str, Any],
+    ) -> StandardExecuteResult | AsyncIterator[bytes]:
+        """Execute the operation and return result.
+
+        Returns:
+            StandardExecuteResult for standard operations (GET, LIST, CREATE, etc.)
+            AsyncIterator[bytes] for download operations
+        """
+        ...
+
+
+class LocalExecutor:
+    """Async executor for Entity×Action operations with direct HTTP execution.
+
+    This is the "local mode" executor that makes direct HTTP calls to external APIs.
+    It performs local entity/action lookups, validation, and request building.
+
+    Implements ExecutorProtocol.
+    """
+
+    def __init__(
+        self,
+        config_path: str | None = None,
+        model: ConnectorModel | None = None,
+        secrets: dict[str, SecretStr] | None = None,
+        auth_config: dict[str, SecretStr] | None = None,
+        auth_scheme: str | None = None,
+        enable_logging: bool = False,
+        log_file: str | None = None,
+        execution_context: str | None = None,
+        max_connections: int = DEFAULT_MAX_CONNECTIONS,
+        max_keepalive_connections: int = DEFAULT_MAX_KEEPALIVE_CONNECTIONS,
+        max_logs: int | None = 10000,
+        config_values: dict[str, str] | None = None,
+        on_token_refresh: TokenRefreshCallback = None,
+        retry_config: RetryConfig | None = None,
+    ):
+        """Initialize async executor.
+
+        Args:
+            config_path: Path to connector.yaml.
+                If neither config_path nor model is provided, an error will be raised.
+            model: ConnectorModel object to execute.
+            secrets: (Legacy) Auth parameters that bypass x-airbyte-auth-config mapping.
+                Directly passed to auth strategies (e.g., {"username": "...", "password": "..."}).
+                Cannot be used together with auth_config.
+            auth_config: User-facing auth configuration following x-airbyte-auth-config spec.
+                Will be transformed via auth_mapping to produce auth parameters.
+                Cannot be used together with secrets.
+            auth_scheme: (Multi-auth only) Explicit security scheme name to use.
+                If None, SDK will auto-select based on provided credentials.
+                Example: auth_scheme="githubOAuth"
+            enable_logging: Enable request/response logging
+            log_file: Path to log file (if enable_logging=True)
+            execution_context: Execution context (mcp, direct, blessed, agent)
+            max_connections: Maximum number of concurrent connections
+            max_keepalive_connections: Maximum number of keepalive connections
+            max_logs: Maximum number of logs to keep in memory before rotation.
+                Set to None for unlimited (not recommended for production).
+                Defaults to 10000.
+            config_values: Optional dict of config values for server variable substitution
+                (e.g., {"subdomain": "acme"} for URLs like https://{subdomain}.api.example.com).
+            on_token_refresh: Optional callback function(new_tokens: dict) called when
+                OAuth2 tokens are refreshed. Use this to persist updated tokens.
+                Can be sync or async. Example: lambda tokens: save_to_db(tokens)
+            retry_config: Optional retry configuration override. If provided, overrides
+                the connector.yaml x-airbyte-retry-config. If None, uses connector.yaml
+                config or SDK defaults.
+        """
+        # Validate mutual exclusivity of secrets and auth_config
+        if secrets is not None and auth_config is not None:
+            raise ValueError(
+                "Cannot provide both 'secrets' and 'auth_config' parameters. "
+                "Use 'auth_config' for user-facing credentials (recommended), "
+                "or 'secrets' for direct auth parameters (legacy)."
+            )
+
+        # Validate mutual exclusivity of config_path and model
+        if config_path is not None and model is not None:
+            raise ValueError("Cannot provide both 'config_path' and 'model' parameters.")
+
+        if config_path is None and model is None:
+            raise ValueError("Must provide either 'config_path' or 'model' parameter.")
+
+        # Load model from path or use provided model
+        if config_path is not None:
+            self.model: ConnectorModel = load_connector_model(config_path)
+        else:
+            self.model: ConnectorModel = model
+
+        self.on_token_refresh = on_token_refresh
+        self.config_values = config_values or {}
+
+        # Handle auth selection for multi-auth or single-auth connectors
+        user_credentials = auth_config if auth_config is not None else secrets
+        selected_auth_config, self.secrets = self._initialize_auth(user_credentials, auth_scheme)
+
+        # Create shared observability session
+        self.session = ObservabilitySession(
+            connector_name=self.model.name,
+            connector_version=getattr(self.model, "version", None),
+            execution_context=(execution_context or os.getenv("AIRBYTE_EXECUTION_CONTEXT", "direct")),
+        )
+
+        # Initialize telemetry tracker
+        self.tracker = SegmentTracker(self.session)
+        self.tracker.track_connector_init(connector_version=getattr(self.model, "version", None))
+
+        # Initialize logger
+        if enable_logging:
+            self.logger = RequestLogger(
+                log_file=log_file,
+                connector_name=self.model.name,
+                max_logs=max_logs,
+            )
+        else:
+            self.logger = NullLogger()
+
+        # Initialize async HTTP client with connection pooling
+        self.http_client = HTTPClient(
+            base_url=self.model.base_url,
+            auth_config=selected_auth_config,
+            secrets=self.secrets,
+            config_values=self.config_values,
+            logger=self.logger,
+            max_connections=max_connections,
+            max_keepalive_connections=max_keepalive_connections,
+            on_token_refresh=on_token_refresh,
+            retry_config=retry_config or self.model.retry_config,
+        )
+
+        # Build O(1) lookup indexes
+        self._entity_index: dict[str, EntityDefinition] = {entity.name: entity for entity in self.model.entities}
+
+        # Build O(1) operation index: (entity, action) -> endpoint
+        self._operation_index: dict[tuple[str, Action], Any] = {}
+        for entity in self.model.entities:
+            for action in entity.actions:
+                endpoint = entity.endpoints.get(action)
+                if endpoint:
+                    self._operation_index[(entity.name, action)] = endpoint
+
+        # Register operation handlers (order matters for can_handle priority)
+        op_context = _OperationContext(self)
+        self._operation_handlers: list[_OperationHandler] = [
+            _DownloadOperationHandler(op_context),
+            _StandardOperationHandler(op_context),
+        ]
+
+    def _apply_auth_config_mapping(self, user_secrets: dict[str, SecretStr]) -> dict[str, SecretStr]:
+        """Apply auth_mapping from x-airbyte-auth-config to transform user secrets.
+
+        This method takes user-provided secrets (e.g., {"api_token": "abc123"}) and
+        transforms them into the auth scheme format (e.g., {"username": "abc123", "password": "api_token"})
+        using the template mappings defined in x-airbyte-auth-config.
+
+        Args:
+            user_secrets: User-provided secrets from config
+
+        Returns:
+            Transformed secrets matching the auth scheme requirements
+        """
+        if not self.model.auth.user_config_spec:
+            # No x-airbyte-auth-config defined, use secrets as-is
+            return user_secrets
+
+        user_config_spec = self.model.auth.user_config_spec
+        auth_mapping = None
+        required_fields: list[str] | None = None
+
+        # Check for single option (direct auth_mapping)
+        if user_config_spec.auth_mapping:
+            auth_mapping = user_config_spec.auth_mapping
+            required_fields = user_config_spec.required
+        # Check for oneOf (multiple auth options)
+        elif user_config_spec.one_of:
+            # Find the matching option based on which required fields are present
+            for option in user_config_spec.one_of:
+                option_required = option.required or []
+                if all(field in user_secrets for field in option_required):
+                    auth_mapping = option.auth_mapping
+                    required_fields = option_required
+                    break
+
+        if not auth_mapping:
+            # No matching auth_mapping found, use secrets as-is
+            return user_secrets
+
+        # If required fields are missing and user provided no credentials,
+        # return as-is (allows empty auth for testing or optional auth)
+        if required_fields and not user_secrets:
+            return user_secrets
+
+        # Convert SecretStr values to plain strings for template processing
+        user_config_values = {
+            key: (value.get_secret_value() if hasattr(value, "get_secret_value") else str(value)) for key, value in user_secrets.items()
+        }
+
+        # Apply the auth_mapping templates, passing required_fields so optional
+        # fields that are not provided can be skipped
+        mapped_values = apply_auth_mapping(auth_mapping, user_config_values, required_fields=required_fields)
+
+        # Convert back to SecretStr
+        mapped_secrets = {key: SecretStr(value) for key, value in mapped_values.items()}
+
+        return mapped_secrets
+
+    def _initialize_auth(
+        self,
+        user_credentials: dict[str, SecretStr] | None,
+        explicit_scheme: str | None,
+    ) -> tuple[AuthConfig, dict[str, SecretStr] | None]:
+        """Initialize authentication for single or multi-auth connectors.
+
+        Handles both legacy single-auth and new multi-auth connectors.
+        For multi-auth, the auth scheme can be explicitly provided or inferred
+        from the provided credentials by matching against each scheme's required fields.
+
+        Args:
+            user_credentials: User-provided credentials (auth_config or secrets)
+            explicit_scheme: Explicit scheme name for multi-auth (optional, will be
+                inferred from credentials if not provided)
+
+        Returns:
+            Tuple of (selected AuthConfig for HTTPClient, transformed secrets)
+
+        Raises:
+            ValueError: If multi-auth connector can't determine which scheme to use
+        """
+        # Multi-auth: explicit scheme selection or inference from credentials
+        if self.model.auth.is_multi_auth():
+            if not user_credentials:
+                available_schemes = [opt.scheme_name for opt in self.model.auth.options]
+                raise ValueError(f"Multi-auth connector requires credentials. Available schemes: {available_schemes}")
+
+            # If explicit scheme provided, use it directly
+            if explicit_scheme:
+                selected_option, transformed_secrets = self._select_auth_option(user_credentials, explicit_scheme)
+            else:
+                # Infer auth scheme from provided credentials
+                selected_option, transformed_secrets = self._infer_auth_scheme(user_credentials)
+
+            # Convert AuthOption to single-auth AuthConfig for HTTPClient
+            selected_auth_config = AuthConfig(
+                type=selected_option.type,
+                config=selected_option.config,
+                user_config_spec=None,  # Not needed by HTTPClient
+            )
+
+            return (selected_auth_config, transformed_secrets)
+
+        # Single-auth: use existing logic
+        if user_credentials is not None:
+            # Apply mapping if this is auth_config (not legacy secrets)
+            transformed_secrets = self._apply_auth_config_mapping(user_credentials)
+        else:
+            transformed_secrets = None
+
+        return (self.model.auth, transformed_secrets)
+
+    def _infer_auth_scheme(
+        self,
+        user_credentials: dict[str, SecretStr],
+    ) -> tuple[AuthOption, dict[str, SecretStr]]:
+        """Infer authentication scheme from provided credentials.
+
+        Matches user credentials against each auth option's required fields
+        to determine which scheme to use.
+
+        Args:
+            user_credentials: User-provided credentials
+
+        Returns:
+            Tuple of (inferred AuthOption, transformed secrets)
+
+        Raises:
+            ValueError: If no scheme matches, or multiple schemes match
+        """
+        options = self.model.auth.options
+        if not options:
+            raise ValueError("No auth options available in multi-auth config")
+
+        # Get the credential keys provided by the user
+        provided_keys = set(user_credentials.keys())
+
+        # Find all options where all required fields are present
+        matching_options: list[AuthOption] = []
+        for option in options:
+            if option.user_config_spec and option.user_config_spec.required:
+                required_fields = set(option.user_config_spec.required)
+                if required_fields.issubset(provided_keys):
+                    matching_options.append(option)
+            elif not option.user_config_spec or not option.user_config_spec.required:
+                # Option has no required fields - it matches any credentials
+                matching_options.append(option)
+
+        # Handle matching results
+        if len(matching_options) == 0:
+            # No matches - provide helpful error message
+            scheme_requirements = []
+            for opt in options:
+                required = opt.user_config_spec.required if opt.user_config_spec and opt.user_config_spec.required else []
+                scheme_requirements.append(f"  - {opt.scheme_name}: requires {required}")
+            raise ValueError(
+                f"Could not infer auth scheme from provided credentials. "
+                f"Provided keys: {list(provided_keys)}. "
+                f"Available schemes and their required fields:\n" + "\n".join(scheme_requirements)
+            )
+
+        if len(matching_options) > 1:
+            # Multiple matches - need explicit scheme
+            matching_names = [opt.scheme_name for opt in matching_options]
+            raise ValueError(
+                f"Multiple auth schemes match the provided credentials: {matching_names}. Please specify 'auth_scheme' explicitly to disambiguate."
+            )
+
+        # Exactly one match - use it
+        selected_option = matching_options[0]
+        transformed_secrets = self._apply_auth_mapping_for_option(user_credentials, selected_option)
+        return (selected_option, transformed_secrets)
+
+    def _select_auth_option(
+        self,
+        user_credentials: dict[str, SecretStr],
+        scheme_name: str,
+    ) -> tuple[AuthOption, dict[str, SecretStr]]:
+        """Select authentication option by explicit scheme name.
+
+        Args:
+            user_credentials: User-provided credentials
+            scheme_name: Explicit scheme name (e.g., "githubOAuth")
+
+        Returns:
+            Tuple of (selected AuthOption, transformed secrets)
+
+        Raises:
+            ValueError: If scheme not found
+        """
+        options = self.model.auth.options
+        if not options:
+            raise ValueError("No auth options available in multi-auth config")
+
+        # Find matching scheme
+        for option in options:
+            if option.scheme_name == scheme_name:
+                transformed_secrets = self._apply_auth_mapping_for_option(user_credentials, option)
+                return (option, transformed_secrets)
+
+        # Scheme not found
+        available = [opt.scheme_name for opt in options]
+        raise ValueError(f"Auth scheme '{scheme_name}' not found. Available schemes: {available}")
+
+    def _apply_auth_mapping_for_option(
+        self,
+        user_credentials: dict[str, SecretStr],
+        option: AuthOption,
+    ) -> dict[str, SecretStr]:
+        """Apply auth mapping for a specific auth option.
+
+        Transforms user credentials using the option's auth_mapping templates.
+
+        Args:
+            user_credentials: User-provided credentials
+            option: AuthOption to apply
+
+        Returns:
+            Transformed secrets after applying auth_mapping
+
+        Raises:
+            ValueError: If required fields are missing or mapping fails
+        """
+        if not option.user_config_spec:
+            # No mapping defined, use credentials as-is
+            return user_credentials
+
+        # Extract auth_mapping and required fields
+        user_config_spec = option.user_config_spec
+        auth_mapping = user_config_spec.auth_mapping
+        required_fields = user_config_spec.required
+
+        if not auth_mapping:
+            raise ValueError(f"No auth_mapping found for scheme '{option.scheme_name}'")
+
+        # Convert SecretStr to plain strings for template processing
+        user_config_values = {
+            key: (value.get_secret_value() if hasattr(value, "get_secret_value") else str(value)) for key, value in user_credentials.items()
+        }
+
+        # Apply the auth_mapping templates
+        mapped_values = apply_auth_mapping(auth_mapping, user_config_values, required_fields=required_fields)
+
+        # Convert back to SecretStr
+        return {key: SecretStr(value) for key, value in mapped_values.items()}
+
+    async def execute(self, config: ExecutionConfig) -> ExecutionResult:
+        """Execute connector operation using handler pattern.
+
+        Args:
+            config: Execution configuration (entity, action, params)
+
+        Returns:
+            ExecutionResult with success/failure status and data
+
+        Example:
+            config = ExecutionConfig(
+                entity="customers",
+                action="list",
+                params={"limit": 10}
+            )
+            result = await executor.execute(config)
+            if result.success:
+                print(result.data)
+        """
+        try:
+            # Check for hosted-only actions before converting to Action enum
+            if config.action == "search":
+                raise NotImplementedError(
+                    "search is only available in hosted execution mode. "
+                    "Initialize the connector with external_user_id, airbyte_client_id, "
+                    "and airbyte_client_secret to use this feature."
+                )
+
+            # Convert config to internal format
+            action = Action(config.action) if isinstance(config.action, str) else config.action
+            params = config.params or {}
+
+            # Dispatch to handler (handlers handle telemetry internally)
+            handler = next((h for h in self._operation_handlers if h.can_handle(action)), None)
+            if not handler:
+                raise ExecutorError(f"No handler registered for action '{action.value}'.")
+
+            # Execute handler
+            result = handler.execute_operation(config.entity, action, params)
+
+            # Check if it's an async generator (download) or awaitable (standard)
+            if inspect.isasyncgen(result):
+                # Download operation: return generator directly
+                return ExecutionResult(
+                    success=True,
+                    data=result,
+                    error=None,
+                    meta=None,
+                )
+            else:
+                # Standard operation: await and extract data and metadata
+                handler_result = await result
+                return ExecutionResult(
+                    success=True,
+                    data=handler_result.data,
+                    error=None,
+                    meta=handler_result.metadata,
+                )
+
+        except (
+            EntityNotFoundError,
+            ActionNotSupportedError,
+            MissingParameterError,
+            InvalidParameterError,
+        ) as e:
+            # These are "expected" execution errors - return them in ExecutionResult
+            return ExecutionResult(success=False, data={}, error=str(e))
+
+    async def _execute_operation(
+        self,
+        entity: str,
+        action: str | Action,
+        params: dict[str, Any] | None = None,
+    ) -> dict[str, Any]:
+        """Internal method: Execute an action on an entity asynchronously.
+
+        This method now delegates to the appropriate handler and extracts just the data.
+        External code should use execute(config) instead for full ExecutionResult with metadata.
+
+        Args:
+            entity: Entity name (e.g., "Customer")
+            action: Action to execute (e.g., "get" or Action.GET)
+            params: Parameters for the operation
+                - For GET: {"id": "cus_123"} for path params
+                - For LIST: {"limit": 10} for query params
+                - For CREATE/UPDATE: {"email": "...", "name": "..."} for body
+
+        Returns:
+            API response as dictionary
+
+        Raises:
+            ValueError: If entity or action not found
+            HTTPClientError: If API request fails
+        """
+        params = params or {}
+        action = Action(action) if isinstance(action, str) else action
+
+        # Delegate to the appropriate handler
+        handler = next((h for h in self._operation_handlers if h.can_handle(action)), None)
+        if not handler:
+            raise ExecutorError(f"No handler registered for action '{action.value}'.")
+
+        # Execute handler and extract just the data for backward compatibility
+        result = await handler.execute_operation(entity, action, params)
+        if isinstance(result, StandardExecuteResult):
+            return result.data
+        else:
+            # Download operation returns AsyncIterator directly
+            return result
+
+    async def execute_batch(self, operations: list[tuple[str, str | Action, dict[str, Any] | None]]) -> list[dict[str, Any] | AsyncIterator[bytes]]:
+        """Execute multiple operations concurrently (supports all action types including download).
+
+        Args:
+            operations: List of (entity, action, params) tuples
+
+        Returns:
+            List of responses in the same order as operations.
+            Standard operations return dict[str, Any].
+            Download operations return AsyncIterator[bytes].
+
+        Raises:
+            ValueError: If any entity or action not found
+            HTTPClientError: If any API request fails
+
+        Example:
+            results = await executor.execute_batch([
+                ("Customer", "list", {"limit": 10}),
+                ("Customer", "get", {"id": "cus_123"}),
+                ("attachments", "download", {"id": "att_456"}),
+            ])
+        """
+        # Build tasks by dispatching directly to handlers
+        tasks = []
+        for entity, action, params in operations:
+            # Convert action to Action enum if needed
+            action = Action(action) if isinstance(action, str) else action
+            params = params or {}
+
+            # Find appropriate handler
+            handler = next((h for h in self._operation_handlers if h.can_handle(action)), None)
+            if not handler:
+                raise ExecutorError(f"No handler registered for action '{action.value}'.")
+
+            # Call handler directly (exceptions propagate naturally)
+            tasks.append(handler.execute_operation(entity, action, params))
+
+        # Execute all tasks concurrently - exceptions propagate via asyncio.gather
+        results = await asyncio.gather(*tasks)
+
+        # Extract data from results
+        extracted_results = []
+        for result in results:
+            if isinstance(result, StandardExecuteResult):
+                # Standard operation: extract data
+                extracted_results.append(result.data)
+            else:
+                # Download operation: return iterator as-is
+                extracted_results.append(result)
+
+        return extracted_results
+
+    def _build_path(self, path_template: str, params: dict[str, Any]) -> str:
+        """Build path by replacing {param} placeholders with URL-encoded values.
+
+        Args:
+            path_template: Path with placeholders (e.g., /v1/customers/{id})
+            params: Parameters containing values for placeholders
+
+        Returns:
+            Completed path with URL-encoded values (e.g., /v1/customers/cus_123)
+
+        Raises:
+            MissingParameterError: If required path parameter is missing
+        """
+        placeholders = re.findall(r"\{(\w+)\}", path_template)
+
+        path = path_template
+        for placeholder in placeholders:
+            if placeholder not in params:
+                raise MissingParameterError(
+                    f"Missing required path parameter '{placeholder}' for path '{path_template}'. Provided parameters: {list(params.keys())}"
+                )
+
+            # Validate parameter value is not None or empty string
+            value = params[placeholder]
+            if value is None or (isinstance(value, str) and value.strip() == ""):
+                raise InvalidParameterError(f"Path parameter '{placeholder}' cannot be None or empty string")
+
+            encoded_value = quote(str(value), safe="")
+            path = path.replace(f"{{{placeholder}}}", encoded_value)
+
+        return path
+
+    def _extract_query_params(self, allowed_params: list[str], params: dict[str, Any]) -> dict[str, Any]:
+        """Extract query parameters from params.
+
+        Args:
+            allowed_params: List of allowed query parameter names
+            params: All parameters
+
+        Returns:
+            Dictionary of query parameters
+        """
+        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, filtering out None values.
+
+        Args:
+            allowed_fields: List of allowed body field names
+            params: All parameters
+
+        Returns:
+            Dictionary of body fields with None values filtered out
+        """
+        return {key: value for key, value in params.items() if key in allowed_fields and value is not None}
+
+    def _extract_header_params(self, endpoint: EndpointDefinition, params: dict[str, Any], body: dict[str, Any] | None = None) -> dict[str, str]:
+        """Extract header parameters from params and schema defaults.
+
+        Also adds Content-Type header when there's a request body (unless already specified
+        as a header parameter in the OpenAPI spec).
+
+        Args:
+            endpoint: Endpoint definition with header_params and header_params_schema
+            params: All parameters
+            body: Request body (if any) - used to determine if Content-Type should be added
+
+        Returns:
+            Dictionary of header name -> value
+        """
+        headers: dict[str, str] = {}
+
+        for header_name in endpoint.header_params:
+            # Check if value is provided in params
+            if header_name in params and params[header_name] is not None:
+                headers[header_name] = str(params[header_name])
+            # Otherwise, use default from schema if available
+            elif header_name in endpoint.header_params_schema:
+                default_value = endpoint.header_params_schema[header_name].get("default")
+                if default_value is not None:
+                    headers[header_name] = str(default_value)
+
+        # Add Content-Type header when there's a request body, but only if not already
+        # specified as a header parameter (which allows custom content types like
+        # application/vnd.spCampaign.v3+json)
+        if body is not None and endpoint.content_type and "Content-Type" not in headers:
+            headers["Content-Type"] = endpoint.content_type.value
+
+        return headers
+
+    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.
+
+        Converts nested dict parameters to the deepObject format expected by APIs
+        like Stripe. For example:
+        - Input: {'created': {'gte': 123, 'lte': 456}}
+        - Output: {'created[gte]': 123, 'created[lte]': 456}
+
+        Args:
+            params: Query parameters dict (may contain nested dicts)
+            deep_object_param_names: List of parameter names that use deepObject style
+
+        Returns:
+            Dictionary with deepObject params serialized to bracket notation
+        """
+        serialized = {}
+
+        for key, value in params.items():
+            if key in deep_object_param_names and isinstance(value, dict):
+                # Serialize nested dict to bracket notation
+                for subkey, subvalue in value.items():
+                    if subvalue is not None:  # Skip None values
+                        serialized[f"{key}[{subkey}]"] = subvalue
+            else:
+                # Keep non-deepObject params as-is (already validated by _extract_query_params)
+                serialized[key] = value
+
+        return serialized
+
+    @staticmethod
+    def _extract_download_url(
+        response: dict[str, Any],
+        file_field: str,
+        entity: str,
+    ) -> str:
+        """Extract download URL from metadata response using x-airbyte-file-url.
+
+        Supports both simple dot notation (e.g., "article.content_url") and array
+        indexing with bracket notation (e.g., "calls[0].media.audioUrl").
+
+        Args:
+            response: Metadata response containing file reference
+            file_field: JSON path to file URL field (from x-airbyte-file-url)
+            entity: Entity name (for error messages)
+
+        Returns:
+            Extracted file URL
+
+        Raises:
+            ExecutorError: If file field not found or invalid
+        """
+        # Navigate nested path (e.g., "article_attachment.content_url" or "calls[0].media.audioUrl")
+        parts = file_field.split(".")
+        current = response
+
+        for i, part in enumerate(parts):
+            # Check if part has array indexing (e.g., "calls[0]")
+            array_match = re.match(r"^(\w+)\[(\d+)\]$", part)
+
+            if array_match:
+                field_name = array_match.group(1)
+                index = int(array_match.group(2))
+
+                # Navigate to the field
+                if not isinstance(current, dict):
+                    raise ExecutorError(
+                        f"Cannot extract download URL for {entity}: Expected dict at '{'.'.join(parts[:i])}', got {type(current).__name__}"
+                    )
+
+                if field_name not in current:
+                    raise ExecutorError(
+                        f"Cannot extract download URL for {entity}: "
+                        f"Field '{field_name}' not found in response. Available fields: {list(current.keys())}"
+                    )
+
+                # Get the array
+                array_value = current[field_name]
+                if not isinstance(array_value, list):
+                    raise ExecutorError(
+                        f"Cannot extract download URL for {entity}: Expected list at '{field_name}', got {type(array_value).__name__}"
+                    )
+
+                # Check index bounds
+                if index >= len(array_value):
+                    raise ExecutorError(
+                        f"Cannot extract download URL for {entity}: Index {index} out of bounds for '{field_name}' (length: {len(array_value)})"
+                    )
+
+                current = array_value[index]
+            else:
+                # Regular dict navigation
+                if not isinstance(current, dict):
+                    raise ExecutorError(
+                        f"Cannot extract download URL for {entity}: Expected dict at '{'.'.join(parts[:i])}', got {type(current).__name__}"
+                    )
+
+                if part not in current:
+                    raise ExecutorError(
+                        f"Cannot extract download URL for {entity}: Field '{part}' not found in response. Available fields: {list(current.keys())}"
+                    )
+
+                current = current[part]
+
+        if not isinstance(current, str):
+            raise ExecutorError(f"Cannot extract download URL for {entity}: Expected string at '{file_field}', got {type(current).__name__}")
+
+        return current
+
+    @staticmethod
+    def _substitute_file_field_params(
+        file_field: str,
+        params: dict[str, Any],
+    ) -> str:
+        """Substitute template variables in file_field with parameter values.
+
+        Uses Jinja2 with custom delimiters to support OpenAPI-style syntax like
+        "attachments[{index}].url" where {index} is replaced with params["index"].
+
+        Args:
+            file_field: File field path with optional template variables
+            params: Parameters from execute() call
+
+        Returns:
+            File field with template variables substituted
+
+        Example:
+            >>> _substitute_file_field_params("attachments[{attachment_index}].url", {"attachment_index": 0})
+            "attachments[0].url"
+        """
+
+        # Use custom delimiters to match OpenAPI path parameter syntax {var}
+        # StrictUndefined raises clear error if a template variable is missing
+        env = Environment(
+            variable_start_string="{",
+            variable_end_string="}",
+            undefined=StrictUndefined,
+        )
+        template = env.from_string(file_field)
+        return template.render(params)
+
+    def _build_request_body(self, endpoint: EndpointDefinition, params: dict[str, Any]) -> dict[str, Any] | None:
+        """Build request body (GraphQL or standard).
+
+        Args:
+            endpoint: Endpoint definition
+            params: Parameters from execute() call
+
+        Returns:
+            Request body dict or None if no body needed
+        """
+        if endpoint.graphql_body:
+            # 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:
+            # Start with defaults from request body schema
+            body = dict(endpoint.request_body_defaults)
+            # Override with user-provided params (filtering out None values)
+            user_body = self._extract_body(endpoint.body_fields, params)
+            body.update(user_body)
+            return body if body else None
+        elif endpoint.request_body_defaults:
+            # If no body_fields but we have defaults, return the defaults
+            return dict(endpoint.request_body_defaults)
+        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
+            body: Request body dict or None
+
+        Returns:
+            Dict with 'json' and/or 'data' keys for http_client.request()
+        """
+        if not body:
+            return {}
+
+        is_graphql = endpoint.graphql_body is not None
+
+        if is_graphql or endpoint.content_type.value == "application/json":
+            return {"json": body}
+        elif endpoint.content_type.value == "application/x-www-form-urlencoded":
+            # Flatten nested structures for form encoding
+            flattened_body = self._flatten_form_data(body)
+            return {"data": flattened_body}
+
+        return {}
+
+    def _process_graphql_fields(self, query: str, graphql_config: dict[str, Any], params: dict[str, Any]) -> str:
+        """Process GraphQL query field selection.
+
+        Handles:
+        - Dynamic fields from params['fields']
+        - Default fields from config
+        - String vs list format for default_fields
+
+        Args:
+            query: GraphQL query string (may contain {{ fields }} placeholder)
+            graphql_config: GraphQL configuration dict
+            params: Parameters from execute() call
+
+        Returns:
+            Processed query string with fields injected
+        """
+        if "{{ fields }}" not in query:
+            return query
+
+        # Check for explicit fields parameter
+        if "fields" in params and params["fields"]:
+            return self._inject_graphql_fields(query, params["fields"])
+
+        # Use default fields if available
+        if "default_fields" not in graphql_config:
+            return query  # Placeholder remains (could raise error in the future)
+
+        default_fields = graphql_config["default_fields"]
+        if isinstance(default_fields, str):
+            # Already in GraphQL format - direct replacement
+            return query.replace("{{ fields }}", default_fields)
+        elif isinstance(default_fields, list):
+            # List format - convert to GraphQL
+            return self._inject_graphql_fields(query, default_fields)
+
+        return query
+
+    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": {...}}
+        """
+        query = graphql_config["query"]
+
+        # Process field selection (dynamic fields or default fields)
+        query = self._process_graphql_fields(query, graphql_config, params)
+
+        body = {"query": query}
+
+        # Substitute variables from params
+        if "variables" in graphql_config and graphql_config["variables"]:
+            variables = self._interpolate_variables(graphql_config["variables"], params, param_defaults)
+            # Filter out None values (optional fields not provided) - matches REST _extract_body() behavior
+            # But preserve None for variables explicitly marked as nullable (e.g., to unassign a user)
+            nullable_vars = set(graphql_config.get("x-airbyte-nullable-variables") or [])
+            body["variables"] = {k: v for k, v in variables.items() if v is not None or k in nullable_vars}
+
+        # Add operation name if specified
+        if "operationName" in graphql_config:
+            body["operationName"] = graphql_config["operationName"]
+
+        return body
+
+    def _convert_nested_field_to_graphql(self, field: str) -> str:
+        """Convert dot-notation field to GraphQL field selection.
+
+        Example: "primaryLanguage.name" -> "primaryLanguage { name }"
+
+        Args:
+            field: Field name in dot notation (e.g., "primaryLanguage.name")
+
+        Returns:
+            GraphQL field selection string
+        """
+        if "." not in field:
+            return field
+
+        parts = field.split(".")
+        result = parts[0]
+        for part in parts[1:]:
+            result += f" {{ {part}"
+        result += " }" * (len(parts) - 1)
+        return result
+
+    def _inject_graphql_fields(self, query: str, fields: list[str]) -> str:
+        """Inject field selection into GraphQL query.
+
+        Replaces field selection placeholders ({{ fields }}) with actual field list.
+        Supports nested fields using dot notation (e.g., "primaryLanguage.name").
+
+        Args:
+            query: GraphQL query string (may contain {{ fields }} placeholder)
+            fields: List of fields to select (e.g., ["id", "name", "primaryLanguage.name"])
+
+        Returns:
+            GraphQL query with fields injected
+
+        Example:
+            Input query: "query { repository { {{ fields }} } }"
+            Fields: ["id", "name", "primaryLanguage { name }"]
+            Output: "query { repository { id name primaryLanguage { name } } }"
+        """
+        # Check if query has field placeholder
+        if "{{ fields }}" not in query:
+            # No placeholder - return query as-is (backward compatible)
+            return query
+
+        # Convert field list to GraphQL field selection
+        graphql_fields = [self._convert_nested_field_to_graphql(field) for field in fields]
+
+        # Replace placeholder with field list
+        fields_str = " ".join(graphql_fields)
+        return query.replace("{{ fields }}", fields_str)
+
+    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).
+
+        Supports:
+        - Direct replacement: "{{ owner }}" → params["owner"] (preserves type)
+        - Nested objects: {"input": {"name": "{{ name }}"}}
+        - Arrays: [{"id": "{{ id }}"}]
+        - 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):
+                # Check for exact template match (preserve type)
+                for key, param_value in params.items():
+                    placeholder = f"{{{{ {key} }}}}"
+                    if value == placeholder:
+                        return param_value  # Return actual value (int, list, etc.)
+                    elif placeholder in value:
+                        # Partial match - do string replacement
+                        value = value.replace(placeholder, str(param_value))
+
+                # Check if any unsubstituted placeholders remain
+                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
+            elif isinstance(value, dict):
+                return {k: interpolate_value(v) for k, v in value.items()}
+            elif isinstance(value, list):
+                return [interpolate_value(item) for item in value]
+            else:
+                return value
+
+        return interpolate_value(variables)
+
+    def _wrap_primitives(self, data: Any) -> dict[str, Any] | list[dict[str, Any]] | None:
+        """Wrap primitive values in dict format for consistent response structure.
+
+        Transforms primitive API responses into dict format so downstream code
+        can always expect dict-based data structures.
+
+        Args:
+            data: Response data (could be primitive, list, dict, or None)
+
+        Returns:
+            - If data is a primitive (str, int, float, bool): {"value": data}
+            - If data is a list: wraps all non-dict elements as {"value": item}
+            - If data is already a dict or list of dicts: unchanged
+            - If data is None: None
+
+        Examples:
+            >>> executor._wrap_primitives(42)
+            {"value": 42}
+            >>> executor._wrap_primitives([1, 2, 3])
+            [{"value": 1}, {"value": 2}, {"value": 3}]
+            >>> executor._wrap_primitives([1, {"id": 2}, 3])
+            [{"value": 1}, {"id": 2}, {"value": 3}]
+            >>> executor._wrap_primitives([[1, 2], 3])
+            [{"value": [1, 2]}, {"value": 3}]
+            >>> executor._wrap_primitives({"id": 1})
+            {"id": 1}  # unchanged
+        """
+        if data is None:
+            return None
+
+        # Handle primitive scalars
+        if isinstance(data, (bool, str, int, float)):
+            return {"value": data}
+
+        # Handle lists - wrap non-dict elements
+        if isinstance(data, list):
+            if not data:
+                return []  # Empty list unchanged
+
+            wrapped = []
+            for item in data:
+                if isinstance(item, dict):
+                    wrapped.append(item)
+                else:
+                    wrapped.append({"value": item})
+            return wrapped
+
+        # Dict - return unchanged
+        if isinstance(data, dict):
+            return data
+
+        # Unknown type - wrap for safety
+        return {"value": data}
+
+    def _extract_records(
+        self,
+        response_data: Any,
+        endpoint: EndpointDefinition,
+    ) -> dict[str, Any] | list[dict[str, Any]] | None:
+        """Extract records from response using record extractor.
+
+        Type inference based on action:
+        - list, search: Returns array ([] if not found)
+        - get, create, update, delete: Returns single record (None if not found)
+
+        Automatically wraps primitive values (int, str, float, bool) in {"value": primitive}
+        format to ensure consistent dict-based responses for downstream code.
+
+        Args:
+            response_data: Full API response (can be dict, list, primitive, or None)
+            endpoint: Endpoint with optional record extractor and action
+
+        Returns:
+            - Extracted data if extractor configured and path found
+            - [] or None if path not found (based on action)
+            - Original response if no extractor configured or on error
+            - Primitives are wrapped as {"value": primitive}
+        """
+        # Check if endpoint has record extractor
+        extractor = endpoint.record_extractor
+        if not extractor:
+            return self._wrap_primitives(response_data)
+
+        # Determine if this action returns array or single record
+        action = endpoint.action
+        if not action:
+            return self._wrap_primitives(response_data)
+
+        is_array_action = action in (Action.LIST, Action.API_SEARCH)
+
+        try:
+            # Parse and apply JSONPath expression
+            jsonpath_expr = parse_jsonpath(extractor)
+            matches = [match.value for match in jsonpath_expr.find(response_data)]
+
+            if not matches:
+                # Path not found - return empty based on action
+                return [] if is_array_action else None
+
+            # Return extracted data with primitive wrapping
+            if is_array_action:
+                # For array actions, return the array (or list of matches)
+                result = matches[0] if len(matches) == 1 else matches
+            else:
+                # For single record actions, return first match
+                result = matches[0]
+
+            return self._wrap_primitives(result)
+
+        except Exception as e:
+            logging.warning(f"Failed to apply record extractor '{extractor}': {e}. Returning original response.")
+            return self._wrap_primitives(response_data)
+
+    def _extract_metadata(
+        self,
+        response_data: dict[str, Any],
+        response_headers: dict[str, str],
+        endpoint: EndpointDefinition,
+    ) -> dict[str, Any] | None:
+        """Extract metadata from response using meta extractor.
+
+        Each field in meta_extractor dict is independently extracted using JSONPath
+        for body extraction, or special prefixes for header extraction:
+        - @link.{rel}: Extract URL from RFC 5988 Link header by rel type
+        - @header.{name}: Extract raw header value by header name
+        - Otherwise: JSONPath expression for body extraction
+
+        Missing or invalid paths result in None for that field (no crash).
+
+        Args:
+            response_data: Full API response (before record extraction)
+            response_headers: HTTP response headers
+            endpoint: Endpoint with optional meta extractor configuration
+
+        Returns:
+            - Dict of extracted metadata if extractor configured
+            - None if no extractor configured
+            - Dict with None values for failed extractions
+
+        Example:
+            meta_extractor = {
+                "pagination": "$.records",
+                "request_id": "$.requestId",
+                "next_page_url": "@link.next",
+                "rate_limit": "@header.X-RateLimit-Remaining"
+            }
+            Returns: {
+                "pagination": {"cursor": "abc", "total": 100},
+                "request_id": "xyz123",
+                "next_page_url": "https://api.example.com/data?cursor=abc",
+                "rate_limit": "99"
+            }
+        """
+        # Check if endpoint has meta extractor
+        if endpoint.meta_extractor is None:
+            return None
+
+        extracted_meta: dict[str, Any] = {}
+
+        # Extract each field independently
+        for field_name, extractor_expr in endpoint.meta_extractor.items():
+            try:
+                if extractor_expr.startswith("@link."):
+                    # RFC 5988 Link header extraction
+                    rel = extractor_expr[6:]
+                    extracted_meta[field_name] = self._extract_link_url(response_headers, rel)
+                elif extractor_expr.startswith("@header."):
+                    # Raw header value extraction (case-insensitive lookup)
+                    header_name = extractor_expr[8:]
+                    extracted_meta[field_name] = self._get_header_value(response_headers, header_name)
+                else:
+                    # JSONPath body extraction
+                    jsonpath_expr = parse_jsonpath(extractor_expr)
+                    matches = [match.value for match in jsonpath_expr.find(response_data)]
+
+                    if matches:
+                        # Return first match (most common case)
+                        extracted_meta[field_name] = matches[0]
+                    else:
+                        # Path not found - set to None
+                        extracted_meta[field_name] = None
+
+            except Exception as e:
+                # Log error but continue with other fields
+                logging.warning(f"Failed to apply meta extractor for field '{field_name}' with expression '{extractor_expr}': {e}. Setting to None.")
+                extracted_meta[field_name] = None
+
+        return extracted_meta
+
+    @staticmethod
+    def _extract_link_url(headers: dict[str, str], rel: str) -> str | None:
+        """Extract URL from RFC 5988 Link header by rel type.
+
+        Parses Link header format: <url>; param1="value1"; rel="next"; param2="value2"
+
+        Supports:
+        - Multiple parameters per link in any order
+        - Both quoted and unquoted rel values
+        - Multiple links separated by commas
+
+        Args:
+            headers: Response headers dict
+            rel: The rel type to extract (e.g., "next", "prev", "first", "last")
+
+        Returns:
+            The URL for the specified rel type, or None if not found
+        """
+        link_header = headers.get("Link") or headers.get("link", "")
+        if not link_header:
+            return None
+
+        for link_segment in re.split(r",(?=\s*<)", link_header):
+            link_segment = link_segment.strip()
+
+            url_match = re.match(r"<([^>]+)>", link_segment)
+            if not url_match:
+                continue
+
+            url = url_match.group(1)
+            params_str = link_segment[url_match.end() :]
+
+            rel_match = re.search(r';\s*rel="?([^";,]+)"?', params_str, re.IGNORECASE)
+            if rel_match and rel_match.group(1).strip() == rel:
+                return url
+
+        return None
+
+    @staticmethod
+    def _get_header_value(headers: dict[str, str], header_name: str) -> str | None:
+        """Get header value with case-insensitive lookup.
+
+        Args:
+            headers: Response headers dict
+            header_name: Header name to look up
+
+        Returns:
+            Header value or None if not found
+        """
+        # Try exact match first
+        if header_name in headers:
+            return headers[header_name]
+
+        # Case-insensitive lookup
+        header_name_lower = header_name.lower()
+        for key, value in headers.items():
+            if key.lower() == header_name_lower:
+                return value
+
+        return None
+
+    def _validate_required_body_fields(self, endpoint: Any, params: dict[str, Any], action: Action, entity: str) -> None:
+        """Validate that required body fields are present for CREATE/UPDATE operations.
+
+        Args:
+            endpoint: Endpoint definition
+            params: Parameters provided
+            action: Operation action
+            entity: Entity name
+
+        Raises:
+            MissingParameterError: If required body fields are missing
+        """
+        # Only validate for operations that typically have required body fields
+        if action not in (Action.CREATE, Action.UPDATE):
+            return
+
+        # Get the request schema to find truly required fields
+        request_schema = endpoint.request_schema
+        if not request_schema:
+            return
+
+        # 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())}"
+            )
+
+    async def close(self):
+        """Close async HTTP client and logger."""
+        self.tracker.track_session_end()
+        await self.http_client.close()
+        self.logger.close()
+
+    async def __aenter__(self):
+        """Async context manager entry."""
+        return self
+
+    async def __aexit__(self, exc_type, exc_val, exc_tb):
+        """Async context manager exit."""
+        await self.close()
+
+
+# =============================================================================
+# Operation Handlers
+# =============================================================================
+
+
+class _StandardOperationHandler:
+    """Handler for standard REST operations (GET, LIST, CREATE, UPDATE, DELETE, API_SEARCH, AUTHORIZE)."""
+
+    def __init__(self, context: _OperationContext):
+        self.ctx = context
+
+    def can_handle(self, action: Action) -> bool:
+        """Check if this handler can handle the given action."""
+        return action in {
+            Action.GET,
+            Action.LIST,
+            Action.CREATE,
+            Action.UPDATE,
+            Action.DELETE,
+            Action.API_SEARCH,
+            Action.AUTHORIZE,
+        }
+
+    async def execute_operation(self, entity: str, action: Action, params: dict[str, Any]) -> StandardExecuteResult:
+        """Execute standard REST operation with full telemetry and error handling."""
+        tracer = trace.get_tracer("airbyte.connector-sdk.executor.local")
+
+        with tracer.start_as_current_span("airbyte.local_executor.execute_operation") as span:
+            # Add span attributes
+            span.set_attribute("connector.name", self.ctx.executor.model.name)
+            span.set_attribute("connector.entity", entity)
+            span.set_attribute("connector.action", action.value)
+            if params:
+                span.set_attribute("connector.param_keys", list(params.keys()))
+
+            # Increment operation counter
+            self.ctx.session.increment_operations()
+
+            # Track operation timing and status
+            start_time = time.time()
+            error_type = None
+            status_code = None
+
+            try:
+                # O(1) entity lookup
+                entity_def = self.ctx.entity_index.get(entity)
+                if not entity_def:
+                    available_entities = list(self.ctx.entity_index.keys())
+                    raise EntityNotFoundError(f"Entity '{entity}' not found in connector. Available entities: {available_entities}")
+
+                # Check if action is supported
+                if action not in entity_def.actions:
+                    supported_actions = [a.value for a in entity_def.actions]
+                    raise ActionNotSupportedError(
+                        f"Action '{action.value}' not supported for entity '{entity}'. Supported actions: {supported_actions}"
+                    )
+
+                # O(1) operation lookup
+                endpoint = self.ctx.operation_index.get((entity, action))
+                if not endpoint:
+                    raise ExecutorError(f"No endpoint defined for {entity}.{action.value}. This is a configuration error.")
+
+                # Validate required body fields for CREATE/UPDATE operations
+                self.ctx.validate_required_body_fields(endpoint, params, action, entity)
+
+                # Build request parameters
+                # Use path_override if available, otherwise use the OpenAPI path
+                actual_path = endpoint.path_override.path if endpoint.path_override else endpoint.path
+                path = self.ctx.build_path(actual_path, params)
+                query_params = self.ctx.extract_query_params(endpoint.query_params, params)
+
+                # Serialize deepObject parameters to bracket notation
+                if endpoint.deep_object_params:
+                    query_params = self.ctx.executor._serialize_deep_object_params(query_params, endpoint.deep_object_params)
+
+                # Build request body (GraphQL or standard)
+                body = self.ctx.build_request_body(endpoint, params)
+
+                # Determine request format (json/data parameters)
+                request_kwargs = self.ctx.determine_request_format(endpoint, body)
+
+                # Extract header parameters from OpenAPI operation (pass body to add Content-Type)
+                header_params = self.ctx.extract_header_params(endpoint, params, body)
+
+                # Execute async HTTP request
+                response_data, response_headers = await self.ctx.http_client.request(
+                    method=endpoint.method,
+                    path=path,
+                    params=query_params if query_params else None,
+                    json=request_kwargs.get("json"),
+                    data=request_kwargs.get("data"),
+                    headers=header_params if header_params else None,
+                )
+
+                # Extract metadata from original response (before record extraction)
+                metadata = self.ctx.executor._extract_metadata(response_data, response_headers, endpoint)
+
+                # Extract records if extractor configured
+                response = self.ctx.extract_records(response_data, endpoint)
+
+                # Assume success with 200 status code if no exception raised
+                status_code = 200
+
+                # Mark span as successful
+                span.set_attribute("connector.success", True)
+                span.set_attribute("http.status_code", status_code)
+
+                # Return StandardExecuteResult with data and metadata
+                return StandardExecuteResult(data=response, metadata=metadata)
+
+            except (EntityNotFoundError, ActionNotSupportedError) as e:
+                # Validation errors - record in span
+                error_type = type(e).__name__
+                span.set_attribute("connector.success", False)
+                span.set_attribute("connector.error_type", error_type)
+                span.record_exception(e)
+                raise
+
+            except Exception as e:
+                # Capture error details
+                error_type = type(e).__name__
+
+                # Try to get status code from HTTP errors
+                if hasattr(e, "response") and hasattr(e.response, "status_code"):
+                    status_code = e.response.status_code
+                    span.set_attribute("http.status_code", status_code)
+
+                span.set_attribute("connector.success", False)
+                span.set_attribute("connector.error_type", error_type)
+                span.record_exception(e)
+                raise
+
+            finally:
+                # Always track operation (success or failure)
+                timing_ms = (time.time() - start_time) * 1000
+                self.ctx.tracker.track_operation(
+                    entity=entity,
+                    action=action.value if isinstance(action, Action) else action,
+                    status_code=status_code,
+                    timing_ms=timing_ms,
+                    error_type=error_type,
+                )
+
+
+class _DownloadOperationHandler:
+    """Handler for download operations.
+
+    Supports two modes:
+    - Two-step (with x-airbyte-file-url): metadata request → extract URL → stream file
+    - One-step (without x-airbyte-file-url): stream file directly from endpoint
+    """
+
+    def __init__(self, context: _OperationContext):
+        self.ctx = context
+
+    def can_handle(self, action: Action) -> bool:
+        """Check if this handler can handle the given action."""
+        return action == Action.DOWNLOAD
+
+    async def execute_operation(self, entity: str, action: Action, params: dict[str, Any]) -> AsyncIterator[bytes]:
+        """Execute download operation (one-step or two-step) with full telemetry."""
+        tracer = trace.get_tracer("airbyte.connector-sdk.executor.local")
+
+        with tracer.start_as_current_span("airbyte.local_executor.execute_operation") as span:
+            # Add span attributes
+            span.set_attribute("connector.name", self.ctx.executor.model.name)
+            span.set_attribute("connector.entity", entity)
+            span.set_attribute("connector.action", action.value)
+            if params:
+                span.set_attribute("connector.param_keys", list(params.keys()))
+
+            # Increment operation counter
+            self.ctx.session.increment_operations()
+
+            # Track operation timing and status
+            start_time = time.time()
+            error_type = None
+            status_code = None
+
+            try:
+                # Look up entity
+                entity_def = self.ctx.entity_index.get(entity)
+                if not entity_def:
+                    raise EntityNotFoundError(f"Entity '{entity}' not found in connector. Available entities: {list(self.ctx.entity_index.keys())}")
+
+                # Look up operation
+                operation = self.ctx.operation_index.get((entity, action))
+                if not operation:
+                    raise ActionNotSupportedError(
+                        f"Action '{action.value}' not supported for entity '{entity}'. Supported actions: {[a.value for a in entity_def.actions]}"
+                    )
+
+                # Common setup for both download modes
+                actual_path = operation.path_override.path if operation.path_override else operation.path
+                path = self.ctx.build_path(actual_path, params)
+                query_params = self.ctx.extract_query_params(operation.query_params, params)
+
+                # Serialize deepObject parameters to bracket notation
+                if operation.deep_object_params:
+                    query_params = self.ctx.executor._serialize_deep_object_params(query_params, operation.deep_object_params)
+
+                # Prepare headers (with optional Range support)
+                range_header = params.get("range_header")
+                headers = {"Accept": "*/*"}
+                if range_header is not None:
+                    headers["Range"] = range_header
+
+                # Check download mode: two-step (with file_field) or one-step (without)
+                file_field = operation.file_field
+
+                if file_field:
+                    # Substitute template variables in file_field (e.g., "attachments[{index}].url")
+                    file_field = LocalExecutor._substitute_file_field_params(file_field, params)
+
+                if file_field:
+                    # Two-step download: metadata → extract URL → stream file
+                    # Step 1: Get metadata (standard request)
+                    request_body = self.ctx.build_request_body(
+                        endpoint=operation,
+                        params=params,
+                    )
+                    request_format = self.ctx.determine_request_format(operation, request_body)
+                    self.ctx.validate_required_body_fields(operation, params, action, entity)
+
+                    metadata_response, _ = await self.ctx.http_client.request(
+                        method=operation.method,
+                        path=path,
+                        params=query_params,
+                        **request_format,
+                    )
+
+                    # Step 2: Extract file URL from metadata
+                    file_url = LocalExecutor._extract_download_url(
+                        response=metadata_response,
+                        file_field=file_field,
+                        entity=entity,
+                    )
+
+                    # Step 3: Stream file from extracted URL
+                    file_response, _ = await self.ctx.http_client.request(
+                        method="GET",
+                        path=file_url,
+                        headers=headers,
+                        stream=True,
+                    )
+                else:
+                    # One-step direct download: stream file directly from endpoint
+                    file_response, _ = await self.ctx.http_client.request(
+                        method=operation.method,
+                        path=path,
+                        params=query_params,
+                        headers=headers,
+                        stream=True,
+                    )
+
+                # Assume success once we start streaming
+                status_code = 200
+
+                # Mark span as successful
+                span.set_attribute("connector.success", True)
+                span.set_attribute("http.status_code", status_code)
+
+                # Stream file chunks
+                default_chunk_size = 8 * 1024 * 1024  # 8 MB
+                async for chunk in file_response.original_response.aiter_bytes(chunk_size=default_chunk_size):
+                    # Log each chunk for cassette recording
+                    self.ctx.logger.log_chunk_fetch(chunk)
+                    yield chunk
+
+            except (EntityNotFoundError, ActionNotSupportedError) as e:
+                # Validation errors - record in span
+                error_type = type(e).__name__
+                span.set_attribute("connector.success", False)
+                span.set_attribute("connector.error_type", error_type)
+                span.record_exception(e)
+
+                # Track the failed operation before re-raising
+                timing_ms = (time.time() - start_time) * 1000
+                self.ctx.tracker.track_operation(
+                    entity=entity,
+                    action=action.value,
+                    status_code=status_code,
+                    timing_ms=timing_ms,
+                    error_type=error_type,
+                )
+                raise
+
+            except Exception as e:
+                # Capture error details
+                error_type = type(e).__name__
+
+                # Try to get status code from HTTP errors
+                if hasattr(e, "response") and hasattr(e.response, "status_code"):
+                    status_code = e.response.status_code
+                    span.set_attribute("http.status_code", status_code)
+
+                span.set_attribute("connector.success", False)
+                span.set_attribute("connector.error_type", error_type)
+                span.record_exception(e)
+
+                # Track the failed operation before re-raising
+                timing_ms = (time.time() - start_time) * 1000
+                self.ctx.tracker.track_operation(
+                    entity=entity,
+                    action=action.value,
+                    status_code=status_code,
+                    timing_ms=timing_ms,
+                    error_type=error_type,
+                )
+                raise
+
+            finally:
+                # Track successful operation (if no exception was raised)
+                # Note: For generators, this runs after all chunks are yielded
+                if error_type is None:
+                    timing_ms = (time.time() - start_time) * 1000
+                    self.ctx.tracker.track_operation(
+                        entity=entity,
+                        action=action.value,
+                        status_code=status_code,
+                        timing_ms=timing_ms,
+                        error_type=None,
+                    )