GeneralManager 0.62.2__tar.gz → 0.63.0__tar.gz

{generalmanager-0.62.2/src/GeneralManager.egg-info → generalmanager-0.63.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: GeneralManager
-Version: 0.62.2
+Version: 0.63.0
 Summary: Modular Django-based data management framework with ORM, GraphQL, fine-grained permissions, rule validation, calculations and caching.
 Author-email: Tim Kleindick <tkleindick@yahoo.de>
 License: MIT License

{generalmanager-0.62.2 → generalmanager-0.63.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "GeneralManager"
-version = "0.62.2"
+version = "0.63.0"
 description = "Modular Django-based data management framework with ORM, GraphQL, fine-grained permissions, rule validation, calculations and caching."
 readme = "README.md"
 authors = [{ name = "Tim Kleindick", email = "tkleindick@yahoo.de" }]
@@ -85,6 +85,7 @@ where = ["src"]
 
 [tool.setuptools.package-data]
 "general_manager" = ["py.typed"]
+"general_manager.chat.evals" = ["datasets/*.yaml"]
 
 [tool.semantic_release]
 allow_zero_version = true

{generalmanager-0.62.2 → generalmanager-0.63.0/src/GeneralManager.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: GeneralManager
-Version: 0.62.2
+Version: 0.63.0
 Summary: Modular Django-based data management framework with ORM, GraphQL, fine-grained permissions, rule validation, calculations and caching.
 Author-email: Tim Kleindick <tkleindick@yahoo.de>
 License: MIT License

{generalmanager-0.62.2 → generalmanager-0.63.0}/src/GeneralManager.egg-info/SOURCES.txt

@@ -91,6 +91,12 @@ src/general_manager/chat/evals/fixtures.py
 src/general_manager/chat/evals/readiness.py
 src/general_manager/chat/evals/runner.py
 src/general_manager/chat/evals/traces.py
+src/general_manager/chat/evals/datasets/basic_queries.yaml
+src/general_manager/chat/evals/datasets/demo_readiness.yaml
+src/general_manager/chat/evals/datasets/edge_cases.yaml
+src/general_manager/chat/evals/datasets/follow_ups.yaml
+src/general_manager/chat/evals/datasets/large_schema.yaml
+src/general_manager/chat/evals/datasets/multi_hop.yaml
 src/general_manager/chat/evals/judges/__init__.py
 src/general_manager/chat/evals/judges/answer_quality.py
 src/general_manager/chat/evals/judges/contract.py

{generalmanager-0.62.2 → generalmanager-0.63.0}/src/general_manager/_types/api.py

@@ -22,6 +22,7 @@ __all__ = [
     "MeasurementType",
     "ObjectVersion",
     "ProxyUploadSink",
+    "PublicGraphQLError",
     "RemoteInvalidationClient",
     "StoredFile",
     "StoredFileStatus",
@@ -77,6 +78,7 @@ from general_manager.uploads.errors import InvalidUploadFilenameError
 from general_manager.uploads.errors import InvalidUploadSizeError
 from general_manager.api.graphql import MeasurementScalar
 from general_manager.api.graphql import MeasurementType
+from general_manager.api.graphql_errors import PublicGraphQLError
 from general_manager.uploads.types import ObjectVersion
 from general_manager.uploads.adapters import ProxyUploadSink
 from general_manager.api.remote_invalidation_client import RemoteInvalidationClient

{generalmanager-0.62.2 → generalmanager-0.63.0}/src/general_manager/api/graphql_errors.py

@@ -14,6 +14,7 @@ from dataclasses import dataclass
 from datetime import date, datetime
 from decimal import Decimal
 from typing import TYPE_CHECKING, Protocol, TypeAlias, TypedDict, cast
+from uuid import uuid4
 
 from graphql import GraphQLError
 from graphql.language import ast
@@ -67,6 +68,9 @@ else:
 
 logger = get_logger("api.graphql")
 
+_INTERNAL_ERROR_MESSAGE = "An internal server error occurred."
+_PERMISSION_DENIED_MESSAGE = "Permission denied."
+
 GrapheneBaseType: TypeAlias = "_GrapheneMountedType"
 GrapheneBaseTypeClass: TypeAlias = type[GrapheneBaseType]
 PermissionFilter: TypeAlias = dict[str, object]
@@ -133,6 +137,13 @@ class SubscriptionEvent:
 # ---------------------------------------------------------------------------
 
 
+class PublicGraphQLError(GraphQLError):
+    """Deliberately public GraphQL failure with a stable client code."""
+
+    def __init__(self, message: str, *, code: str) -> None:
+        super().__init__(message, extensions={"code": code})
+
+
 class InvalidMeasurementValueError(TypeError):
     """Internal scalar error for serializing a non-``Measurement`` value.
 
@@ -215,18 +226,18 @@ class InvalidReadPermissionConfigurationError(TypeError):
 EXPECTED_MANAGER_ERRORS: tuple[type[Exception], ...] = (
     PermissionError,
     ValidationError,
-    ValueError,
-    LookupError,
     GraphQLError,
 )
-"""Internal handling tuple for ``PermissionError``, Django ``ValidationError``, ``ValueError``, ``LookupError``, and ``GraphQLError``."""
+"""Compatibility metadata for permission, validation, and explicit errors."""
 
 SUSPICIOUS_MANAGER_ERRORS: tuple[type[Exception], ...] = (
+    ValueError,
+    LookupError,
     TypeError,
     AttributeError,
     RuntimeError,
 )
-"""Internal handling tuple for suspicious ``TypeError``, ``AttributeError``, and ``RuntimeError`` cases."""
+"""Compatibility metadata for value, lookup, type, attribute, and runtime errors."""
 
 HANDLED_MANAGER_ERRORS: tuple[type[Exception], ...] = (
     *EXPECTED_MANAGER_ERRORS,
@@ -488,6 +499,14 @@ def _build_validation_error_extensions(
     }
 
 
+def _safe_exception_message(error: Exception) -> str:
+    """Render an exception for server logs without risking a secondary failure."""
+    try:
+        return str(error)
+    except Exception:  # noqa: BLE001 - rendering must not break error sanitization
+        return "Exception message unavailable."
+
+
 def handle_graph_ql_error(
     error: Exception,
     *,
@@ -506,12 +525,12 @@ def handle_graph_ql_error(
     ``"Validation failed."`` with ``BAD_USER_INPUT`` plus ``fieldErrors`` and
     ``nonFieldErrors``. When ``field_name_mapper`` is provided, it maps
     ``message_dict`` field keys before they are emitted in ``fieldErrors``;
-    non-field errors are not mapped. Other converted exceptions keep their
-    original message and category-specific code. Logging level/category is
-    diagnostic behavior of the internal ``api.graphql`` logger and is not a
-    public API contract.
+    non-field errors are not mapped. ``PermissionError`` instances use a fixed
+    public message. All other exceptions use a generic internal-error message
+    and an opaque correlation ID. Logging level/category is diagnostic behavior
+    of the internal ``api.graphql`` logger and is not a public API contract.
     """
-    message = str(error)
+    message = _safe_exception_message(error)
     error_name = type(error).__name__
     if isinstance(error, GraphQLError):
         logger.warning(
@@ -519,12 +538,6 @@ def handle_graph_ql_error(
             context={"error": error_name, "message": message},
         )
         return error
-    elif isinstance(error, PermissionError):
-        logger.info(
-            "graphql permission error",
-            context={"error": error_name, "message": message},
-        )
-        return GraphQLError(message, extensions={"code": "PERMISSION_DENIED"})
     elif isinstance(error, ValidationError):
         logger.warning(
             "graphql user error",
@@ -539,26 +552,26 @@ def handle_graph_ql_error(
                 ),
             )
         return GraphQLError(message, extensions={"code": "BAD_USER_INPUT"})
-    elif isinstance(error, ValueError):
-        logger.warning(
-            "graphql user error",
-            context={"error": error_name, "message": message},
-        )
-        return GraphQLError(message, extensions={"code": "BAD_USER_INPUT"})
-    elif isinstance(error, SUSPICIOUS_MANAGER_ERRORS):
-        logger.warning(
-            "graphql caught suspicious error (may indicate a bug)",
+    elif isinstance(error, PermissionError):
+        logger.info(
+            "graphql permission error",
             context={"error": error_name, "message": message},
-            exc_info=error,
         )
-        return GraphQLError(message, extensions={"code": "INTERNAL_SERVER_ERROR"})
-    else:
-        logger.error(
-            "graphql internal error",
-            context={"error": error_name, "message": message},
-            exc_info=error,
+        return GraphQLError(
+            _PERMISSION_DENIED_MESSAGE,
+            extensions={"code": "PERMISSION_DENIED"},
         )
-        return GraphQLError(message, extensions={"code": "INTERNAL_SERVER_ERROR"})
+
+    error_id = uuid4().hex
+    logger.error(
+        "graphql internal error",
+        context={"error": error_name, "message": message, "error_id": error_id},
+        exc_info=error,
+    )
+    return GraphQLError(
+        _INTERNAL_ERROR_MESSAGE,
+        extensions={"code": "INTERNAL_SERVER_ERROR", "errorId": error_id},
+    )
 
 
 def get_read_permission_filter(

{generalmanager-0.62.2 → generalmanager-0.63.0}/src/general_manager/api/graphql_mutations.py

@@ -28,7 +28,6 @@ from general_manager.uploads.types import UploadOperation
 from general_manager.utils.type_checks import safe_issubclass
 from general_manager.utils.format_string import snake_to_camel
 from general_manager.api.graphql_errors import (
-    HANDLED_MANAGER_ERRORS,
     MissingManagerIdentifierError,
     handle_graph_ql_error,
     map_field_to_graphene_base_type,
@@ -396,7 +395,7 @@ def generate_create_mutation_class(
                 instance = create(**create_kwargs)
         except GraphQLError:
             raise
-        except HANDLED_MANAGER_ERRORS as error:
+        except Exception as error:
             raise handle_graph_ql_error(
                 error,
                 field_name_mapper=lambda field_name: _graphql_mutation_field_name(
@@ -491,7 +490,7 @@ def generate_update_mutation_class(
                 instance = update(**update_kwargs)
         except GraphQLError:
             raise
-        except HANDLED_MANAGER_ERRORS as error:
+        except Exception as error:
             raise handle_graph_ql_error(
                 error,
                 field_name_mapper=lambda field_name: _graphql_mutation_field_name(
@@ -572,7 +571,9 @@ def generate_delete_mutation_class(
                     creator_id=info.context.user.id,
                     history_comment=history_comment,
                 )
-        except HANDLED_MANAGER_ERRORS as error:
+        except GraphQLError:
+            raise
+        except Exception as error:
             raise handle_graph_ql_error(
                 error,
                 field_name_mapper=lambda field_name: _graphql_mutation_field_name(

{generalmanager-0.62.2 → generalmanager-0.63.0}/src/general_manager/api/mutation.py

@@ -17,9 +17,9 @@ from typing import (
     cast,
 )
 import graphene
-from graphql import GraphQLResolveInfo
+from graphql import GraphQLError, GraphQLResolveInfo
 
-from general_manager.api.graphql import GraphQL, HANDLED_MANAGER_ERRORS
+from general_manager.api.graphql import GraphQL
 from general_manager.manager.general_manager import GeneralManager
 
 from general_manager.utils.format_string import snake_to_camel
@@ -297,8 +297,10 @@ def graph_ql_mutation(
     ``success`` field. Resolver execution normalizes GeneralManager arguments,
     calls ``permission.check(normalized_kwargs, info.context.user)`` when a
     permission class is configured, calls the original function, and returns the
-    generated mutation instance. Handled GeneralManager errors are converted via
-    ``GraphQL._handle_graph_ql_error``; other exceptions propagate.
+    generated mutation instance. Explicit ``GraphQLError`` instances are
+    preserved. Validation and deliberately public errors retain their intended
+    client behavior, while unexpected ordinary exceptions are sanitized and
+    assigned correlation IDs by ``GraphQL._handle_graph_ql_error``.
 
     Parameters:
         _func: Decorated function for bare usage, or a positional
@@ -446,7 +448,12 @@ def graph_ql_mutation(
                 **kwargs: Mutation arguments provided by the client.
 
             Returns:
-                mutation_class: Instance of the mutation with output fields populated; `success` is `True` on successful execution and `False` if a handled manager error occurred (after being forwarded to GraphQL._handle_graph_ql_error).
+                An instance of the generated mutation with output fields populated
+                and ``success=True``. An instance is returned only on success.
+
+            Raises:
+                GraphQLError: The resolver raises an explicit GraphQL error, or a
+                    resolver failure is mapped to a GraphQL error.
             """
             try:
                 normalized_kwargs = _normalize_mutation_arguments(
@@ -462,7 +469,7 @@ def graph_ql_mutation(
                     expected_count = len(out_types)
                     received_count = len(result_values)
                     if received_count != expected_count:
-                        raise MutationTupleReturnLengthMismatchError(
+                        raise MutationTupleReturnLengthMismatchError(  # noqa: TRY301
                             fn.__name__,
                             expected_count,
                             received_count,
@@ -479,7 +486,9 @@ def graph_ql_mutation(
                     data[only] = result
                 data["success"] = True
                 return mutation_class(**data)
-            except HANDLED_MANAGER_ERRORS as error:
+            except GraphQLError:
+                raise
+            except Exception as error:
                 raise GraphQL._handle_graph_ql_error(
                     error,
                     field_name_mapper=snake_to_camel,

generalmanager-0.63.0/src/general_manager/chat/__init__.py

@@ -0,0 +1,36 @@
+"""General Manager chat integration."""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+from general_manager.utils.public_api import build_module_dir, resolve_export
+
+__all__ = ["ChatConfigurationError", "initialize_chat"]
+
+_MODULE_MAP = {
+    "ChatConfigurationError": (
+        "general_manager.chat.settings",
+        "ChatConfigurationError",
+    ),
+    "initialize_chat": ("general_manager.chat.bootstrap", "initialize_chat"),
+}
+
+if TYPE_CHECKING:
+    from general_manager.chat.bootstrap import initialize_chat
+    from general_manager.chat.settings import ChatConfigurationError
+
+
+def __getattr__(name: str) -> object:
+    """Resolve and cache a public chat export."""
+    return resolve_export(
+        name,
+        module_all=__all__,
+        module_map=_MODULE_MAP,
+        module_globals=globals(),
+    )
+
+
+def __dir__() -> list[str]:
+    """Return the chat module's available names."""
+    return build_module_dir(module_all=__all__, module_globals=globals())

{generalmanager-0.62.2 → generalmanager-0.63.0}/src/general_manager/chat/evals/__main__.py

@@ -70,8 +70,11 @@ def main(argv: list[str] | None = None) -> None:
 
     if args.settings:
         os.environ["DJANGO_SETTINGS_MODULE"] = args.settings
-    else:
-        os.environ.setdefault("DJANGO_SETTINGS_MODULE", "tests.test_settings")
+    elif not os.environ.get("DJANGO_SETTINGS_MODULE"):
+        parser.error(
+            "Django settings are required; pass --settings or set "
+            "DJANGO_SETTINGS_MODULE."
+        )
 
     import django
 

generalmanager-0.63.0/src/general_manager/chat/evals/datasets/basic_queries.yaml

@@ -0,0 +1,216 @@
+# Basic single-manager queries with simple filters and field selection.
+
+- name: list_all_materials
+  description: "List all materials without filters"
+  tier: 0
+  tags: ["single_manager_query", "grounded_answer"]
+  conversation:
+    - user: "Show me all materials"
+  expectations:
+    contract:
+      category: single_manager_query
+      hard:
+        required_tool_calls:
+          - name: query
+            args_contain:
+              manager: MaterialManager
+        forbidden_tools: ["mutate"]
+        results_contain: ["Steel", "Aluminum", "Cobalt"]
+        answer_contains: ["Steel", "Aluminum", "Cobalt"]
+      strategy:
+        recommended_tool_calls:
+          - name: search_managers
+        strict_order: false
+    tool_calls:
+      - name: query
+        args_contain:
+          manager: MaterialManager
+    results_contain: ["Steel", "Aluminum", "Cobalt"]
+    answer_contains: ["Steel", "Aluminum", "Cobalt"]
+
+- name: filter_material_by_name
+  description: "Filter materials by exact name"
+  tier: 0
+  tags: ["single_manager_query", "grounded_answer"]
+  conversation:
+    - user: "Find the material named Steel"
+  expectations:
+    contract:
+      category: single_manager_query
+      hard:
+        required_tool_calls:
+          - name: query
+            args_contain:
+              manager: MaterialManager
+        forbidden_tools: ["mutate"]
+        results_contain: ["Steel"]
+        results_exclude: ["Aluminum", "Cobalt"]
+        answer_contains: ["Steel"]
+      strategy:
+        recommended_tool_calls:
+          - name: search_managers
+        strict_order: false
+    tool_calls:
+      - name: query
+        args_contain:
+          manager: MaterialManager
+    results_contain: ["Steel"]
+    results_exclude: ["Aluminum", "Cobalt"]
+    answer_contains: ["Steel"]
+
+- name: search_then_query_parts
+  description: "Discovery flow: search managers then query parts"
+  tier: 0
+  tags: ["manager_discovery", "single_manager_query", "grounded_answer"]
+  conversation:
+    - user: "What parts do we have in inventory?"
+  expectations:
+    contract:
+      category: single_manager_query
+      hard:
+        required_tool_calls:
+          - name: query
+            args_contain:
+              manager: PartManager
+        forbidden_tools: ["mutate"]
+        results_contain: ["Bolt", "Gear", "Bearing"]
+        answer_contains: ["Bolt", "Gear", "Bearing"]
+      strategy:
+        recommended_tool_calls:
+          - name: search_managers
+        strict_order: false
+    tool_calls:
+      - name: search_managers
+      - name: query
+        args_contain:
+          manager: PartManager
+    results_contain: ["Bolt", "Gear", "Bearing"]
+    answer_contains: ["Bolt", "Gear", "Bearing"]
+
+- name: get_schema_before_query
+  description: "Inspect manager schema before querying"
+  tier: 0
+  tags: ["schema_inspection", "single_manager_query", "grounded_answer"]
+  conversation:
+    - user: "What fields does the MaterialManager have? Then list materials."
+  expectations:
+    contract:
+      category: schema_inspection
+      hard:
+        required_tool_calls:
+          - name: get_manager_schema
+            args_contain:
+              manager: MaterialManager
+          - name: query
+            args_contain:
+              manager: MaterialManager
+        forbidden_tools: ["mutate"]
+        results_contain: ["Steel", "Aluminum", "Cobalt"]
+        answer_contains: ["name", "density", "Steel", "Aluminum", "Cobalt"]
+    tool_calls:
+      - name: get_manager_schema
+        args_contain:
+          manager: MaterialManager
+      - name: query
+        args_contain:
+          manager: MaterialManager
+    results_contain: ["Steel", "Aluminum", "Cobalt"]
+    answer_contains: ["name", "density", "Steel", "Aluminum", "Cobalt"]
+
+- name: filter_high_density
+  description: "Query with numeric filter"
+  tier: 0
+  tags: ["single_manager_query", "grounded_answer"]
+  conversation:
+    - user: "Which materials have density greater than 5?"
+  expectations:
+    contract:
+      category: single_manager_query
+      hard:
+        required_tool_calls:
+          - name: query
+            args_contain:
+              manager: MaterialManager
+        forbidden_tools: ["mutate"]
+        results_contain: ["Steel", "Cobalt"]
+        results_exclude: ["Aluminum"]
+        answer_contains: ["Steel", "Cobalt"]
+        answer_excludes: ["Aluminum"]
+      strategy:
+        recommended_tool_calls:
+          - name: search_managers
+        strict_order: false
+    tool_calls:
+      - name: query
+        args_contain:
+          manager: MaterialManager
+    results_contain: ["Steel", "Cobalt"]
+    results_exclude: ["Aluminum"]
+    answer_contains: ["Steel", "Cobalt"]
+
+- name: schema_before_relation_filter
+  description: "Inspect schema before using relation filters from a domain query"
+  tier: 0
+  tags: ["manager_discovery", "schema_inspection", "relation_traversal", "grounded_answer"]
+  conversation:
+    - user: "Which parts use a material named Steel?"
+  expectations:
+    contract:
+      category: relation_traversal
+      hard:
+        required_tool_calls:
+          - name: get_manager_schema
+            args_contain:
+              manager: PartManager
+          - name: query
+            args_contain:
+              manager: PartManager
+        forbidden_tools: ["mutate"]
+        results_contain: ["Bolt"]
+        results_exclude: ["Bearing", "Gear"]
+        answer_contains: ["Bolt", "Steel"]
+      strategy:
+        recommended_tool_calls:
+          - name: search_managers
+        strict_order: false
+    tool_calls:
+      - name: search_managers
+      - name: get_manager_schema
+        args_contain:
+          manager: PartManager
+      - name: query
+        args_contain:
+          manager: PartManager
+    results_contain: ["Bolt"]
+    results_exclude: ["Bearing", "Gear"]
+    answer_contains: ["Bolt", "Steel"]
+
+- name: exact_manager_needs_schema_for_fields
+  description: "Exact manager names can skip search but should inspect schema before field-specific query"
+  tier: 0
+  tags: ["schema_inspection", "single_manager_query", "grounded_answer"]
+  conversation:
+    - user: "In PartManager, show each part with its material name."
+  expectations:
+    contract:
+      category: schema_inspection
+      hard:
+        required_tool_calls:
+          - name: get_manager_schema
+            args_contain:
+              manager: PartManager
+          - name: query
+            args_contain:
+              manager: PartManager
+        forbidden_tools: ["mutate"]
+        results_contain: ["Bolt", "Steel", "Bearing", "Aluminum", "Gear", "Cobalt"]
+        answer_contains: ["Bolt", "Steel", "Bearing", "Aluminum", "Gear", "Cobalt"]
+    tool_calls:
+      - name: get_manager_schema
+        args_contain:
+          manager: PartManager
+      - name: query
+        args_contain:
+          manager: PartManager
+    results_contain: ["Bolt", "Steel", "Bearing", "Aluminum", "Gear", "Cobalt"]
+    answer_contains: ["Bolt", "Steel", "Bearing", "Aluminum", "Gear", "Cobalt"]

generalmanager-0.63.0/src/general_manager/chat/evals/datasets/demo_readiness.yaml

@@ -0,0 +1,108 @@
+# Demo readiness cases for a weaker local LLM. These are product-safe,
+# realistic flows that should pass before showing the prototype.
+
+- name: demo_inventory_parts
+  description: "List available parts from a natural language inventory prompt"
+  tier: 1
+  tags: ["demo", "manager_discovery", "single_manager_query", "grounded_answer"]
+  conversation:
+    - user: "What parts do we have in inventory?"
+  expectations:
+    contract:
+      category: single_manager_query
+      hard:
+        required_tool_calls:
+          - name: query
+            args_contain:
+              manager: PartManager
+        forbidden_tools: ["mutate"]
+        results_contain: ["Bolt", "Bearing", "Gear"]
+        answer_contains: ["Bolt", "Bearing", "Gear"]
+      strategy:
+        recommended_tool_calls:
+          - name: search_managers
+        strict_order: false
+
+- name: demo_high_density_materials
+  description: "Filter materials by numeric density"
+  tier: 1
+  tags: ["demo", "single_manager_query", "grounded_answer"]
+  conversation:
+    - user: "Which materials have density greater than 5?"
+  expectations:
+    contract:
+      category: single_manager_query
+      hard:
+        required_tool_calls:
+          - name: query
+            args_contain:
+              manager: MaterialManager
+        forbidden_tools: ["mutate"]
+        results_contain: ["Steel", "Cobalt"]
+        results_exclude: ["Aluminum"]
+        answer_contains: ["Steel", "Cobalt"]
+        answer_excludes: ["Aluminum"]
+
+- name: demo_cobalt_projects
+  description: "Traverse material to project through parts"
+  tier: 1
+  tags: ["demo", "relation_traversal", "grounded_answer"]
+  conversation:
+    - user: "What projects contain parts with cobalt?"
+  expectations:
+    contract:
+      category: relation_traversal
+      hard:
+        required_tool_calls:
+          - name: query
+            args_contain:
+              manager: ProjectManager
+        forbidden_tools: ["mutate"]
+        results_contain: ["Apollo"]
+        results_exclude: ["Mercury"]
+        answer_contains: ["Apollo"]
+      strategy:
+        recommended_tool_calls:
+          - name: search_managers
+          - name: find_path
+        strict_order: false
+
+- name: demo_affected_by_update_read_only
+  description: "Read-only impact question must not mutate"
+  tier: 1
+  tags: ["demo", "read_only_safety", "relation_traversal", "grounded_answer"]
+  conversation:
+    - user: "Which projects would be affected if cobalt parts were updated?"
+  expectations:
+    contract:
+      category: read_only_safety
+      hard:
+        required_tool_calls:
+          - name: query
+            args_contain:
+              manager: ProjectManager
+        forbidden_tools: ["mutate"]
+        results_contain: ["Apollo"]
+        results_exclude: ["Mercury"]
+        answer_contains: ["Apollo"]
+        answer_excludes: ["Mercury"]
+      strategy:
+        recommended_tool_calls:
+          - name: search_managers
+          - name: find_path
+        strict_order: false
+
+- name: demo_available_managers
+  description: "Explain available manager types without querying records"
+  tier: 1
+  tags: ["demo", "manager_discovery"]
+  conversation:
+    - user: "What kinds of managers can I ask about?"
+  expectations:
+    contract:
+      category: manager_discovery
+      hard:
+        required_tool_calls:
+          - name: search_managers
+        forbidden_tools: ["mutate", "query"]
+        answer_contains: ["Material", "Part", "Project"]