Update haiway

Author: KaQuMiQ

docs/getting-started/first-steps.md

@@ -31,11 +31,11 @@ class AppConfig(State):
     max_retries: int = 3
 ```
 
-State instances are immutable—use `.updated()` to create tweaked copies.
+State instances are immutable—use `.updating()` to create tweaked copies.
 
 ```python
 config = AppConfig(environment="staging")
-updated = config.updated(max_retries=5)
+updated = config.updating(max_retries=5)
 ```
 
 ## Model your data

docs/getting-started/index.md

@@ -34,7 +34,7 @@ Throughout the getting-started journey you will assemble:
 ## Core Concepts
 
 1. **State management** – mutable-looking, immutable-under-the-hood `State` classes represent
-    configuration and runtime data. You update them with methods like `State.updated(...)` to keep
+    configuration and runtime data. You update them with methods like `State.updating(...)` to keep
     histories, snapshots, and metrics consistent.
 1. **Context scoping** – `ctx.scope(...)` activates a stack of `State` instances and disposables for
     a logical unit of work, ensuring structured concurrency and clean teardown.

docs/guides/AdvancedState.md

@@ -29,9 +29,9 @@ initial: Mutable = Mutable(
     value=42,
 )
 # update one of the fields by creating a copy
-updated: Mutable = initial.updated(identifier="post")
+updated: Mutable = initial.updating(identifier="post")
 # update initial state once more - this will be another copy
-final: Mutable = initial.updated(value=21)
+final: Mutable = initial.updating(value=21)
 
 print("initial", initial)
 print("updated", updated)

docs/guides/BasicStageUsage.md

@@ -9,7 +9,7 @@ reliable pipelines with confidence.
 
 - **Stage** – an async callable that produces or transforms multimodal content.
 - **StageState** – an immutable snapshot holding context entries and the latest result. Always
-    return `state.updated(...)` instead of mutating in place.
+    return `state.updating(...)` instead of mutating in place.
 - **MultimodalContent** – container for text, images, artifacts, and resources used as model inputs
     or outputs.
 - **`ctx.scope(...)`** – binds providers, disposables, and logging/metrics for the lifetime of your
@@ -189,7 +189,7 @@ async def merge_results(
 ) -> StageState:
     successful = [branch for branch in branches if isinstance(branch, StageState)]
     combined = MultimodalContent.of(*[state.result for state in successful])
-    return successful[0].updated(result=combined)
+    return successful[0].updating(result=combined)
 
 concurrent_stage = Stage.concurrent(
     Stage.completion("Analyze aspect A."),
@@ -292,7 +292,7 @@ from draive import MultimodalContent, StageState, stage
 @stage
 async def custom_processor(*, state: StageState) -> StageState:
     processed = MultimodalContent.of(f"Processed: {state.result.as_string()}")
-    return state.updated(result=processed)
+    return state.updating(result=processed)
 ```
 
 Any async function decorated with `@stage` gains the same API as built-in stages and can be mixed

docs/guides/BasicUsage.md

@@ -78,10 +78,10 @@ async with ctx.scope(  # prepare the new context
     print("RESULT GPT 3.5 | temperature 0.4:", result)
 
     # we can update the configuration to change any parameter for nested context
-    with ctx.updated(
+    with ctx.updating(
         # we are updating the current context value instead of making a new one
         # this allows to preserve other elements of the configuration
-        ctx.state(OpenAIResponsesConfig).updated(
+        ctx.state(OpenAIResponsesConfig).updating(
             model="gpt-5",
         ),
     ):
@@ -135,8 +135,8 @@ async with ctx.scope(  # prepare the context and see the execution metrics repor
         text="Roses are red...",
     )
 
-    with ctx.updated(
-        ctx.state(OpenAIResponsesConfig).updated(
+    with ctx.updating(
+        ctx.state(OpenAIResponsesConfig).updating(
             model="gpt-5",
         ),
     ):

docs/guides/Basics.md

@@ -16,7 +16,7 @@ properties and an appropriate `__init__` function will be generated as well. Add
 types come with built-in validation. Each object is automatically validated during the
 initialization ensuring proper types and values for each field. Both types are also immutable by
 default - you will receive linting and runtime errors when trying to mutate those. To make a
-mutation of an instance of either of those we can use a dedicated `updated` method which makes a
+mutation of an instance of either of those we can use a dedicated `updating` method which makes a
 copy on the fly and validates mutation as well. Let's have a look:
 
 ```python
@@ -32,7 +32,7 @@ basic_state: BasicState = BasicState(
 # )
 
 # prepare an update
-updated_state: BasicState = basic_state.updated(
+updated_state: BasicState = basic_state.updating(
     value=21
 )  # value of `identifier` field won't change
 
@@ -150,7 +150,7 @@ def do_something_contextually() -> None:
 ```
 
 What will be the current state is defined by the context of the execution. We can change it locally
-by using another function from `ctx` called `updated` which allows us to update the state by copying
+by using another function from `ctx` called `updating` which allows us to update the state by copying
 the context and allowing to enter a new scope with it:
 
 ```python
@@ -160,7 +160,7 @@ async with ctx.scope("basics", basic_state):
     do_something_contextually()
 
     # then we can update it locally
-    with ctx.updated(basic_state.updated(identifier="updated")):
+    with ctx.updating(basic_state.updating(identifier="updated")):
         print("Updated:")
         # and access its updated version
         do_something_contextually()

llms.txt

@@ -11,7 +11,7 @@ load_env()
 ```
 
 ## Core primitives
-- `State` (from haiway): immutable runtime config/resources. Update via `.updated(...)`.
+- `State` (from haiway): immutable runtime config/resources. Update via `.updating(...)`.
 - `ctx.scope(label, *state, disposables=(...), observability=...)`: structured async context that activates states and disposables; nests safely.
 - `Disposables`: async resources (e.g., `OpenAI()`, `QdrantClient()`) auto-cleaned on scope exit.
 - `DataModel`: immutable, validated data containers with `.json_schema()` / `.simplified_schema()` and `.to_str()`.
@@ -189,5 +189,5 @@ all_ok = all(r.passed for r in results)
 - Always run inside `ctx.scope`.
 - Keep public APIs strictly typed (no implicit `Any`).
 - Avoid blocking calls; everything I/O is async.
-- Mutate state via `.updated(...)`; states and data models are immutable by design.
+- Mutate state via `.updating(...)`; states and data models are immutable by design.
 - Use `MultimodalContent` for any model input/output that mixes text/media/tools; use `.to_str()` for safe logging (binary redacted).

pyproject.toml

@@ -5,7 +5,7 @@ build-backend = "uv_build"
 [project]
 name = "draive"
 description = "Framework designed to simplify and accelerate the development of LLM-based applications."
-version = "0.98.0"
+version = "0.99.0"
 readme = "README.md"
 maintainers = [
   { name = "Kacper Kaliński", email = "[email protected]" },
@@ -24,7 +24,7 @@ classifiers = [
   "Topic :: Software Development :: Libraries :: Application Frameworks",
 ]
 license = { file = "LICENSE" }
-dependencies = ["numpy~=2.3", "haiway~=0.42.0"]
+dependencies = ["numpy~=2.3", "haiway~=0.43.0"]
 
 [project.urls]
 Homepage = "https://miquido.com"

src/draive/helpers/instruction_refinement.py

@@ -264,7 +264,7 @@ async def tree_initialization(
         )
 
         # Setup initial state
-        return state.updated(
+        return state.updating(
             _RefinementState(
                 root=root_node,
                 nodes={root_node.identifier: root_node},
@@ -324,7 +324,7 @@ async def explore(
             }
 
         # Update the refinement state with the updated nodes
-        refinement_state = refinement_state.updated(nodes=updated_nodes)
+        refinement_state = refinement_state.updating(nodes=updated_nodes)
 
         # Log tree statistics
         total_nodes: int = len(refinement_state.nodes)
@@ -336,7 +336,7 @@ async def explore(
             f"{active_nodes} active, {pruned_count} pruned"
         )
 
-        return state.updated(refinement_state)
+        return state.updating(refinement_state)
 
     async def condition(
         *,
@@ -419,7 +419,7 @@ async def _explore_node[Parameters: DataModel](
         # Evaluate with focused suite
         ctx.log_info(f"Evaluating strategy '{strategy_name}'...")
         focused_evaluation: EvaluatorSuiteResult
-        with ctx.updated(child_node.patched_instructions_repository):
+        with ctx.updating(child_node.patched_instructions_repository):
             focused_evaluation = await evaluator_suite(focused_suite_cases)
 
         # Check for performance drop
@@ -431,7 +431,7 @@ async def _explore_node[Parameters: DataModel](
         pruned: bool = performance_ratio < performance_drop_threshold
 
         # update node with evaluation data
-        children[child_node.identifier] = child_node.updated(
+        children[child_node.identifier] = child_node.updating(
             focused_evaluation=focused_evaluation,
             pruned=pruned,
         )
@@ -656,14 +656,14 @@ async def tree_finalization(
             )
 
             complete_evaluation: EvaluatorSuiteResult
-            with ctx.updated(candidate_node.patched_instructions_repository):
+            with ctx.updating(candidate_node.patched_instructions_repository):
                 complete_evaluation = await evaluator_suite()
 
             # Update node with full eval score
-            updated_node: _RefinementTreeNode = candidate_node.updated(
+            updated_node: _RefinementTreeNode = candidate_node.updating(
                 complete_evaluation=complete_evaluation
             )
-            refinement_state = refinement_state.updated(
+            refinement_state = refinement_state.updating(
                 nodes={
                     **refinement_state.nodes,
                     # update node in tree
@@ -688,7 +688,7 @@ async def tree_finalization(
         )
 
         # Update result with best instructions and updated state
-        return state.updated(
+        return state.updating(
             refinement_state,
             result=MultimodalContent.of(
                 TextContent.of(

src/draive/models/generative.py

@@ -340,7 +340,7 @@ async def _loop(
 
             if not pending_tools:
                 ctx.log_debug("...loop finished...")
-                result: ModelOutput = generated.updated(
+                result: ModelOutput = generated.updating(
                     # replace generated with forced result
                     blocks=(
                         *result_extension,
@@ -384,7 +384,7 @@ async def _loop(
 
             if tools_result:  # tools direct result
                 ctx.log_debug("...loop finished by tools...")
-                result: ModelOutput = generated.updated(
+                result: ModelOutput = generated.updating(
                     # replace generated with forced result
                     blocks=(
                         *result_extension,
@@ -746,27 +746,27 @@ def _decoded(  # noqa: PLR0911
             return result
 
         case "text":
-            return result.updated(blocks=(MultimodalContent.of(*result.content.texts()),))
+            return result.updating(blocks=(MultimodalContent.of(*result.content.texts()),))
 
         case "image":
-            return result.updated(blocks=(MultimodalContent.of(*result.content.images()),))
+            return result.updating(blocks=(MultimodalContent.of(*result.content.images()),))
 
         case "audio":
-            return result.updated(blocks=(MultimodalContent.of(*result.content.audio()),))
+            return result.updating(blocks=(MultimodalContent.of(*result.content.audio()),))
 
         case "video":
-            return result.updated(blocks=(MultimodalContent.of(*result.content.video()),))
+            return result.updating(blocks=(MultimodalContent.of(*result.content.video()),))
 
         case selection if isinstance(selection, Collection) and not isinstance(selection, str):
             selected_parts: tuple[MultimodalContentPart, ...] = tuple(
                 part
                 for part in result.content.parts
                 if _matches_modalities(part, allowed=set(selection))
             )
-            return result.updated(blocks=(MultimodalContent.of(*selected_parts),))
+            return result.updating(blocks=(MultimodalContent.of(*selected_parts),))
 
         case "json":
-            return result.updated(
+            return result.updating(
                 blocks=(
                     MultimodalContent.of(
                         ArtifactContent.of(
@@ -779,7 +779,7 @@ def _decoded(  # noqa: PLR0911
 
         case model:
             assert isinstance(model, type)  # nosec: B101
-            return result.updated(
+            return result.updating(
                 blocks=(
                     MultimodalContent.of(
                         ArtifactContent.of(