tunacode-cli 0.2.1__tar.gz → 0.2.2__tar.gz

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/AGENTS.md

@@ -1,5 +1,5 @@
 # AGENTS.md
-Last Updated: 2026-04-27
+Last Updated: 2026-06-22
 
 ## Repository Orientation
 - This is `tunacode-cli`, a terminal AI coding agent with a Textual UI and tiny-agent tool loop.
@@ -61,7 +61,7 @@ Last Updated: 2026-04-27
   - Allowed import direction: `ui -> core -> tools`, with shared-layer imports from `utils`, `types`, `configuration`, `constants`, `exceptions`.
 - Import ordering test: `tests/architecture/test_import_order.py`.
   - Layer order constant includes shared modules (`configuration, constants, exceptions, prompts, skills, types, utils`) then layered modules (`tools, infrastructure, core, ui`).
-- Public-init constraint: `tests/architecture/test_init_bloat.py`.
+- Public-init constraint: `tests/architecture/test_imports_in_init.py`.
 - Dependency map regeneration pipeline: `scripts/grimp_layers_report.py`, committed to `docs/architecture/dependencies/` by workflow.
 
 ## Documentation Sources to Trust
@@ -98,7 +98,7 @@ Last Updated: 2026-04-27
   - `uv run pytest`
   - `uv run pytest tests/test_dependency_layers.py -v`
   - `uv run pytest tests/architecture/test_import_order.py`
-  - `uv run pytest tests/architecture/test_init_bloat.py`
+  - `uv run pytest tests/architecture/test_imports_in_init.py`
   - `uv run python scripts/run_gates.py`
   - `uv run python scripts/check_agents_freshness.py`
   - `uv run python scripts/generate_structure_tree.py`

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/CHANGELOG.md

@@ -4,7 +4,7 @@ summary: Release history for TunaCode versions and notable changes.
 when_to_read:
   - When preparing a release note
   - When reviewing past changes or version history
-last_updated: "2026-04-27"
+last_updated: "2026-06-22"
 ---
 
 # Changelog
@@ -16,6 +16,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.2.2] - 2026-06-22
+
+### Changed
+- Moved thinking panel state management out of the app layer and addressed follow-up review feedback.
+- Bumped `tiny-agent-os` to `1.2.28` and moved max-iteration enforcement to tinyagent's `should_stop_after_turn` hook so capped tool loops end through the normal turn-completion path.
+- Updated the system prompt to tell the agent how to report progress when configured turn limits stop further tool calls.
+
 ## [0.2.1] - 2026-04-27
 
 ### Added

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tunacode-cli
-Version: 0.2.1
+Version: 0.2.2
 Summary: Your agentic CLI developer.
 Project-URL: Homepage, https://tunacode.xyz/
 Project-URL: Repository, https://github.com/alchemiststudiosDOTai/tunacode
@@ -32,7 +32,7 @@ Requires-Dist: rich<15.0.0,>=14.2.0
 Requires-Dist: ruff>=0.14.0
 Requires-Dist: textual-autocomplete>=4.0.6
 Requires-Dist: textual<5.0.0,>=4.0.0
-Requires-Dist: tiny-agent-os>=1.2.27
+Requires-Dist: tiny-agent-os>=1.2.28
 Requires-Dist: typer>=0.15.0
 Provides-Extra: dev
 Requires-Dist: autoflake>=2.0.0; extra == 'dev'

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/docs/modules/core/core.md

@@ -8,7 +8,7 @@ when_to_read:
   - Debugging agent behavior
   - Modifying the request lifecycle
   - Changing how session state is persisted
-last_updated: "2026-04-04"
+last_updated: "2026-06-21"
 ---
 
 # Core Layer
@@ -28,8 +28,9 @@ The engine. Takes a user message, routes it through a tinyagent `Agent`, handles
 | `main.py` | `RequestOrchestrator` -- the main request lifecycle. `process_request()` is the public entry point. Handles: history coercion, pre-request compaction, streaming event dispatch, abort cleanup, empty-response intervention, context-overflow retry. |
 | `helpers.py` | Pure helpers for `main.py`: history coercion/validation, usage parsing, context-overflow detection, tool-result display helpers, and `_TinyAgentStreamState` (per-stream mutable orchestration state). |
 | `agent_components/__init__.py` | Re-exports from sub-modules. |
-| `agent_components/agent_config.py` | `get_or_create_agent()` -- builds or retrieves a cached tinyagent `Agent`. Configures: system prompt, native tool definitions, model, stream function, API key resolver, compaction transform, and skill prompt injection. `invalidate_agent_cache()` clears both module and session caches after abort/timeout. `_build_tools()` constructs the tool list (bash, discover, read_file, hashline_edit, web_fetch, write_file). `_build_skills_prompt_state()` renders active and available skill blocks, and validation helpers include `_coerce_request_delay()`, `_coerce_global_request_timeout()`, `_compute_agent_version()`. |
+| `agent_components/agent_config.py` | `get_or_create_agent()` -- builds or retrieves a cached tinyagent `Agent`. Configures: system prompt, native tool definitions, model, stream function, API key resolver, compaction transform, tinyagent turn-stop control, and skill prompt injection. `invalidate_agent_cache()` clears both module and session caches after abort/timeout. `_build_tools()` constructs the tool list (bash, discover, read_file, hashline_edit, web_fetch, write_file). `_build_skills_prompt_state()` renders active and available skill blocks, and validation helpers include `_coerce_request_delay()`, `_coerce_global_request_timeout()`, `_compute_agent_version()`. |
 | `agent_components/agent_helpers.py` | Human-readable tool descriptions for UI panels. `create_empty_response_message()` builds the intervention prompt when the model returns nothing. |
+| `agent_components/agent_turn_control.py` | tinyagent host-side turn-control callbacks, including the `settings.max_iterations` `should_stop_after_turn` hook. |
 | `resume/sanitize.py` | Cleans persisted session messages for safe resume (removes dangling tool calls, fixes structural violations). |
 | `resume/sanitize_debug.py` | Debug instrumentation for sanitization. |
 
@@ -165,6 +166,8 @@ Selected skills are treated as active instructions, and their fingerprint is inc
 
 **Agent version hashing:** `_compute_agent_version()` generates a cache key from configuration that affects agent behavior: `max_retries`, `tool_strict_validation`, `request_delay`, `global_request_timeout`, `max_tokens`, and the computed skills prompt fingerprint.
 
+**Turn limit control:** `agent_config.py` wires tinyagent's `should_stop_after_turn` host hook so `settings.max_iterations` ends the tool loop through the normal `TurnEndEvent` -> `AgentEndEvent` path. The stream event handler observes turn-end events but no longer calls `agent.abort()` for the iteration cap.
+
 ## Why
 
 The `RequestOrchestrator` class exists to keep the streaming event loop testable and the callback wiring explicit. Each event type has its own handler method -- no giant switch statement.

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/pyproject.toml

@@ -5,7 +5,7 @@ build-backend = "hatchling.build"
 [project]
 name = "tunacode-cli"
 
-version = "0.2.1"
+version = "0.2.2"
 description = "Your agentic CLI developer."
 keywords = ["cli", "agent", "development", "automation"]
 readme = "README.md"
@@ -28,7 +28,7 @@ dependencies = [
     "typer>=0.15.0",
     "click>=8.1.0,<8.2.0",
     "prompt_toolkit>=3.0.52,<4.0.0",
-    "tiny-agent-os>=1.2.27",
+    "tiny-agent-os>=1.2.28",
     "pygments>=2.19.2,<3.0.0",
     "rich>=14.2.0,<15.0.0",
     "textual>=4.0.0,<5.0.0",

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/src/tunacode/constants.py

@@ -18,7 +18,7 @@ KB = 1024
 MB = KB * 1024
 
 APP_NAME = "TunaCode"
-APP_VERSION = "0.2.1"
+APP_VERSION = "0.2.2"
 
 
 AGENTS_MD = "AGENTS.md"

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/src/tunacode/core/agents/agent_components/agent_config.py

@@ -69,6 +69,7 @@ from .agent_session_config import (
     _compute_agent_version,
     _normalize_session_config,
 )
+from .agent_turn_control import build_should_stop_after_turn as _build_should_stop_after_turn
 
 __all__ = [
     "get_or_create_agent",
@@ -507,6 +508,7 @@ def _build_agent_options(
         session_id=session.session_id,
         get_api_key=_build_api_key_resolver(config.env),
         transform_context=_build_transform_context(state_manager),
+        should_stop_after_turn=_build_should_stop_after_turn(session),
     )
 
 

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/src/tunacode/core/agents/agent_components/agent_streaming.py

@@ -148,16 +148,10 @@ class AgentStreamMixin:
         *,
         agent: Agent,
         state: _TinyAgentStreamState,
-        max_iterations: int,
         baseline_message_count: int,
     ) -> bool:
-        _ = (event_obj, baseline_message_count)
-        state.runtime.iteration_count += 1
-        state.runtime.current_iteration = state.runtime.iteration_count
-        if state.runtime.iteration_count <= max_iterations:
-            return False
-        agent.abort()
-        raise RuntimeError(f"Max iterations exceeded ({max_iterations}); aborted")
+        _ = (event_obj, agent, state, baseline_message_count)
+        return False
 
     async def _handle_stream_message_update(
         self,
@@ -281,7 +275,6 @@ class AgentStreamMixin:
         event: AgentEvent,
         agent: Agent,
         state: _TinyAgentStreamState,
-        max_iterations: int,
         baseline_message_count: int,
     ) -> bool:
         if is_turn_end_event(event):
@@ -289,7 +282,6 @@ class AgentStreamMixin:
                 event,
                 agent=agent,
                 state=state,
-                max_iterations=max_iterations,
                 baseline_message_count=baseline_message_count,
             )
         if isinstance(event, MessageUpdateEvent):
@@ -340,7 +332,6 @@ class AgentStreamMixin:
         self,
         *,
         agent: Agent,
-        max_iterations: int,
         baseline_message_count: int,
     ) -> Agent:
         """Run ``agent.stream``; on success, persistence is via AgentEnd; on failure, via cleanup.
@@ -376,7 +367,6 @@ class AgentStreamMixin:
                     event=event,
                     agent=agent,
                     state=state,
-                    max_iterations=max_iterations,
                     baseline_message_count=baseline_message_count,
                 )
                 if should_stop:

tunacode_cli-0.2.2/src/tunacode/core/agents/agent_components/agent_turn_control.py

@@ -0,0 +1,37 @@
+"""tinyagent host-side turn control callbacks."""
+
+from __future__ import annotations
+
+from collections.abc import Callable
+
+from tinyagent.agent_types import AgentContext, AgentMessage, AssistantMessage, ToolResultMessage
+
+from tunacode.core.logging.manager import get_logger
+from tunacode.core.types.state import SessionStateProtocol
+
+from .agent_session_config import _coerce_max_iterations
+
+
+def build_should_stop_after_turn(
+    session: SessionStateProtocol,
+) -> Callable[
+    [AssistantMessage, list[ToolResultMessage], AgentContext, list[AgentMessage]],
+    bool,
+]:
+    def _should_stop_after_turn(
+        message: AssistantMessage,
+        tool_results: list[ToolResultMessage],
+        context: AgentContext,
+        new_messages: list[AgentMessage],
+    ) -> bool:
+        _ = (message, tool_results, context, new_messages)
+        runtime = session.runtime
+        runtime.iteration_count += 1
+        runtime.current_iteration = runtime.iteration_count
+        max_iterations = _coerce_max_iterations(session)
+        if runtime.iteration_count <= max_iterations:
+            return False
+        get_logger().warning(f"Max iterations exceeded ({max_iterations}); ending agent loop")
+        return True
+
+    return _should_stop_after_turn

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/src/tunacode/core/agents/main.py

@@ -33,7 +33,7 @@ from tunacode.core.logging.manager import get_logger
 from tunacode.core.types.state import StateManagerProtocol
 
 from . import agent_components as ac
-from .agent_components.agent_config import _coerce_global_request_timeout, _coerce_max_iterations
+from .agent_components.agent_config import _coerce_global_request_timeout
 from .agent_components.agent_streaming import AgentStreamMixin
 from .helpers import (
     CONTEXT_OVERFLOW_FAILURE_NOTICE,
@@ -91,7 +91,7 @@ class RequestOrchestrator(AgentStreamMixin):
             logger.lifecycle("Agent cache invalidated after timeout")
 
     async def _run_impl(self) -> Agent:
-        max_iterations = self._initialize_request()
+        self._initialize_request()
         logger = get_logger()
         logger.info("Request started", request_id=self.state_manager.session.runtime.request_id)
 
@@ -139,17 +139,15 @@ class RequestOrchestrator(AgentStreamMixin):
 
         await self._run_stream(
             agent=agent,
-            max_iterations=max_iterations,
             baseline_message_count=baseline_message_count,
         )
         await self._retry_after_context_overflow_if_needed(
             agent=agent,
-            max_iterations=max_iterations,
             pre_request_history=pre_request_history,
         )
         return agent
 
-    def _initialize_request(self) -> int:
+    def _initialize_request(self) -> None:
         request_id = str(uuid.uuid4())[:REQUEST_ID_LENGTH]
         session = self.state_manager.session
         runtime = session.runtime
@@ -160,7 +158,6 @@ class RequestOrchestrator(AgentStreamMixin):
         session.usage.last_call_usage = UsageMetrics()
         if not session.task.original_query:
             session.task.original_query = self.message
-        return _coerce_max_iterations(session)
 
     def _maybe_emit_compaction_notice(self, outcome: CompactionOutcome) -> None:
         if self.notice_callback is None:
@@ -193,7 +190,6 @@ class RequestOrchestrator(AgentStreamMixin):
         self,
         *,
         agent: Agent,
-        max_iterations: int,
         pre_request_history: list[AgentMessage],
     ) -> None:
         error_text = self._agent_error_text(agent)
@@ -212,7 +208,6 @@ class RequestOrchestrator(AgentStreamMixin):
         self.state_manager.session._debug_raw_stream_accum = ""
         await self._run_stream(
             agent=agent,
-            max_iterations=max_iterations,
             baseline_message_count=len(conversation.messages),
         )
 

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/src/tunacode/exceptions.py

@@ -331,4 +331,3 @@ class ToolRetryError(TunaCodeError):
     def __init__(self, message: str, original_error: OriginalError = None):
         super().__init__(message)
         self.original_error = original_error
-

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/src/tunacode/prompts/system_prompt.md

@@ -49,6 +49,7 @@ You MUST reuse paths exactly as returned by discover.
 - If more than 3 independent read-only calls are needed, queue them and launch the next call only after one finishes.
 - Do not batch hashline_edit or write_file calls. Execute file mutations one at a time in deterministic order.
 - Execute tool calls immediately. Do not narrate them.
+- Respect configured turn limits. If the host stops further tool calls after a turn, report completed work, remaining work, and the next required step.
 
 ###Output###
 - No emojis.

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/src/tunacode/ui/app.py

@@ -65,6 +65,8 @@ from tunacode.ui.slopgotchi import (
 )
 from tunacode.ui.streaming import StreamingHandler
 from tunacode.ui.styles import STYLE_PRIMARY, STYLE_SUCCESS, STYLE_WARNING
+from tunacode.ui.thinking_state import ThinkingState
+
 from tunacode.ui.widgets import (
     ChatContainer,
     CommandAutoComplete,
@@ -149,10 +151,9 @@ class TextualReplApp(App[None]):
         self._slopgotchi_timer: Timer | None = None
         self._request_bridge: RequestUiBridge | None = None
         self._delta_flush_timer: Timer | None = None
-
-        self._current_thinking_text: str = ""
-        self._last_thinking_update: float = 0.0
         self._last_editor_keypress_at: float = 0.0
+
+        self._thinking_state = ThinkingState(self)
         self._request_debug = RequestDebugTracer(self)
 
     def compose(self) -> ComposeResult:
@@ -298,7 +299,7 @@ class TextualReplApp(App[None]):
         thinking_callback_ms = 0.0
         if thinking_batch.has_data:
             thinking_started_at = time.monotonic()
-            await self._thinking_callback(thinking_batch.text)
+            await self._thinking_state.callback(thinking_batch.text)
             thinking_callback_ms = (
                 time.monotonic() - thinking_started_at
             ) * self.MILLISECONDS_PER_SECOND
@@ -326,7 +327,7 @@ class TextualReplApp(App[None]):
         if not self._loading_indicator_shown:
             self._request_debug.loading_shown(reason="request_start")
         self._show_loading_indicator()
-        self._clear_thinking_state()
+        self._thinking_state.clear()
         bridge = RequestUiBridge(self)
         self._request_bridge = bridge
         self._start_delta_flush_timer()
@@ -398,7 +399,7 @@ class TextualReplApp(App[None]):
             self._hide_loading_indicator()
             self.query_one("#viewport").remove_class(RICHLOG_CLASS_STREAMING)
             self.streaming.reset()
-            self._finalize_thinking_state_after_request()
+            self._thinking_state.finalize()
             self._update_compaction_status(False)
             response_panel_ms = 0.0
             response_char_count = 0
@@ -765,28 +766,3 @@ class TextualReplApp(App[None]):
         )
         if self._context_panel_visible:
             self._refresh_context_panel()
-
-    def _hide_thinking_output(self) -> None:
-        from tunacode.ui.thinking_state import hide_thinking_output
-
-        hide_thinking_output(self)
-
-    def _clear_thinking_state(self) -> None:
-        from tunacode.ui.thinking_state import clear_thinking_state
-
-        clear_thinking_state(self)
-
-    def _finalize_thinking_state_after_request(self) -> None:
-        from tunacode.ui.thinking_state import finalize_thinking_state_after_request
-
-        finalize_thinking_state_after_request(self)
-
-    def _refresh_thinking_output(self, force: bool = False) -> None:
-        from tunacode.ui.thinking_state import refresh_thinking_output
-
-        refresh_thinking_output(self, force)
-
-    async def _thinking_callback(self, delta: str) -> None:
-        from tunacode.ui.thinking_state import thinking_callback
-
-        await thinking_callback(self, delta)

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/src/tunacode/ui/commands/thoughts.py

@@ -22,9 +22,9 @@ class ThoughtsCommand(Command):
         session.show_thoughts = not session.show_thoughts
 
         if session.show_thoughts:
-            app._refresh_thinking_output(force=True)
+            app._thinking_state.refresh(force=True)
             app.notify("Thought panel: ON")
             return
 
-        app._hide_thinking_output()
+        app._thinking_state.hide()
         app.notify("Thought panel: OFF")

tunacode_cli-0.2.2/src/tunacode/ui/thinking_state.py

@@ -0,0 +1,138 @@
+"""Thinking-panel state management for TunaCode TUI."""
+
+from __future__ import annotations
+
+import time
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from textual.widgets import Static
+
+    from tunacode.ui.app import TextualReplApp
+
+
+THINKING_PANEL_HORIZONTAL_MARGIN = 2
+
+
+class ThinkingState:
+    """Encapsulates live thinking-panel state and rendering logic."""
+
+    def __init__(self, app: TextualReplApp) -> None:
+        self._app = app
+        self._text: str = ""
+        self._last_update: float = 0.0
+
+    @property
+    def _widget(self) -> Static | None:
+        widget = self._app._thinking_panel_widget
+        if widget is None:
+            return None
+        return widget
+
+    def _editor_has_draft(self) -> bool:
+        return bool(self._app.editor.value.strip())
+
+    def _has_recent_editor_keypress(self) -> bool:
+        if not self._editor_has_draft():
+            return False
+        last_keypress_at = self._app._last_editor_keypress_at
+        if last_keypress_at <= 0.0:
+            return False
+        elapsed_ms = (time.monotonic() - last_keypress_at) * self._app.MILLISECONDS_PER_SECOND
+        return elapsed_ms < self._app.THINKING_DEFER_AFTER_KEYPRESS_MS
+
+    def _throttle_ms(self) -> float:
+        if self._editor_has_draft():
+            return self._app.THINKING_THROTTLE_WHILE_DRAFTING_MS
+        return self._app.THINKING_THROTTLE_MS
+
+    def hide(self) -> None:
+        """Hide the live thinking panel without removing the widget."""
+        widget = self._widget
+        if widget is None:
+            return
+        widget.update("")
+        widget.remove_class("active")
+
+    def clear(self) -> None:
+        """Reset thinking buffer and hide the widget."""
+        self._text = ""
+        self._last_update = 0.0
+        self.hide()
+
+    def refresh(self, *, force: bool = False) -> None:
+        """Throttled render of the live thinking panel."""
+        if not self._app.state_manager.session.show_thoughts:
+            return
+        if not self._text:
+            self.hide()
+            return
+
+        widget = self._widget
+        if widget is None:
+            return
+
+        if not force and self._has_recent_editor_keypress():
+            return
+
+        now = time.monotonic()
+        elapsed_ms = (now - self._last_update) * self._app.MILLISECONDS_PER_SECOND
+        if not force and elapsed_ms < self._throttle_ms():
+            return
+
+        from tunacode.ui.renderers.thinking import render_thinking_panel
+
+        self._last_update = now
+        content, meta = render_thinking_panel(
+            self._text,
+            max_lines=self._app.THINKING_MAX_RENDER_LINES,
+            max_chars=self._app.THINKING_MAX_RENDER_CHARS,
+        )
+        widget.border_title = meta.border_title
+        widget.border_subtitle = meta.border_subtitle
+        widget.update(content)
+        widget.add_class("active")
+
+    def _final_panel_width(self) -> int:
+        width_candidates = [
+            self._app.chat_container.content_region.width,
+            self._app.chat_container.size.width,
+            self._app.size.width,
+        ]
+        content_width = next((width for width in width_candidates if width > 0), 1)
+        return max(1, content_width - THINKING_PANEL_HORIZONTAL_MARGIN)
+
+    def finalize(self) -> None:
+        """After a request completes, persist a final thinking panel into chat history."""
+        if not self._app.state_manager.session.show_thoughts:
+            self.clear()
+            return
+        if not self._text:
+            self.hide()
+            return
+
+        from tunacode.ui.renderers.thinking import render_thinking_panel
+
+        content, meta = render_thinking_panel(
+            self._text,
+            max_lines=self._app.THINKING_MAX_RENDER_LINES,
+            max_chars=self._app.THINKING_MAX_RENDER_CHARS,
+        )
+        self._app.chat_container.write(
+            content,
+            width=self._final_panel_width(),
+            panel_meta=meta,
+        )
+        self.clear()
+
+    async def callback(self, delta: str) -> None:
+        """Accumulate thinking text and refresh the panel."""
+        self._text += delta
+        overflow = len(self._text) - self._app.THINKING_BUFFER_CHAR_LIMIT
+        if overflow > 0:
+            self._text = self._text[overflow:]
+
+        if not self._app.state_manager.session.show_thoughts:
+            return
+
+        self.refresh()

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/src/tunacode/ui/widgets/chat.py

@@ -276,6 +276,7 @@ class ChatContainer(VerticalScroll):
         renderable: RenderableType | tuple[RenderableType, PanelMeta],
         *,
         expand: bool = False,
+        width: int | None = None,
         panel_meta: PanelMeta | None = None,
     ) -> CopyOnSelectStatic:
         """Append a renderable to the chat container.
@@ -285,6 +286,7 @@ class ChatContainer(VerticalScroll):
         Args:
             renderable: Rich renderable or ``(RenderableType, PanelMeta)`` tuple.
             expand: If True, widget expands to fill available width.
+            width: Optional explicit widget width in cells.
             panel_meta: Optional panel metadata for CSS-styled borders.
 
         Returns:
@@ -294,7 +296,9 @@ class ChatContainer(VerticalScroll):
 
         widget = CopyOnSelectStatic(panel_content)
         widget.add_class("chat-message")
-        if expand:
+        if width is not None:
+            widget.styles.width = max(1, width)
+        elif expand:
             widget.add_class("expand")
 
         if panel_meta is not None:

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/tests/unit/core/test_tool_concurrency_limit.py

@@ -4,14 +4,17 @@ import asyncio
 
 import pytest
 from tinyagent.agent_types import (
+    AgentContext,
     AgentTool,
     AgentToolResult,
     AgentToolUpdateCallback,
+    AssistantMessage,
     JsonObject,
     TextContent,
 )
 
 from tunacode.core.agents.agent_components import agent_config
+from tunacode.core.session import StateManager
 
 
 async def _wait_for_entered_count(entered_order: list[str], expected: int) -> None:
@@ -70,3 +73,18 @@ async def test_apply_tool_concurrency_limit_caps_in_flight_and_preserves_queue_o
 def test_apply_tool_concurrency_limit_rejects_non_positive_max_parallel() -> None:
     with pytest.raises(ValueError, match="max_parallel_tool_calls must be >= 1"):
         agent_config._apply_tool_concurrency_limit([], max_parallel_tool_calls=0)
+
+
+def test_should_stop_after_turn_updates_runtime_and_stops_after_limit() -> None:
+    state_manager = StateManager()
+    state_manager.session.user_config["settings"]["max_iterations"] = 1
+    should_stop = agent_config._build_should_stop_after_turn(state_manager.session)
+    message = AssistantMessage(content=[TextContent(text="done")])
+
+    assert should_stop(message, [], AgentContext(), []) is False
+    assert state_manager.session.runtime.iteration_count == 1
+    assert state_manager.session.runtime.current_iteration == 1
+
+    assert should_stop(message, [], AgentContext(), []) is True
+    assert state_manager.session.runtime.iteration_count == 2
+    assert state_manager.session.runtime.current_iteration == 2

{tunacode_cli-0.2.1 → tunacode_cli-0.2.2}/tests/unit/ui/test_request_threading.py

@@ -52,6 +52,22 @@ class _FakeBridgeApp:
         return True
 
 
+class _FakeThinkingState:
+    def __init__(self) -> None:
+        self.chunks: list[str] = []
+        self.cleared = False
+        self.finalized = False
+
+    async def callback(self, chunk: str) -> None:
+        self.chunks.append(chunk)
+
+    def clear(self) -> None:
+        self.cleared = True
+
+    def finalize(self) -> None:
+        self.finalized = True
+
+
 class _FakeToolCallbackApp:
     def __init__(self) -> None:
         self.messages: list[object] = []
@@ -177,12 +193,8 @@ async def test_flush_timer_applies_queued_deltas_to_streaming_handler() -> None:
     app = TextualReplApp(state_manager=StateManager())
     app._request_bridge = RequestUiBridge(_FakeBridgeApp())
     app.streaming = _FakeStreamingHandler()  # type: ignore[assignment]
-    thinking_chunks: list[str] = []
-
-    async def _fake_thinking_callback(chunk: str) -> None:
-        thinking_chunks.append(chunk)
-
-    app._thinking_callback = _fake_thinking_callback  # type: ignore[method-assign]
+    thinking_state = _FakeThinkingState()
+    app._thinking_state = thinking_state  # type: ignore[assignment]
 
     await app._request_bridge.streaming_callback("hello")
     await app._request_bridge.streaming_callback(" world")
@@ -191,7 +203,7 @@ async def test_flush_timer_applies_queued_deltas_to_streaming_handler() -> None:
     await app._flush_request_deltas()
 
     assert app.streaming.chunks == ["hello world"]
-    assert thinking_chunks == ["trace data"]
+    assert thinking_state.chunks == ["trace data"]
 
 
 def test_escape_handler_cancels_worker_handle() -> None:
@@ -226,8 +238,7 @@ async def test_process_request_runs_in_worker_and_sets_current_request_handle()
     app.set_interval = lambda *_args, **_kwargs: timer  # type: ignore[method-assign]
     app._show_loading_indicator = lambda: None  # type: ignore[method-assign]
     app._hide_loading_indicator = lambda: None  # type: ignore[method-assign]
-    app._clear_thinking_state = lambda: None  # type: ignore[method-assign]
-    app._finalize_thinking_state_after_request = lambda: None  # type: ignore[method-assign]
+    app._thinking_state = _FakeThinkingState()  # type: ignore[assignment]
     app._update_resource_bar = lambda: None  # type: ignore[method-assign]
     app._get_latest_response_text = lambda: None  # type: ignore[method-assign]
     app._update_compaction_status = compaction_updates.append  # type: ignore[method-assign]