deepagents-code 0.1.48__tar.gz → 0.1.49__tar.gz

{deepagents_code-0.1.48 → deepagents_code-0.1.49}/AGENTS.md

@@ -32,7 +32,7 @@ IMPORTANT: `Content` requires **Textual's** `Style` (`textual.style.Style`) for
 
 ### `App.notify()` defaults to `markup=True`
 
-Textual's `App.notify(message)` parses the message string as Rich markup by default. Any dynamic content (exception messages, file paths, user input, command strings) containing brackets `[]`, ANSI escape codes, or `=` will cause a `MarkupError` crash in Textual's Toast renderer. Always pass `markup=False` when the message contains f-string interpolated variables. Hardcoded string literals are safe with the default.
+Textual's `App.notify(message)` parses the message string as Rich markup by default. Any dynamic content (exception messages, file paths, user input, command strings) containing square brackets is unsafe: tag-shaped text like `[bold]` is silently swallowed, and closing-tag-shaped text like `[/tmp/x]` raises a `MarkupError` crash in Textual's Toast renderer. Always pass `markup=False` when the message contains f-string interpolated variables. Hardcoded string literals are safe with the default.
 
 ### Rich `console.print()` and number highlighting
 

{deepagents_code-0.1.48 → deepagents_code-0.1.49}/CHANGELOG.md

@@ -2,6 +2,21 @@
 
 # Deep Agents Code Changelog
 
+## [0.1.49](https://github.com/langchain-ai/deepagents/compare/deepagents-code==0.1.48...deepagents-code==0.1.49) (2026-07-27)
+
+### Features
+
+- Added recognition for LangSmith Gateway credentials. ([#5042](https://github.com/langchain-ai/deepagents/issues/5042))
+- Added slash commands for mode changes and an option to escape to manual mode from the YOLO notice. ([#5092](https://github.com/langchain-ai/deepagents/issues/5092))
+- Added an `install` subcommand. ([#5080](https://github.com/langchain-ai/deepagents/issues/5080))
+
+### Bug Fixes
+
+- Kept the post-install restart modal responsive. ([#5086](https://github.com/langchain-ai/deepagents/issues/5086))
+- Moved server graph bootstrap `cwd` calls off the event loop. ([#5081](https://github.com/langchain-ai/deepagents/issues/5081))
+- Improved handling of long transcript user messages with toast notifications and expansion. ([#5073](https://github.com/langchain-ai/deepagents/issues/5073))
+- Show model no-op notices as toasts instead of inline messages. ([#5074](https://github.com/langchain-ai/deepagents/issues/5074))
+
 ## [0.1.48](https://github.com/langchain-ai/deepagents/compare/deepagents-code==0.1.47...deepagents-code==0.1.48) (2026-07-27)
 
 ### Features

{deepagents_code-0.1.48 → deepagents_code-0.1.49}/COMMANDS.md

@@ -8,12 +8,13 @@ Regenerate this file with `make commands-catalog` after changing command names,
 aliases, descriptions, visibility, or hidden-command metadata.
 
 
-## Public (34)
+## Public (37)
 
 | Command | Aliases | Description |
 | --- | --- | --- |
 | `/agents` |  | Browse and switch between available agents |
 | `/auth` | `/connect` | Connect and manage provider and service credentials |
+| `/auto` |  | Switch to Auto approval mode |
 | `/auto-update` |  | Turn automatic updates on or off |
 | `/changelog` |  | Open the changelog in a browser |
 | `/clear` |  | Clear the chat and start a new thread |
@@ -26,6 +27,7 @@ aliases, descriptions, visibility, or hidden-command metadata.
 | `/goal` |  | Set and manage a persistent objective with acceptance criteria |
 | `/help` |  | Show help and available commands |
 | `/install` |  | Install an optional integration |
+| `/manual` |  | Switch to Manual approval mode |
 | `/mcp` |  | Manage MCP servers and authentication |
 | `/model` |  | Switch models or edit model settings |
 | `/notifications` |  | Configure startup warnings |
@@ -46,6 +48,7 @@ aliases, descriptions, visibility, or hidden-command metadata.
 | `/trace` |  | Open this thread in LangSmith |
 | `/update` |  | Check for and install updates |
 | `/version` | `/about` | Show version information |
+| `/yolo` |  | Switch to YOLO approval mode (no review) |
 
 ## Hidden (2)
 

{deepagents_code-0.1.48 → deepagents_code-0.1.49}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: deepagents-code
-Version: 0.1.48
+Version: 0.1.49
 Summary: Terminal coding agent built on the Deep Agents SDK - works with any model, with persistent memory, customizable skills, subagents, and approval controls.
 Project-URL: Homepage, https://docs.langchain.com/oss/python/deepagents/overview
 Project-URL: Documentation, https://reference.langchain.com/python/deepagents/
@@ -32,10 +32,10 @@ Requires-Dist: deepagents==0.7.0b2
 Requires-Dist: filelock<3.30,>=3.29.7
 Requires-Dist: httpx<1.0.0,>=0.28.1
 Requires-Dist: langchain-anthropic<2.0.0,>=1.5.2
-Requires-Dist: langchain-google-genai<5.0.0,>=4.3.1
+Requires-Dist: langchain-google-genai<5.0.0,>=4.3.2
 Requires-Dist: langchain-mcp-adapters<1.0.0,>=0.3.0
 Requires-Dist: langchain-openai<2.0.0,>=1.4.1
-Requires-Dist: langchain-quickjs<0.4.0,>=0.3.3
+Requires-Dist: langchain-quickjs<0.4.0,>=0.3.4
 Requires-Dist: langchain<2.0.0,>=1.3.14
 Requires-Dist: langgraph-checkpoint-sqlite<4.0.0,>=3.1.0
 Requires-Dist: langgraph-cli[inmem]<1.0.0,>=0.4.31
@@ -46,7 +46,7 @@ Requires-Dist: markdownify<2.0.0,>=1.2.3
 Requires-Dist: mcp<2.0.0,>=1.28.1
 Requires-Dist: packaging>=26.2
 Requires-Dist: pillow<13.0.0,>=12.3.0
-Requires-Dist: prompt-toolkit<4.0.0,>=3.0.52
+Requires-Dist: prompt-toolkit<4.0.0,>=3.0.53
 Requires-Dist: pydantic<2.14.0a0,>=2.12.5
 Requires-Dist: pyperclip<2.0.0,>=1.11.0
 Requires-Dist: python-dotenv<2.0.0,>=1.2.2
@@ -65,11 +65,11 @@ Provides-Extra: all-providers
 Requires-Dist: aiohttp<3.15.0,>=3.14.3; extra == 'all-providers'
 Requires-Dist: langchain-anthropic<2.0.0,>=1.5.2; extra == 'all-providers'
 Requires-Dist: langchain-aws<2.0.0,>=1.6.3; extra == 'all-providers'
-Requires-Dist: langchain-baseten<1.0.0,>=0.2.1; extra == 'all-providers'
+Requires-Dist: langchain-baseten<1.0.0,>=0.2.3; extra == 'all-providers'
 Requires-Dist: langchain-cohere<1.0.0,>=0.6.0; extra == 'all-providers'
 Requires-Dist: langchain-deepseek<2.0.0,>=1.1.0; extra == 'all-providers'
-Requires-Dist: langchain-fireworks<2.0.0,>=1.5.1; extra == 'all-providers'
-Requires-Dist: langchain-google-genai<5.0.0,>=4.3.1; extra == 'all-providers'
+Requires-Dist: langchain-fireworks<2.0.0,>=1.5.2; extra == 'all-providers'
+Requires-Dist: langchain-google-genai<5.0.0,>=4.3.2; extra == 'all-providers'
 Requires-Dist: langchain-google-vertexai<4.0.0,>=3.2.4; extra == 'all-providers'
 Requires-Dist: langchain-groq<2.0.0,>=1.1.3; extra == 'all-providers'
 Requires-Dist: langchain-huggingface<2.0.0,>=1.2.2; extra == 'all-providers'
@@ -93,7 +93,7 @@ Requires-Dist: langchain-vercel-sandbox>=0.0.1; extra == 'all-sandboxes'
 Provides-Extra: anthropic
 Requires-Dist: langchain-anthropic<2.0.0,>=1.5.2; extra == 'anthropic'
 Provides-Extra: baseten
-Requires-Dist: langchain-baseten<1.0.0,>=0.2.1; extra == 'baseten'
+Requires-Dist: langchain-baseten<1.0.0,>=0.2.3; extra == 'baseten'
 Provides-Extra: bedrock
 Requires-Dist: langchain-aws<2.0.0,>=1.6.3; extra == 'bedrock'
 Provides-Extra: cohere
@@ -103,9 +103,9 @@ Requires-Dist: langchain-daytona>=0.0.7; extra == 'daytona'
 Provides-Extra: deepseek
 Requires-Dist: langchain-deepseek<2.0.0,>=1.1.0; extra == 'deepseek'
 Provides-Extra: fireworks
-Requires-Dist: langchain-fireworks<2.0.0,>=1.5.1; extra == 'fireworks'
+Requires-Dist: langchain-fireworks<2.0.0,>=1.5.2; extra == 'fireworks'
 Provides-Extra: google-genai
-Requires-Dist: langchain-google-genai<5.0.0,>=4.3.1; extra == 'google-genai'
+Requires-Dist: langchain-google-genai<5.0.0,>=4.3.2; extra == 'google-genai'
 Provides-Extra: groq
 Requires-Dist: langchain-groq<2.0.0,>=1.1.3; extra == 'groq'
 Provides-Extra: huggingface

{deepagents_code-0.1.48 → deepagents_code-0.1.49}/deepagents_code/_build_info.py

@@ -1,3 +1,3 @@
 """Generated at build time. Do not edit or commit."""
 
-BUILD_COMMIT = "803b832"
+BUILD_COMMIT = "44910bc"

{deepagents_code-0.1.48 → deepagents_code-0.1.49}/deepagents_code/_version.py

@@ -2,7 +2,7 @@
 
 # Keep the `x-release-please-version` annotation — release-please uses it to
 # bump `__version__` in sync with `pyproject.toml` on every release PR.
-__version__ = "0.1.48"  # x-release-please-version
+__version__ = "0.1.49"  # x-release-please-version
 
 DOCS_URL = "https://docs.langchain.com/oss/python/deepagents/code"
 """URL for `deepagents-code` documentation."""

{deepagents_code-0.1.48 → deepagents_code-0.1.49}/deepagents_code/app.py

@@ -571,7 +571,13 @@ class _ConfigWriteResult:
 ScreenResultT = TypeVar("ScreenResultT")
 
 if TYPE_CHECKING:
-    from collections.abc import AsyncIterator, Awaitable, Callable, Mapping
+    from collections.abc import (
+        AsyncIterator,
+        Awaitable,
+        Callable,
+        Coroutine,
+        Mapping,
+    )
 
     from deepagents.backends import CompositeBackend
     from langchain_core.messages import BaseMessage
@@ -3161,8 +3167,14 @@ class DeepAgentsApp(App):
         )
         """Per-turn model params override set via startup or `/model` params."""
 
-        self._last_model_unchanged_message: str | None = None
-        """Most recent same-model notice, used to suppress duplicates."""
+        self._last_model_unchanged: tuple[str, float] | None = None
+        """Most recent same-model toast, as `(text, monotonic timestamp)`.
+
+        Used to suppress duplicates. Suppression only lasts for the toast
+        lifetime (`NOTIFICATION_TIMEOUT`); after expiry, a later identical
+        no-op selection can toast again. Text and timestamp are one field so
+        they cannot drift out of sync.
+        """
 
         self._model_install_switching = False
         """True while a provider extra install-then-switch flow is active."""
@@ -6179,9 +6191,14 @@ class DeepAgentsApp(App):
         # restart. `_offer_restart_after_install` owns all follow-up messaging
         # (the prompt's button is the call to action when shown; it mounts a
         # `/restart`-or-relaunch hint itself when it can't show the prompt), so
-        # a redundant "Run /restart" line is never appended here.
+        # a redundant "Run /restart" line is never appended here. The offer is
+        # scheduled off the message pump so the modal stays responsive (see
+        # `_schedule_restart_offer`); the interactive offer never affects this
+        # method's return value, so detaching it does not change the contract.
         await self._mount_message(AppMessage(f"Installed extra '{extra}'."))
-        await self._offer_restart_after_install(extra)
+        self._schedule_restart_offer(
+            self._offer_restart_after_install(extra), context=f"extra:{extra}"
+        )
         return True
 
     async def _handle_install_package(self, package: str, *, force: bool) -> None:
@@ -6272,7 +6289,11 @@ class DeepAgentsApp(App):
                 "now, or relaunch dcode.",
             ),
         )
-        await self._offer_restart_after_install(package)
+        # Scheduled off the message pump so the restart modal stays responsive
+        # to Enter/Esc (see `_schedule_restart_offer`).
+        self._schedule_restart_offer(
+            self._offer_restart_after_install(package), context=f"package:{package}"
+        )
 
     async def _confirm_install_package(self, package: str) -> bool:
         """Ask the user to confirm installing an arbitrary package.
@@ -12379,6 +12400,10 @@ class DeepAgentsApp(App):
             await self._handle_version_command()
         elif cmd == "/agents":
             await self._show_agent_selector()
+        elif cmd in {"/manual", "/auto", "/yolo"}:
+            from deepagents_code.approval_mode import ApprovalMode
+
+            await self._handle_approval_mode_command(ApprovalMode(cmd.lstrip("/")))
         elif cmd == "/goal" or cmd.startswith("/goal "):
             await self._handle_goal_command(command)
         elif cmd in {"/rubric", "/criteria"} or cmd.startswith(
@@ -13821,6 +13846,24 @@ class DeepAgentsApp(App):
         )
         await self._mount_message(user_message)
         self._active_user_message = user_message
+        # Toast only on submit when the transcript will collapse the body —
+        # the model still receives the full text; the UI is head+tail until
+        # the user expands. Asking the widget (rather than re-deriving from
+        # `message`) keeps this in step with what `render()` actually collapses,
+        # including its mode-prefix handling.
+        if user_message.has_expandable_body:
+            # "Shortened", not "collapsed": a large paste already toasted
+            # "Large paste collapsed" in the composer (see chat_input), and
+            # that means something different (a placeholder chip, not a
+            # head+tail elision in the transcript).
+            self.notify(
+                "Long message shortened in the transcript — click "
+                "'show full message' to expand it. The full text was still "
+                "sent to the model.",
+                severity="information",
+                markup=False,
+                timeout=8,
+            )
         await self._send_to_agent(message)
 
     async def _send_to_agent(
@@ -15723,6 +15766,23 @@ class DeepAgentsApp(App):
             )
             self._schedule_message_height_measurement(event.widget.id)
 
+    def on_user_message_expansion_changed(
+        self,
+        event: UserMessage.ExpansionChanged,
+    ) -> None:
+        """Keep long-prompt expansion state across transcript virtualization.
+
+        Also re-measures the row: expanding a collapsed prompt can add hundreds
+        of lines, and the spacer math that sizes the scrollbar reads a cached
+        height that `refresh(layout=True)` alone does not update.
+        """
+        if event.widget.id:
+            self._message_store.update_message(
+                event.widget.id,
+                user_expanded=event.expanded,
+            )
+            self._schedule_message_height_measurement(event.widget.id)
+
     async def _clear_messages(self) -> None:
         """Clear the messages area and message store."""
         # Drop buffered `!` shell output so it never leaks across a thread
@@ -16865,8 +16925,8 @@ class DeepAgentsApp(App):
             text = f"Auto denied [{event.get('category', 'policy')}]: {reason}"
         elif kind == "unavailable":
             # Reason is a short cause fragment from the server; keep the UI
-            # line outcome-focused so the blocked action is obvious.
-            text = f"Auto classifier unavailable: {reason} (action blocked)"
+            # line outcome-focused so fail-closed denial is obvious.
+            text = f"Auto classifier unavailable: {reason} — tool not executed"
         else:
             text = f"Auto warning: {reason}"
         await self._mount_message(AppMessage(text))
@@ -17032,25 +17092,82 @@ class DeepAgentsApp(App):
                 await self._auto_accept_pending_goal_rubric()
         return True
 
+    async def _handle_approval_mode_command(self, target: ApprovalMode) -> None:
+        """Switch approval mode from a slash command (`/manual`, `/auto`, `/yolo`).
+
+        Mirrors the Shift+Tab switcher: Auto is refused with a sandbox, YOLO
+        honors the `startup.yolo_switcher` policy and prompts the first-run
+        acknowledgement modal before unrestricted mode becomes active. Feedback
+        is delivered via toasts and the status bar, not chat messages, so the
+        command stays quiet like the keybinding.
+
+        Args:
+            target: The approval mode requested by the command.
+        """
+        from deepagents_code.approval_mode import (
+            ApprovalMode,
+            has_yolo_acknowledgement,
+        )
+        from deepagents_code.config import is_yolo_switcher_enabled
+
+        if self._approval_mode is target:
+            self.notify(
+                f"Already in {target.value.capitalize()} mode.",
+                severity="information",
+                markup=False,
+            )
+            return
+
+        if target is ApprovalMode.AUTO and not self._auto_mode_eligible:
+            self._warn_live_approval_mode_unavailable(
+                "Auto is unavailable with a sandbox."
+            )
+            return
+
+        if target is ApprovalMode.YOLO:
+            if not is_yolo_switcher_enabled():
+                self._warn_live_approval_mode_unavailable(
+                    "YOLO is disabled in the approval switcher."
+                )
+                return
+            if not has_yolo_acknowledgement():
+                self._prompt_yolo_switcher_acknowledgement()
+                return
+
+        await self._set_approval_mode(target)
+
     def _prompt_yolo_switcher_acknowledgement(self) -> None:
         """Show the first-run YOLO acknowledgement before entering unrestricted mode.
 
         YOLO stays inactive until Enter persists the acknowledgement and the
-        follow-up mode write succeeds. Esc (or a non-true dismiss) leaves the
-        current mode untouched and does not store the acknowledgement.
+        follow-up mode write succeeds. `m` switches to Manual instead. Esc (or a
+        non-acknowledge dismiss) leaves the current mode untouched and does not
+        store the acknowledgement.
         """
         from deepagents_code.approval_mode import (
             ApprovalMode,
             save_yolo_acknowledgement,
         )
-        from deepagents_code.tui.widgets.yolo_mode_notice import YoloModeNoticeScreen
+        from deepagents_code.tui.widgets.yolo_mode_notice import (
+            YoloModeNoticeResult,
+            YoloModeNoticeScreen,
+        )
 
         if getattr(self, "_yolo_mode_notice_pending", False):
             return
 
-        def handle_result(accepted: bool | None) -> None:
+        def handle_result(result: YoloModeNoticeResult | None) -> None:
             self._yolo_mode_notice_pending = False
-            if accepted is not True:
+            if result is YoloModeNoticeResult.MANUAL:
+                # Already-Manual is a no-op: skip the live store write so a
+                # failed redundant write cannot flip `_approval_mode_blocked`
+                # and cancel an active run after the user declined YOLO.
+                if self._approval_mode is ApprovalMode.MANUAL:
+                    return
+                task = asyncio.create_task(self._set_approval_mode(ApprovalMode.MANUAL))
+                task.add_done_callback(_log_task_exception)
+                return
+            if result is not YoloModeNoticeResult.ACKNOWLEDGE:
                 self.notify(
                     "Stayed in the current approval mode.",
                     severity="information",
@@ -17139,6 +17256,13 @@ class DeepAgentsApp(App):
                 if child.has_output:
                     child.toggle_output()
                     return
+            # Long UserMessages are collapsible. This scan is newest-first and
+            # first-match-wins, so any later tool/skill/group row claims Ctrl+O
+            # and the prompt is then reachable only by click. Assistant rows are
+            # not in the chain, so they do not block it.
+            if isinstance(child, UserMessage) and child.has_expandable_body:
+                child.toggle_expanded()
+                return
 
     # Approval menu action handlers (delegated from App-level bindings)
     # NOTE: These only activate when approval widget is pending
@@ -20759,6 +20883,28 @@ class DeepAgentsApp(App):
         except ModuleNotFoundError:
             logger.warning("Could not preload restart_prompt modal", exc_info=True)
 
+    def _schedule_restart_offer(
+        self, coro: Coroutine[Any, Any, None], *, context: str
+    ) -> None:
+        """Run a post-install restart offer as a detached background task.
+
+        The `/install` slash command reaches the restart offer through the App's
+        `on_chat_input_submitted` handler, which is awaited inline on the App
+        message pump. Awaiting the restart modal there blocks the pump, so the
+        modal never receives the Enter/Esc key events it needs to resolve and
+        appears frozen. Detaching the offer onto its own task lets the command
+        handler return and frees the pump to route keys to the modal — the same
+        reason the post-`/auth` web-search offer is scheduled rather than
+        awaited (see `_launch_web_search_restart_prompt`).
+
+        Args:
+            coro: The restart-offer coroutine to run off the message pump.
+            context: Short description used if the task is logged on failure.
+        """
+        task = asyncio.create_task(coro, name=f"restart-offer:{context}")
+        self._track_server_restart_task(task)
+        task.add_done_callback(_log_task_exception)
+
     async def _offer_restart_after_install(self, label: str) -> None:
         """Offer a one-keypress restart after a restart-capable install.
 
@@ -22263,7 +22409,7 @@ class DeepAgentsApp(App):
                 (e.g., `'anthropic:claude-sonnet-4-5'`) or just the model name
                 for auto-detection.
             extra_kwargs: Extra constructor kwargs from `--model-params`.
-            announce_unchanged: Whether to mount a message when the requested
+            announce_unchanged: Whether to toast a notice when the requested
                 model is already active.
             persist: Whether to write the model to the user's recent/default
                 config.
@@ -22390,9 +22536,21 @@ class DeepAgentsApp(App):
                 params_suffix = _format_model_params(extra_kwargs)
                 if announce_unchanged:
                     message = f"Already using {current}{params_suffix}"
-                    if message != self._last_model_unchanged_message:
-                        await self._mount_message(AppMessage(message))
-                        self._last_model_unchanged_message = message
+                    # Suppress only while the previous identical toast is
+                    # presumed still on-screen. Once it expires, a later
+                    # intentional no-op selection must be able to toast again.
+                    now = _monotonic()
+                    last = self._last_model_unchanged
+                    if (
+                        last is None
+                        or last[0] != message
+                        or (now - last[1]) >= self.NOTIFICATION_TIMEOUT
+                    ):
+                        # A no-op re-selection is transient feedback, not part
+                        # of the conversation, so surface it as a toast rather
+                        # than an inline chat message.
+                        self.notify(message, markup=False)
+                        self._last_model_unchanged = (message, now)
                 logger.info(
                     "Model unchanged (%s); model_params=%s",
                     current,
@@ -22444,7 +22602,7 @@ class DeepAgentsApp(App):
 
             self._sync_status_model()
 
-            self._last_model_unchanged_message = None
+            self._last_model_unchanged = None
             params_suffix = _format_model_params(extra_kwargs)
             if not persist:
                 # Session-only switch (e.g. adopting a resumed thread's model):

{deepagents_code-0.1.48 → deepagents_code-0.1.49}/deepagents_code/auto_mode.py

@@ -506,9 +506,10 @@ def classifier_unavailable_reason(exc: BaseException, *, timeout_seconds: float)
 
     Provider exception text stays out of the reason (it can carry secrets or
     noisy HTML). Only real local deadline expiry
-    (`_ClassifierDeadlineExceededError`) includes the configured wait budget; a
-    bare provider `TimeoutError` stays type-only so we do not claim dcode's
-    deadline fired when the model failed first.
+    (`_ClassifierDeadlineExceededError`) says the classifier did not respond
+    within the configured wait budget; a bare provider `TimeoutError` stays
+    type-only so we do not claim dcode's deadline fired when the model failed
+    first.
 
     Args:
         exc: Failure raised while invoking or validating the classifier.
@@ -518,7 +519,7 @@ def classifier_unavailable_reason(exc: BaseException, *, timeout_seconds: float)
         Compact single-line reason for tool messages and TUI events.
     """
     if isinstance(exc, _ClassifierDeadlineExceededError):
-        return f"no decision within Auto's {timeout_seconds:g}s limit"
+        return f"classifier did not respond within {timeout_seconds:g}s"
     return f"failed ({type(exc).__name__})"