deepagents-code 0.1.28__tar.gz → 0.1.30__tar.gz

{deepagents_code-0.1.28 → deepagents_code-0.1.30}/CHANGELOG.md

@@ -2,6 +2,36 @@
 
 # Deep Agents Code Changelog
 
+## [0.1.30](https://github.com/langchain-ai/deepagents/compare/deepagents-code==0.1.29...deepagents-code==0.1.30) (2026-07-02)
+
+### Features
+
+* Add reasoning effort selector ([#4403](https://github.com/langchain-ai/deepagents/issues/4403)) ([6ee0ac4](https://github.com/langchain-ai/deepagents/commit/6ee0ac4cca998f84e157187d565084e2a1912626))
+* Collapse completed tool calls into group summaries ([#4373](https://github.com/langchain-ai/deepagents/issues/4373)) ([3735829](https://github.com/langchain-ai/deepagents/commit/3735829a0c14fb1537daf930140c66b647c8222c))
+* Auto-detect MCP OAuth from 401 challenge ([#4364](https://github.com/langchain-ai/deepagents/issues/4364)) ([9763ffc](https://github.com/langchain-ai/deepagents/commit/9763ffceab8800526ae9fe10decdb2f32d8e4707))
+* Toast when opening a clicked URL ([#4368](https://github.com/langchain-ai/deepagents/issues/4368)) ([434f29e](https://github.com/langchain-ai/deepagents/commit/434f29e5cb3d94988e2a4d80b78452bee4b10812))
+
+### Bug Fixes
+
+* Allow `/remember` with args when no conversation history ([#4418](https://github.com/langchain-ai/deepagents/issues/4418)) ([5df9c83](https://github.com/langchain-ai/deepagents/commit/5df9c83b58f2529561848ae9f1364961e1c80641))
+* Persist rubric model after server restart ([#4419](https://github.com/langchain-ai/deepagents/issues/4419)) ([5605a68](https://github.com/langchain-ai/deepagents/commit/5605a681c9fec888d8c6dfbde2decd4692c8cef0))
+* Persist resume model state privately ([#4400](https://github.com/langchain-ai/deepagents/issues/4400)) ([bbd0f0d](https://github.com/langchain-ai/deepagents/commit/bbd0f0d1ddc9d22d399a819d7b4363f7997d119e))
+* Quiet routine ripgrep installer output ([#4417](https://github.com/langchain-ai/deepagents/issues/4417)) ([a52c18d](https://github.com/langchain-ai/deepagents/commit/a52c18d3ef8fe462d1289dca8798c51450e0368a))
+* Honest MCP OAuth callback close message ([#4410](https://github.com/langchain-ai/deepagents/issues/4410)) ([ef637f4](https://github.com/langchain-ai/deepagents/commit/ef637f4a3f38ee155ad7d31e64a5cdd7bfaee62c))
+
+## [0.1.29](https://github.com/langchain-ai/deepagents/compare/deepagents-code==0.1.28...deepagents-code==0.1.29) (2026-07-01)
+
+### Features
+
+* Add rubric iteration controls ([#4405](https://github.com/langchain-ai/deepagents/issues/4405)) ([d6692a7](https://github.com/langchain-ai/deepagents/commit/d6692a7c713490f170b17510d613e02ee37574ab))
+* Add Claude Sonnet 5 to model picker ([#4386](https://github.com/langchain-ai/deepagents/issues/4386)) ([64758dd](https://github.com/langchain-ai/deepagents/commit/64758dddf8b46c5b68c345fb1eac1fb7bcbf7f7e))
+
+### Bug Fixes
+
+* Full-width chat messages, hide scrollbar, flush input bg ([#4374](https://github.com/langchain-ai/deepagents/issues/4374)) ([1f8e8dc](https://github.com/langchain-ai/deepagents/commit/1f8e8dc942ac54c4014ac48f9345ee0201be0b1b))
+* Resolve editable SDK version metadata ([#4394](https://github.com/langchain-ai/deepagents/issues/4394)) ([3239bf4](https://github.com/langchain-ai/deepagents/commit/3239bf4edf2be3037bb18315463ac6a0c3537e5c))
+* Unblock MCP force reconnect modal ([#4396](https://github.com/langchain-ai/deepagents/issues/4396)) ([8b7eab0](https://github.com/langchain-ai/deepagents/commit/8b7eab023db686f577d16f4f9ac872205baf136f))
+
 ## [0.1.28](https://github.com/langchain-ai/deepagents/compare/deepagents-code==0.1.27...deepagents-code==0.1.28) (2026-06-29)
 
 ### Features

{deepagents_code-0.1.28 → deepagents_code-0.1.30}/COMMANDS.md

@@ -8,7 +8,7 @@ Regenerate this file with `make commands-catalog` after changing command names,
 aliases, descriptions, visibility, or hidden-command metadata.
 
 
-## Public (30)
+## Public (32)
 
 | Command | Aliases | Description |
 | --- | --- | --- |
@@ -20,6 +20,7 @@ aliases, descriptions, visibility, or hidden-command metadata.
 | `/copy` |  | Copy the latest assistant message to clipboard |
 | `/docs` |  | Open the docs |
 | `/editor` |  | Open prompt in an external editor ($EDITOR) |
+| `/effort` |  | Set reasoning effort for the current model |
 | `/feedback` |  | Send feedback or report an issue |
 | `/force-clear` |  | Stop active work, clear the chat, and start a new thread |
 | `/goal` |  | Set a persistent objective by drafting acceptance criteria |
@@ -34,6 +35,7 @@ aliases, descriptions, visibility, or hidden-command metadata.
 | `/remember` |  | Save useful context to memory or skills |
 | `/restart` |  | Restart the agent server |
 | `/rubric` | `/criteria` | Set explicit acceptance criteria for rubric grading |
+| `/scrollbar` |  | Show or hide the chat scrollbar |
 | `/skill-creator` |  | Create or refine agent skills |
 | `/theme` |  | Change color theme |
 | `/threads` |  | Browse and resume past threads |

{deepagents_code-0.1.28 → deepagents_code-0.1.30}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: deepagents-code
-Version: 0.1.28
+Version: 0.1.30
 Summary: Terminal interface for Deep Agents - interactive AI agent with file operations, shell access, and sub-agent capabilities.
 Project-URL: Homepage, https://docs.langchain.com/oss/python/deepagents/overview
 Project-URL: Documentation, https://reference.langchain.com/python/deepagents/
@@ -27,7 +27,7 @@ Classifier: Topic :: Terminals
 Requires-Python: <4.0,>=3.11
 Requires-Dist: aiosqlite<1.0.0,>=0.22.1
 Requires-Dist: deepagents-acp<1.0.0,>=0.0.8
-Requires-Dist: deepagents==0.7.0a2
+Requires-Dist: deepagents==0.7.0a3
 Requires-Dist: httpx<1.0.0,>=0.28.1
 Requires-Dist: langchain-anthropic<2.0.0,>=1.4.7
 Requires-Dist: langchain-google-genai<5.0.0,>=4.2.6
@@ -60,6 +60,7 @@ Requires-Dist: uuid-utils<1.0.0,>=0.16.2
 Provides-Extra: agentcore
 Requires-Dist: langchain-agentcore-codeinterpreter<1.0.0,>=0.0.5; extra == 'agentcore'
 Provides-Extra: all-providers
+Requires-Dist: aiohttp<3.15.0,>=3.14.1; extra == 'all-providers'
 Requires-Dist: langchain-anthropic<2.0.0,>=1.4.7; extra == 'all-providers'
 Requires-Dist: langchain-aws<2.0.0,>=1.6.1; extra == 'all-providers'
 Requires-Dist: langchain-baseten<1.0.0,>=0.2.1; extra == 'all-providers'
@@ -73,7 +74,7 @@ Requires-Dist: langchain-huggingface<2.0.0,>=1.2.2; extra == 'all-providers'
 Requires-Dist: langchain-ibm<2.0.0,>=1.1.0; extra == 'all-providers'
 Requires-Dist: langchain-litellm<2.0.0,>=0.7.0; extra == 'all-providers'
 Requires-Dist: langchain-mistralai<2.0.0,>=1.1.5; extra == 'all-providers'
-Requires-Dist: langchain-nvidia-ai-endpoints<2.0.0,>=1.4.2; extra == 'all-providers'
+Requires-Dist: langchain-nvidia-ai-endpoints<2.0.0,>=1.4.3; extra == 'all-providers'
 Requires-Dist: langchain-ollama<2.0.0,>=1.1.0; extra == 'all-providers'
 Requires-Dist: langchain-openai<2.0.0,>=1.3.3; extra == 'all-providers'
 Requires-Dist: langchain-openrouter<2.0.0,>=0.2.4; extra == 'all-providers'
@@ -115,7 +116,8 @@ Requires-Dist: langchain-mistralai<2.0.0,>=1.1.5; extra == 'mistralai'
 Provides-Extra: modal
 Requires-Dist: langchain-modal>=0.0.5; extra == 'modal'
 Provides-Extra: nvidia
-Requires-Dist: langchain-nvidia-ai-endpoints<2.0.0,>=1.4.2; extra == 'nvidia'
+Requires-Dist: aiohttp<3.15.0,>=3.14.1; extra == 'nvidia'
+Requires-Dist: langchain-nvidia-ai-endpoints<2.0.0,>=1.4.3; extra == 'nvidia'
 Provides-Extra: ollama
 Requires-Dist: langchain-ollama<2.0.0,>=1.1.0; extra == 'ollama'
 Provides-Extra: openai

{deepagents_code-0.1.28 → deepagents_code-0.1.30}/deepagents_code/_build_info.py

@@ -1,3 +1,3 @@
 """Generated at build time. Do not edit or commit."""
 
-BUILD_COMMIT = "14a0fb1"
+BUILD_COMMIT = "e4709c2"

{deepagents_code-0.1.28 → deepagents_code-0.1.30}/deepagents_code/_cli_context.py

@@ -35,8 +35,6 @@ class CLIContextSchema:
 
     model_params: dict[str, Any] = field(default_factory=dict)
 
-    effective_model: str | None = None
-
     auto_approve: bool = False
 
     approval_mode_key: str | None = None
@@ -63,17 +61,6 @@ class CLIContext(TypedDict, total=False):
     """Invocation params (e.g. `temperature`, `max_tokens`) to merge
     into `model_settings`."""
 
-    effective_model: str | None
-    """Resolved `provider:model` spec actually in use for this invocation.
-
-    Unlike `model` (a swap *instruction* that is `None` when the base model is
-    used), this carries the model in effect — whether from a `/model` override
-    or the startup default — so `ResumeStateMiddleware` can record it to
-    checkpoint state for restore on resume. `None` when no usable spec is
-    resolved yet (e.g. credentials not configured), in which case nothing is
-    recorded rather than a malformed spec.
-    """
-
     auto_approve: bool
     """Whether gated tool calls should skip the human-approval interrupt.
 

{deepagents_code-0.1.28 → deepagents_code-0.1.30}/deepagents_code/_env_vars.py

@@ -206,6 +206,25 @@ Defaults to enabled; set to a falsy value (`0`, `false`, `no`, `off`, or empty)
 to hide replica tracing details from the splash while leaving tracing active.
 """
 
+SHOW_SCROLLBAR = "DEEPAGENTS_CODE_SHOW_SCROLLBAR"
+"""Show the vertical scrollbar in the chat area when enabled.
+
+Off by default; use the `/scrollbar` slash command or `[ui].show_scrollbar` in
+config.toml to toggle. Parsed by `classify_env_bool` (an unrecognized or empty
+value falls through to the config value rather than forcing the default).
+
+When set, this env var takes precedence over the persisted `[ui].show_scrollbar`
+config value on launch, so a `/scrollbar` toggle will not appear to "stick"
+across restarts while the env var remains set.
+"""
+
+SHOW_URL_OPEN_TOAST = "DEEPAGENTS_CODE_SHOW_URL_OPEN_TOAST"
+"""Show a confirmation toast after clicking a URL that opens in a browser.
+
+Defaults to enabled; set to a falsy value (`0`, `false`, `no`, `off`, or empty)
+to suppress the success toast while still opening URLs normally.
+"""
+
 THEME = "DEEPAGENTS_CODE_THEME"
 """Force the CLI to launch with this theme name when set."""
 

{deepagents_code-0.1.28 → deepagents_code-0.1.30}/deepagents_code/_server_config.py

@@ -80,7 +80,7 @@ def _read_env_str(suffix: str) -> str | None:
     return os.environ.get(f"{SERVER_ENV_PREFIX}{suffix}")
 
 
-def _read_env_int(suffix: str, *, default: int) -> int:
+def _read_env_int(suffix: str, *, default: int | None) -> int | None:
     """Read a `DEEPAGENTS_CODE_SERVER_*` integer from the environment.
 
     Args:
@@ -88,7 +88,7 @@ def _read_env_int(suffix: str, *, default: int) -> int:
         default: Value when the variable is absent or malformed.
 
     Returns:
-        Parsed integer, or the default when parsing fails.
+        Parsed integer, or the default when absent or parsing fails.
     """
     raw = os.environ.get(f"{SERVER_ENV_PREFIX}{suffix}")
     if raw is None:
@@ -264,9 +264,8 @@ class ServerConfig:
     `None` reuses the main agent model.
     """
 
-    rubric_max_iterations: int = 3
-    """Grader iterations per rubric attempt before the agent stops with
-    `'max_iterations_reached'`."""
+    rubric_max_iterations: int | None = None
+    """Explicit grader iterations per rubric attempt; `None` uses the SDK default."""
 
     sandbox_type: str | None = None
     """Sandbox backend identifier (e.g. `'daytona'`); `None` runs tools on the
@@ -304,13 +303,21 @@ class ServerConfig:
         """Normalize fields and validate invariants.
 
         Raises:
-            ValueError: If `shell_allow_list` is an empty list.
+            TypeError: If `rubric_max_iterations` is a boolean.
+            ValueError: If `shell_allow_list` is an empty list or
+                `rubric_max_iterations` is non-positive.
         """
         if self.sandbox_type == "none":
             object.__setattr__(self, "sandbox_type", None)
         if self.shell_allow_list is not None and len(self.shell_allow_list) == 0:
             msg = "shell_allow_list must be None or non-empty"
             raise ValueError(msg)
+        if isinstance(self.rubric_max_iterations, bool):
+            msg = "rubric_max_iterations must be None or a positive integer"
+            raise TypeError(msg)
+        if self.rubric_max_iterations is not None and self.rubric_max_iterations <= 0:
+            msg = "rubric_max_iterations must be None or a positive integer"
+            raise ValueError(msg)
 
     # ------------------------------------------------------------------
     # Serialization
@@ -356,7 +363,11 @@ class ServerConfig:
                 self.interpreter_ptc_acknowledge_unsafe
             ).lower(),
             "RUBRIC_MODEL": self.rubric_model,
-            "RUBRIC_MAX_ITERATIONS": str(self.rubric_max_iterations),
+            "RUBRIC_MAX_ITERATIONS": (
+                str(self.rubric_max_iterations)
+                if self.rubric_max_iterations is not None
+                else None
+            ),
             "SANDBOX_TYPE": self.sandbox_type,
             "SANDBOX_ID": self.sandbox_id,
             "SANDBOX_SNAPSHOT_NAME": self.sandbox_snapshot_name,
@@ -406,7 +417,7 @@ class ServerConfig:
                 "INTERPRETER_PTC_ACKNOWLEDGE_UNSAFE"
             ),
             rubric_model=_read_env_str("RUBRIC_MODEL") or None,
-            rubric_max_iterations=_read_env_int("RUBRIC_MAX_ITERATIONS", default=3),
+            rubric_max_iterations=_read_env_int("RUBRIC_MAX_ITERATIONS", default=None),
             sandbox_type=_read_env_str("SANDBOX_TYPE"),
             sandbox_id=_read_env_str("SANDBOX_ID"),
             sandbox_snapshot_name=_read_env_str("SANDBOX_SNAPSHOT_NAME") or None,
@@ -443,7 +454,7 @@ class ServerConfig:
         interpreter_ptc: str | list[str] | None = None,
         interpreter_ptc_acknowledge_unsafe: bool = False,
         rubric_model: str | None = None,
-        rubric_max_iterations: int = 3,
+        rubric_max_iterations: int | None = None,
         mcp_config_path: str | None,
         no_mcp: bool,
         trust_project_mcp: bool | None,
@@ -478,7 +489,8 @@ class ServerConfig:
             interpreter_ptc_acknowledge_unsafe: Mirror of
                 `settings.interpreter_ptc_acknowledge_unsafe`.
             rubric_model: Grader model spec; `None` reuses the main model.
-            rubric_max_iterations: Grader iterations per rubric attempt.
+            rubric_max_iterations: Explicit grader iterations per rubric attempt;
+                `None` uses the SDK default.
             mcp_config_path: Path to MCP config.
             no_mcp: Disable MCP.
             trust_project_mcp: Trust project MCP servers.

deepagents_code-0.1.30/deepagents_code/_testing_models.py

@@ -0,0 +1,317 @@
+"""Internal chat models used by local integration tests."""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any
+
+from langchain_core.language_models.fake_chat_models import GenericFakeChatModel
+from langchain_core.messages import AIMessage, BaseMessage
+from langchain_core.outputs import ChatGeneration, ChatResult
+from pydantic import Field
+
+if TYPE_CHECKING:
+    from collections.abc import Callable, Sequence
+
+    from langchain_core.callbacks import CallbackManagerForLLMRun
+    from langchain_core.language_models import LanguageModelInput
+    from langchain_core.runnables import Runnable
+    from langchain_core.tools import BaseTool
+
+
+# Prompt markers that drive `ToolCallingIntegrationChatModel`. Each marker is the
+# full token (including the trailing `=`); the file path follows on the same line,
+# e.g. `DCA_TEST_WRITE_FILE=/tmp/out.txt`. These are the single source of truth
+# shared with the integration tests, so the model and tests cannot drift apart.
+DCA_TEST_WRITE_FILE_MARKER = "DCA_TEST_WRITE_FILE="
+DCA_TEST_DELEGATE_WRITE_MARKER = "DCA_TEST_DELEGATE_WRITE="
+DCA_SUBAGENT_WRITE_FILE_MARKER = "DCA_SUBAGENT_WRITE_FILE="
+
+# Distinct file contents per write path, so a test asserting on file content can
+# prove which branch executed — in particular that subagent mode delegated through
+# the `task` tool rather than writing directly.
+TOP_LEVEL_WRITE_CONTENT = "auto-approved"
+SUBAGENT_WRITE_CONTENT = "auto-approved-subagent"
+
+
+class DeterministicIntegrationChatModel(GenericFakeChatModel):
+    """Deterministic chat model for integration tests.
+
+    This subclasses LangChain's `GenericFakeChatModel` so the implementation
+    stays aligned with the core fake-chat-model test surface, while overriding
+    generation to remain prompt-driven and restart-safe for real CLI server
+    integration tests.
+
+    Why the existing `langchain_core` fakes cannot be reused here:
+
+    1. Every core fake (`GenericFakeChatModel`, `FakeListChatModel`,
+        `FakeMessagesListChatModel`) pops from an iterator or cycles an index —
+        the actual prompt is ignored. App integration tests start and stop the
+        server process, which resets in-memory state. An iterator-based model
+        either raises `StopIteration` or replays from the beginning after a
+        restart, producing wrong or missing responses. This model derives output
+        solely from the prompt text, so identical input always produces
+        identical output regardless of process lifecycle.
+
+    2. The agent runtime calls `model.bind_tools(schemas)` during
+        initialization. None of the core fakes implement `bind_tools`, so they
+        raise `AttributeError` in any agent-loop context. This model provides a
+        no-op passthrough.
+
+    3. The app server reads `model.profile` for capability negotiation (e.g.
+        `tool_calling`, `max_input_tokens`). Core fakes have no such attribute,
+        causing `AttributeError` or silent misconfiguration at runtime.
+
+    Additionally, the compact middleware issues summarization prompts mid-
+    conversation. A list-based model cannot distinguish these from normal user
+    turns without pre-knowledge of exact call ordering, whereas this model
+    detects summary requests by inspecting the prompt content.
+    """
+
+    model: str = "fake"
+    # Required by `GenericFakeChatModel`, but our override does not consume it.
+    messages: object = Field(default_factory=lambda: iter(()))
+    profile: dict[str, Any] | None = Field(
+        default_factory=lambda: {
+            "tool_calling": True,
+            "max_input_tokens": 8000,
+        }
+    )
+
+    def bind_tools(
+        self,
+        tools: Sequence[dict[str, Any] | type | Callable | BaseTool],  # noqa: ARG002
+        *,
+        tool_choice: str | None = None,  # noqa: ARG002
+        **kwargs: Any,  # noqa: ARG002
+    ) -> Runnable[LanguageModelInput, AIMessage]:
+        """Return self so the agent can bind tool schemas during tests."""
+        return self
+
+    def _generate(
+        self,
+        messages: list[BaseMessage],
+        stop: list[str] | None = None,  # noqa: ARG002
+        run_manager: CallbackManagerForLLMRun | None = None,  # noqa: ARG002
+        **kwargs: Any,  # noqa: ARG002
+    ) -> ChatResult:
+        """Produce a deterministic reply derived from the prompt text.
+
+        Returns:
+            A single-message `ChatResult` with deterministic content.
+        """
+        prompt = "\n".join(
+            text
+            for message in messages
+            if (text := self._stringify_message(message)).strip()
+        )
+        if self._looks_like_summary_request(prompt):
+            content = "integration summary"
+        else:
+            excerpt = " ".join(prompt.split()[-18:])
+            if excerpt:
+                content = f"integration reply: {excerpt}"
+            else:
+                content = "integration reply"
+
+        return ChatResult(
+            generations=[ChatGeneration(message=AIMessage(content=content))]
+        )
+
+    @property
+    def _llm_type(self) -> str:
+        """LangChain model type identifier."""
+        return "deterministic-integration"
+
+    @staticmethod
+    def _stringify_message(message: BaseMessage) -> str:
+        """Flatten message content into plain text for deterministic responses.
+
+        Returns:
+            Plain-text content extracted from the message.
+        """
+        content = message.content
+        if isinstance(content, str):
+            return content
+        if isinstance(content, list):
+            parts: list[str] = []
+            for block in content:
+                if isinstance(block, str):
+                    parts.append(block)
+                elif isinstance(block, dict) and block.get("type") == "text":
+                    text = block.get("text")
+                    if isinstance(text, str):
+                        parts.append(text)
+            return " ".join(parts)
+        return str(content)
+
+    @staticmethod
+    def _looks_like_summary_request(prompt: str) -> bool:
+        """Detect the middleware's summary-generation prompt.
+
+        Returns:
+            `True` when the prompt appears to be a summarization request.
+        """
+        lowered = prompt.lower()
+        return (
+            "messages to summarize" in lowered
+            or "condense the following conversation" in lowered
+            or "<summary>" in lowered
+        )
+
+
+def _extract_marker_path(prompt: str, marker: str) -> str:
+    """Extract the file path that follows a prompt marker on the same line.
+
+    Args:
+        prompt: The flattened prompt text.
+        marker: The marker token (including its trailing `=`) to locate.
+
+    Returns:
+        The stripped file path immediately following the marker.
+
+    Raises:
+        ValueError: If the marker is present but not followed by a path, so a
+            malformed test prompt fails loudly here instead of silently
+            degrading to a `"done"` reply or raising an opaque `IndexError`.
+    """
+    _, _, tail = prompt.partition(marker)
+    lines = tail.splitlines()
+    file_path = lines[0].strip() if lines else ""
+    if not file_path:
+        msg = (
+            f"Test model saw marker {marker!r} but found no file path after it; "
+            f"check the integration-test prompt construction."
+        )
+        raise ValueError(msg)
+    return file_path
+
+
+def _tool_call_result(name: str, args: dict[str, Any], call_id: str) -> ChatResult:
+    """Build a single-tool-call `ChatResult`.
+
+    Returns:
+        A `ChatResult` wrapping an `AIMessage` with exactly one tool call.
+    """
+    return ChatResult(
+        generations=[
+            ChatGeneration(
+                message=AIMessage(
+                    content="",
+                    tool_calls=[
+                        {
+                            "name": name,
+                            "args": args,
+                            "id": call_id,
+                            "type": "tool_call",
+                        }
+                    ],
+                )
+            )
+        ]
+    )
+
+
+class ToolCallingIntegrationChatModel(DeterministicIntegrationChatModel):
+    """Deterministic tool-calling model for auto-approve integration tests.
+
+    Generation is driven entirely by prompt markers (the module-level `DCA_*`
+    constants), so output is restart-safe and independent of call ordering — the
+    same rationale as the parent `DeterministicIntegrationChatModel`:
+
+    - `DCA_TEST_WRITE_FILE=<path>` emits a top-level `write_file` call.
+    - `DCA_TEST_DELEGATE_WRITE=<path>` emits a `task` call delegating to the
+      `general-purpose` subagent, whose prompt then carries
+      `DCA_SUBAGENT_WRITE_FILE=<path>` to trigger the subagent's `write_file`.
+    - `DCA_SUBAGENT_WRITE_FILE=<path>` emits the subagent's `write_file` call.
+
+    Each marker fires only on the agent's first turn (no prior `ToolMessage`),
+    so once the tool result returns the model replies with a plain `"done"` and
+    the agent loop terminates instead of re-issuing the tool call.
+    """
+
+    # Only `_generate` is overridden; the inherited `_stream` would bypass this
+    # marker dispatch entirely, so streaming must be disabled.
+    disable_streaming: bool = True
+
+    def _generate(
+        self,
+        messages: list[BaseMessage],
+        stop: list[str] | None = None,  # noqa: ARG002
+        run_manager: CallbackManagerForLLMRun | None = None,  # noqa: ARG002
+        **kwargs: Any,  # noqa: ARG002
+    ) -> ChatResult:
+        """Emit a deterministic tool call (or terminal reply) from prompt markers.
+
+        The `has_tool_result` guard ensures each marker fires only on the
+        agent's first turn; once a `ToolMessage` is present the model returns
+        `"done"` so the agent loop terminates.
+
+        A recognized marker with no file path raises `ValueError` (via
+        `_extract_marker_path`) rather than degrading to `"done"`.
+
+        Returns:
+            A single-message `ChatResult`: an `AIMessage` carrying a `task` or
+            `write_file` tool call when a marker matches and no tool result is
+            present yet, otherwise a plain `"done"` reply.
+        """
+        prompt = "\n".join(
+            text
+            for message in messages
+            if (text := self._stringify_message(message)).strip()
+        )
+        has_tool_result = any(message.type == "tool" for message in messages)
+        if not has_tool_result:
+            for marker, build_tool_call in self._marker_dispatch():
+                if marker in prompt:
+                    name, args, call_id = build_tool_call(
+                        _extract_marker_path(prompt, marker)
+                    )
+                    return _tool_call_result(name, args, call_id)
+
+        return ChatResult(
+            generations=[ChatGeneration(message=AIMessage(content="done"))]
+        )
+
+    @staticmethod
+    def _marker_dispatch() -> tuple[
+        tuple[str, Callable[[str], tuple[str, dict[str, Any], str]]], ...
+    ]:
+        """Return ordered `(marker, tool-call builder)` pairs.
+
+        The delegate marker is checked before the plain write markers because
+        its emitted `task` description embeds `DCA_SUBAGENT_WRITE_FILE=`; the
+        first marker found in the prompt wins.
+
+        Returns:
+            Marker-to-builder pairs in precedence order. Each builder maps an
+            extracted file path to a `(tool_name, args, call_id)` triple.
+        """
+        return (
+            (
+                DCA_TEST_DELEGATE_WRITE_MARKER,
+                lambda path: (
+                    "task",
+                    {
+                        "description": f"{DCA_SUBAGENT_WRITE_FILE_MARKER}{path}",
+                        "subagent_type": "general-purpose",
+                    },
+                    "call_task",
+                ),
+            ),
+            (
+                DCA_SUBAGENT_WRITE_FILE_MARKER,
+                lambda path: (
+                    "write_file",
+                    {"file_path": path, "content": SUBAGENT_WRITE_CONTENT},
+                    "call_write_file",
+                ),
+            ),
+            (
+                DCA_TEST_WRITE_FILE_MARKER,
+                lambda path: (
+                    "write_file",
+                    {"file_path": path, "content": TOP_LEVEL_WRITE_CONTENT},
+                    "call_write_file",
+                ),
+            ),
+        )

{deepagents_code-0.1.28 → deepagents_code-0.1.30}/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__ = "0.1.28"  # x-release-please-version
+__version__ = "0.1.30"  # x-release-please-version
 
 DOCS_URL = "https://docs.langchain.com/oss/python/deepagents/code"
 """URL for `deepagents-code` documentation."""