klaude-code 1.7.0__tar.gz → 1.7.1__tar.gz

{klaude_code-1.7.0 → klaude_code-1.7.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: klaude-code
-Version: 1.7.0
+Version: 1.7.1
 Summary: Minimal code agent CLI
 Requires-Dist: anthropic>=0.66.0
 Requires-Dist: chardet>=5.2.0
@@ -146,20 +146,62 @@ klaude config
 
 ##### Adding Models to Built-in Providers
 
-You can add custom models to existing providers without redefining the entire provider:
+You can add custom models to existing built-in providers without redefining the entire provider. Just reference the `provider_name` and add your `model_list`:
 
 ```yaml
-# Just specify provider_name and your new models - no need for protocol/api_key
+# ~/.klaude/klaude-config.yaml
 provider_list:
+  - provider_name: openrouter  # Reference existing built-in provider
+    model_list:
+      - model_name: seed
+        model_params:
+          model: bytedance-seed/seed-1.6  # Model ID from OpenRouter
+          context_limit: 262000
+          cost:
+            input: 0.25
+            output: 2
+```
+
+**How merging works:**
+- Your models are merged with the built-in models for that provider
+- You only need `provider_name` and `model_list` - protocol, api_key, etc. are inherited from the built-in config
+- To override a built-in model, use the same `model_name` (e.g., `sonnet` to customize the built-in sonnet)
+
+**More examples:**
+
+```yaml
+provider_list:
+  # Add multiple models to OpenRouter
   - provider_name: openrouter
     model_list:
-      - model_name: my-custom-model
+      - model_name: qwen-coder
+        model_params:
+          model: qwen/qwen-2.5-coder-32b-instruct
+          context_limit: 131072
+          cost:
+            input: 0.3
+            output: 0.9
+      - model_name: llama-405b
+        model_params:
+          model: meta-llama/llama-3.1-405b-instruct
+          context_limit: 131072
+          cost:
+            input: 0.8
+            output: 0.8
+
+  # Add models to Anthropic provider
+  - provider_name: anthropic
+    model_list:
+      - model_name: haiku@ant
         model_params:
-          model: some-provider/some-model-id
+          model: claude-3-5-haiku-20241022
           context_limit: 200000
+          cost:
+            input: 1.0
+            output: 5.0
 ```
 
-Your models are merged with built-in models. To override a built-in model, use the same `model_name`.
+After adding models, run `klaude list` to verify they appear in the model list.
 
 ##### Overriding Provider Settings
 
@@ -230,6 +272,8 @@ provider_list:
 - `openai` - OpenAI-compatible API
 - `responses` - OpenAI Responses API (for o-series, GPT-5, Codex)
 - `openrouter` - OpenRouter API
+- `google` - Google Gemini API
+- `bedrock` - AWS Bedrock (uses AWS credentials instead of api_key)
 - `codex` - OpenAI Codex CLI (OAuth-based)
 
 List configured providers and models:

{klaude_code-1.7.0 → klaude_code-1.7.1}/README.md

@@ -125,20 +125,62 @@ klaude config
 
 ##### Adding Models to Built-in Providers
 
-You can add custom models to existing providers without redefining the entire provider:
+You can add custom models to existing built-in providers without redefining the entire provider. Just reference the `provider_name` and add your `model_list`:
 
 ```yaml
-# Just specify provider_name and your new models - no need for protocol/api_key
+# ~/.klaude/klaude-config.yaml
 provider_list:
+  - provider_name: openrouter  # Reference existing built-in provider
+    model_list:
+      - model_name: seed
+        model_params:
+          model: bytedance-seed/seed-1.6  # Model ID from OpenRouter
+          context_limit: 262000
+          cost:
+            input: 0.25
+            output: 2
+```
+
+**How merging works:**
+- Your models are merged with the built-in models for that provider
+- You only need `provider_name` and `model_list` - protocol, api_key, etc. are inherited from the built-in config
+- To override a built-in model, use the same `model_name` (e.g., `sonnet` to customize the built-in sonnet)
+
+**More examples:**
+
+```yaml
+provider_list:
+  # Add multiple models to OpenRouter
   - provider_name: openrouter
     model_list:
-      - model_name: my-custom-model
+      - model_name: qwen-coder
+        model_params:
+          model: qwen/qwen-2.5-coder-32b-instruct
+          context_limit: 131072
+          cost:
+            input: 0.3
+            output: 0.9
+      - model_name: llama-405b
+        model_params:
+          model: meta-llama/llama-3.1-405b-instruct
+          context_limit: 131072
+          cost:
+            input: 0.8
+            output: 0.8
+
+  # Add models to Anthropic provider
+  - provider_name: anthropic
+    model_list:
+      - model_name: haiku@ant
         model_params:
-          model: some-provider/some-model-id
+          model: claude-3-5-haiku-20241022
           context_limit: 200000
+          cost:
+            input: 1.0
+            output: 5.0
 ```
 
-Your models are merged with built-in models. To override a built-in model, use the same `model_name`.
+After adding models, run `klaude list` to verify they appear in the model list.
 
 ##### Overriding Provider Settings
 
@@ -209,6 +251,8 @@ provider_list:
 - `openai` - OpenAI-compatible API
 - `responses` - OpenAI Responses API (for o-series, GPT-5, Codex)
 - `openrouter` - OpenRouter API
+- `google` - Google Gemini API
+- `bedrock` - AWS Bedrock (uses AWS credentials instead of api_key)
 - `codex` - OpenAI Codex CLI (OAuth-based)
 
 List configured providers and models:

{klaude_code-1.7.0 → klaude_code-1.7.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "uv_build"
 
 [project]
 name = "klaude-code"
-version = "1.7.0"
+version = "1.7.1"
 description = "Minimal code agent CLI"
 readme = "README.md"
 requires-python = ">=3.13"

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/cli/main.py

@@ -95,10 +95,20 @@ def read_input_content(cli_argument: str) -> str | None:
     return content
 
 
+ENV_HELP = """\
+Environment Variables:
+
+  KLAUDE_READ_GLOBAL_LINE_CAP  Max lines to read (default: 2000)
+
+  KLAUDE_READ_MAX_CHARS        Max total chars to read (default: 50000)
+"""
+
 app = typer.Typer(
     add_completion=False,
     pretty_exceptions_enable=False,
     no_args_is_help=False,
+    rich_markup_mode="rich",
+    epilog=ENV_HELP,
 )
 
 # Register subcommands from modules

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/cli/runtime.py

@@ -379,7 +379,7 @@ async def run_interactive(init_config: AppInitConfig, session_id: str | None = N
                 model_name=model_name,
                 save_as_default=False,
                 defer_thinking_selection=True,
-                emit_welcome_event=False,
+                emit_welcome_event=True,
                 emit_switch_message=False,
             )
         )
@@ -398,7 +398,7 @@ async def run_interactive(init_config: AppInitConfig, session_id: str | None = N
             op.ChangeThinkingOperation(
                 session_id=sid,
                 thinking=thinking,
-                emit_welcome_event=False,
+                emit_welcome_event=True,
                 emit_switch_message=False,
             )
         )

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/command/fork_session_cmd.py

@@ -7,6 +7,7 @@ from prompt_toolkit.styles import Style
 
 from klaude_code.command.command_abc import Agent, CommandABC, CommandResult
 from klaude_code.protocol import commands, events, model
+from klaude_code.ui.modes.repl.clipboard import copy_to_clipboard
 from klaude_code.ui.terminal.selector import SelectItem, select_one
 
 FORK_SELECT_STYLE = Style(
@@ -215,6 +216,9 @@ class ForkSessionCommand(CommandABC):
             new_session = agent.session.fork()
             await new_session.wait_for_flush()
 
+            resume_cmd = f"klaude --resume-by-id {new_session.id}"
+            copy_to_clipboard(resume_cmd)
+
             event = events.DeveloperMessageEvent(
                 session_id=agent.session.id,
                 item=model.DeveloperMessageItem(
@@ -247,6 +251,9 @@ class ForkSessionCommand(CommandABC):
         # Build result message
         fork_description = "entire conversation" if selected is None else f"up to message index {selected}"
 
+        resume_cmd = f"klaude --resume-by-id {new_session.id}"
+        copy_to_clipboard(resume_cmd)
+
         event = events.DeveloperMessageEvent(
             session_id=agent.session.id,
             item=model.DeveloperMessageItem(

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/config/assets/builtin_config.yaml

@@ -87,6 +87,30 @@ provider_list:
         input: 1.75
         output: 14.0
         cache_read: 0.17
+  - model_name: gpt-5.2-medium
+    model_params:
+      model: openai/gpt-5.2
+      max_tokens: 128000
+      context_limit: 400000
+      verbosity: high
+      thinking:
+        reasoning_effort: medium
+      cost:
+        input: 1.75
+        output: 14.0
+        cache_read: 0.17
+  - model_name: gpt-5.2-low
+    model_params:
+      model: openai/gpt-5.2
+      max_tokens: 128000
+      context_limit: 400000
+      verbosity: low
+      thinking:
+        reasoning_effort: low
+      cost:
+        input: 1.75
+        output: 14.0
+        cache_read: 0.17
   - model_name: gpt-5.2-fast
     model_params:
       model: openai/gpt-5.2

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/config/config.py

@@ -257,6 +257,11 @@ def get_example_config() -> UserConfig:
                             model="model-id-from-provider",
                             max_tokens=16000,
                             context_limit=200000,
+                            cost=llm_param.Cost(
+                                input=1,
+                                output=10,
+                                cache_read=0.1,
+                            ),
                         ),
                     ),
                 ],

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/const.py

@@ -4,8 +4,21 @@ This module consolidates all magic numbers and configuration values
 that were previously scattered across the codebase.
 """
 
+import os
 from pathlib import Path
 
+
+def _get_int_env(name: str, default: int) -> int:
+    """Get an integer value from environment variable, or return default."""
+    val = os.environ.get(name)
+    if val is None:
+        return default
+    try:
+        return int(val)
+    except ValueError:
+        return default
+
+
 # =============================================================================
 # Agent Configuration
 # =============================================================================
@@ -47,10 +60,12 @@ TODO_REMINDER_TOOL_CALL_THRESHOLD = 10
 READ_CHAR_LIMIT_PER_LINE = 2000
 
 # Maximum number of lines to read from a file
-READ_GLOBAL_LINE_CAP = 2000
+# Can be overridden via KLAUDE_READ_GLOBAL_LINE_CAP environment variable
+READ_GLOBAL_LINE_CAP = _get_int_env("KLAUDE_READ_GLOBAL_LINE_CAP", 2000)
 
 # Maximum total characters to read (truncates beyond this limit)
-READ_MAX_CHARS = 50000
+# Can be overridden via KLAUDE_READ_MAX_CHARS environment variable
+READ_MAX_CHARS = _get_int_env("KLAUDE_READ_MAX_CHARS", 50000)
 
 # Maximum image file size in bytes (4MB)
 READ_MAX_IMAGE_BYTES = 4 * 1024 * 1024

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/core/executor.py

@@ -9,6 +9,7 @@ from __future__ import annotations
 
 import asyncio
 import subprocess
+import sys
 from collections.abc import Callable
 from dataclasses import dataclass
 from pathlib import Path
@@ -427,14 +428,26 @@ class ExecutorContext:
         return build_export_html(agent.session, system_prompt, tool_schemas, model_name)
 
     def _open_file(self, path: Path) -> None:
+        # Select platform-appropriate command
+        if sys.platform == "darwin":
+            cmd = "open"
+        elif sys.platform == "win32":
+            cmd = "start"
+        else:
+            cmd = "xdg-open"
+
         try:
             # Detach stdin to prevent interference with prompt_toolkit's terminal state
-            subprocess.run(["open", str(path)], stdin=subprocess.DEVNULL, check=True)
+            if sys.platform == "win32":
+                # Windows 'start' requires shell=True
+                subprocess.run(f'start "" "{path}"', shell=True, stdin=subprocess.DEVNULL, check=True)
+            else:
+                subprocess.run([cmd, str(path)], stdin=subprocess.DEVNULL, check=True)
         except FileNotFoundError as exc:  # pragma: no cover
-            msg = "`open` command not found; please open the HTML manually."
+            msg = f"`{cmd}` command not found; please open the HTML manually."
             raise RuntimeError(msg) from exc
         except subprocess.CalledProcessError as exc:  # pragma: no cover
-            msg = f"Failed to open HTML with `open`: {exc}"
+            msg = f"Failed to open HTML with `{cmd}`: {exc}"
             raise RuntimeError(msg) from exc
 
     async def handle_interrupt(self, operation: op.InterruptOperation) -> None:

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/core/task.py

@@ -220,7 +220,9 @@ class TaskExecutor:
                         error_msg = f"Retrying {attempt + 1}/{const.MAX_FAILED_TURN_RETRIES} in {delay:.1f}s"
                         if last_error_message:
                             error_msg = f"{error_msg} - {last_error_message}"
-                        yield events.ErrorEvent(error_message=error_msg, can_retry=True)
+                        yield events.ErrorEvent(
+                            error_message=error_msg, can_retry=True, session_id=session_ctx.session_id
+                        )
                         await asyncio.sleep(delay)
                 finally:
                     self._current_turn = None
@@ -234,7 +236,7 @@ class TaskExecutor:
                 final_error = f"Turn failed after {const.MAX_FAILED_TURN_RETRIES} retries."
                 if last_error_message:
                     final_error = f"{last_error_message}\n{final_error}"
-                yield events.ErrorEvent(error_message=final_error, can_retry=False)
+                yield events.ErrorEvent(error_message=final_error, can_retry=False, session_id=session_ctx.session_id)
                 return
 
             if turn is None or turn.task_finished:
@@ -244,7 +246,7 @@ class TaskExecutor:
                         error_msg = "Sub-agent returned empty result, retrying..."
                     else:
                         error_msg = "Agent returned empty result, retrying..."
-                    yield events.ErrorEvent(error_message=error_msg, can_retry=True)
+                    yield events.ErrorEvent(error_message=error_msg, can_retry=True, session_id=session_ctx.session_id)
                     continue
                 break
 

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/core/tool/shell/command_safety.py

@@ -275,12 +275,10 @@ def _is_safe_argv(argv: list[str]) -> SafetyCheckResult:
             "tag",
             "clone",
             "worktree",
+            "push",
+            "pull",
+            "remote",
         }
-        # Block remote operations
-        blocked_git_cmds = {"push", "pull", "remote"}
-
-        if sub in blocked_git_cmds:
-            return SafetyCheckResult(False, f"git: Remote operation '{sub}' not allowed")
         if sub not in allowed_git_cmds:
             return SafetyCheckResult(False, f"git: Subcommand '{sub}' not in allow list")
         return SafetyCheckResult(True)

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/protocol/events.py

@@ -16,6 +16,7 @@ class EndEvent(BaseModel):
 class ErrorEvent(BaseModel):
     error_message: str
     can_retry: bool = False
+    session_id: str | None = None
 
 
 class TaskStartEvent(BaseModel):

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/session/export.py

@@ -308,13 +308,17 @@ def _try_render_todo_args(arguments: str, tool_name: str) -> str | None:
         return None
 
 
-def _render_sub_agent_result(content: str) -> str:
+def _render_sub_agent_result(content: str, description: str | None = None) -> str:
     # Try to format as JSON for better readability
     try:
         parsed = json.loads(content)
         formatted = "```json\n" + json.dumps(parsed, ensure_ascii=False, indent=2) + "\n```"
     except (json.JSONDecodeError, TypeError):
         formatted = content
+
+    if description:
+        formatted = f"# {description}\n\n{formatted}"
+
     encoded = _escape_html(formatted)
     return (
         f'<div class="sub-agent-result-container">'
@@ -628,7 +632,15 @@ def _format_tool_call(tool_call: model.ToolCallItem, result: model.ToolResultIte
 
         if result.output and not should_hide_text:
             if is_sub_agent_tool(tool_call.name):
-                items_to_render.append(_render_sub_agent_result(result.output))
+                description = None
+                try:
+                    args = json.loads(tool_call.arguments)
+                    if isinstance(args, dict):
+                        typed_args = cast(dict[str, Any], args)
+                        description = cast(str | None, typed_args.get("description"))
+                except (json.JSONDecodeError, TypeError):
+                    pass
+                items_to_render.append(_render_sub_agent_result(result.output, description))
             else:
                 items_to_render.append(_render_text_block(result.output))
 

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/session/session.py

@@ -62,6 +62,7 @@ class Session(BaseModel):
     need_todo_not_used_cooldown_counter: int = Field(exclude=True, default=0)
 
     _messages_count_cache: int | None = PrivateAttr(default=None)
+    _user_messages_cache: list[str] | None = PrivateAttr(default=None)
     _store: JsonlSessionStore = PrivateAttr(default_factory=get_default_store)
 
     @property
@@ -78,6 +79,20 @@ class Session(BaseModel):
     def _invalidate_messages_count_cache(self) -> None:
         self._messages_count_cache = None
 
+    @property
+    def user_messages(self) -> list[str]:
+        """All user message contents in this session.
+
+        This is used for session selection UI and search, and is also persisted
+        in meta.json to avoid scanning events.jsonl for every session.
+        """
+
+        if self._user_messages_cache is None:
+            self._user_messages_cache = [
+                it.content for it in self.conversation_history if isinstance(it, model.UserMessageItem) and it.content
+            ]
+        return self._user_messages_cache
+
     @staticmethod
     def _project_key() -> str:
         return _project_key_from_cwd()
@@ -178,6 +193,18 @@ class Session(BaseModel):
         self.conversation_history.extend(items)
         self._invalidate_messages_count_cache()
 
+        new_user_messages = [
+            it.content for it in items if isinstance(it, model.UserMessageItem) and it.content
+        ]
+        if new_user_messages:
+            if self._user_messages_cache is None:
+                # Build from full history once to ensure correctness when resuming older sessions.
+                self._user_messages_cache = [
+                    it.content for it in self.conversation_history if isinstance(it, model.UserMessageItem) and it.content
+                ]
+            else:
+                self._user_messages_cache.extend(new_user_messages)
+
         if self.created_at <= 0:
             self.created_at = time.time()
         self.updated_at = time.time()
@@ -188,6 +215,7 @@ class Session(BaseModel):
             sub_agent_state=self.sub_agent_state,
             file_tracker=self.file_tracker,
             todos=list(self.todos),
+            user_messages=self.user_messages,
             created_at=self.created_at,
             updated_at=self.updated_at,
             messages_count=self.messages_count,
@@ -311,7 +339,7 @@ class Session(BaseModel):
                 case model.DeveloperMessageItem() as dm:
                     yield events.DeveloperMessageEvent(session_id=self.id, item=dm)
                 case model.StreamErrorItem() as se:
-                    yield events.ErrorEvent(error_message=se.error, can_retry=False)
+                    yield events.ErrorEvent(error_message=se.error, can_retry=False, session_id=self.id)
                 case _:
                     continue
             prev_item = it
@@ -378,6 +406,17 @@ class Session(BaseModel):
                 pass
             return messages
 
+        def _maybe_backfill_user_messages(*, meta_path: Path, meta: dict[str, Any], user_messages: list[str]) -> None:
+            if isinstance(meta.get("user_messages"), list):
+                return
+            meta["user_messages"] = user_messages
+            try:
+                tmp_path = meta_path.with_suffix(".json.tmp")
+                tmp_path.write_text(json.dumps(meta, ensure_ascii=False, indent=2), encoding="utf-8")
+                tmp_path.replace(meta_path)
+            except OSError:
+                return
+
         items: list[Session.SessionMetaBrief] = []
         for meta_path in store.iter_meta_files():
             data = _read_json_dict(meta_path)
@@ -390,7 +429,15 @@ class Session(BaseModel):
             created = float(data.get("created_at", meta_path.stat().st_mtime))
             updated = float(data.get("updated_at", meta_path.stat().st_mtime))
             work_dir = str(data.get("work_dir", ""))
-            user_messages = _get_user_messages(sid)
+
+            user_messages_raw = data.get("user_messages")
+            if isinstance(user_messages_raw, list) and all(
+                isinstance(m, str) for m in cast(list[object], user_messages_raw)
+            ):
+                user_messages = cast(list[str], user_messages_raw)
+            else:
+                user_messages = _get_user_messages(sid)
+                _maybe_backfill_user_messages(meta_path=meta_path, meta=data, user_messages=user_messages)
             messages_count = int(data.get("messages_count", -1))
             model_name = data.get("model_name") if isinstance(data.get("model_name"), str) else None
 

{klaude_code-1.7.0 → klaude_code-1.7.1}/src/klaude_code/session/store.py

@@ -193,6 +193,7 @@ def build_meta_snapshot(
     sub_agent_state: model.SubAgentState | None,
     file_tracker: dict[str, model.FileStatus],
     todos: list[model.TodoItem],
+    user_messages: list[str],
     created_at: float,
     updated_at: float,
     messages_count: int,
@@ -206,6 +207,8 @@ def build_meta_snapshot(
         "sub_agent_state": sub_agent_state.model_dump(mode="json") if sub_agent_state else None,
         "file_tracker": {path: status.model_dump(mode="json") for path, status in file_tracker.items()},
         "todos": [todo.model_dump(mode="json", exclude_defaults=True) for todo in todos],
+        # Cache user messages to avoid scanning events.jsonl during session listing.
+        "user_messages": list(user_messages),
         "created_at": created_at,
         "updated_at": updated_at,
         "messages_count": messages_count,