fix: empty response handling, /no_think model gating, per-model profiles

- Detect empty LLM responses (no content, no tool calls) instead of
  silently treating them as task completion. Retries once without tools
  before warning the user.
- Gate /no_think system message and chat_template_kwargs to Qwen/QwQ
  models only — sending /no_think to llama3.x caused empty responses.
- Add model_profiles config section for per-model overrides (token
  budget, thinking, temperature, max_tokens) matched by name prefix.
  Applied at startup and on /model switch.
- Update SessionManager on /model switch so session files record the
  correct model.
- Add NDJSON fallback in SSE stream parser for Ollama compatibility.
- Improve read_file error to suggest find_files on FileNotFoundError.
- Add diagnostic logging for empty streams and empty results.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

app/agent/loop.py

@@ -105,17 +105,25 @@ class AgentLoop:
             )
         return prompt
 
+    # Models whose chat templates understand /no_think directives.
+    _THINKING_MODEL_PREFIXES = ("qwen", "qwq")
+
+    def _model_supports_no_think(self) -> bool:
+        """Check if the current model uses a thinking chat template."""
+        model_lower = self._config.llm.model.lower()
+        return any(model_lower.startswith(p) for p in self._THINKING_MODEL_PREFIXES)
+
     def _get_messages_with_system_prompt(self) -> list[Message]:
         """Prepend the system prompt to conversation history.
 
-        When thinking is disabled, appends a system-level /no_think directive
+        When thinking is disabled on a model that supports it, appends a
-        after the last user message so Qwen 3.x (and similar) chat templates
+        system-level /no_think directive after the last user message so
-        see it, without polluting the user's actual message content.
+        Qwen 3.x (and similar) chat templates see it.
         """
         system_msg = Message(role="system", content=self._system_prompt)
         history = self._ctx.get_history()
 
-        if not self._config.llm.thinking and history:
+        if not self._config.llm.thinking and self._model_supports_no_think() and history:
             history = list(history)
             # Find last user message and insert a system hint after it
             for i in range(len(history) - 1, -1, -1):
@@ -140,6 +148,7 @@ class AgentLoop:
 
         max_iter = self._config.agent.max_iterations
         reasoning_only_streak = 0
+        empty_streak = 0
         for iteration in range(1, max_iter + 1):
             if self._cancelled:
                 if self._display:
@@ -230,6 +239,36 @@ class AgentLoop:
             # Successful response — reset streak
             reasoning_only_streak = 0
 
+            # Detect completely empty response (no content, no tool calls)
+            if not assistant_msg.content and not assistant_msg.tool_calls:
+                empty_streak += 1
+                self._ctx.pop_last_message()  # Don't keep empty messages
+                if empty_streak >= 2:
+                    if self._display:
+                        self._display.write_warning(
+                            "Model returned repeated empty responses — "
+                            "try a different model or check Ollama logs."
+                        )
+                    break
+                if self._display:
+                    self._display.write_warning("Model returned empty response. Retrying without tools...")
+                # Retry without tool schemas — some models return empty when
+                # tools are in the payload but the model can't handle them.
+                assistant_msg = await self._llm_step(skip_tools=True)
+                if assistant_msg is None:
+                    break
+                if assistant_msg.content:
+                    self._ctx.add_message("assistant", assistant_msg.content)
+                    if self._display:
+                        self._display.write_assistant_message(assistant_msg.content)
+                    self._handler.reset()
+                    break
+                # Still empty even without tools
+                self._handler.reset()
+                continue
+
+            empty_streak = 0  # reset on successful non-empty response
+
             # Display any assistant text content (even if tool calls follow)
             if self._display and assistant_msg.content:
                 self._display.write_assistant_message(assistant_msg.content)
@@ -263,21 +302,25 @@ class AgentLoop:
             if self._display:
                 self._display.write_warning(f"Agent reached maximum iterations ({max_iter}). Stopping.")
 
-    async def _llm_step(self) -> Message | None:
+    async def _llm_step(self, *, skip_tools: bool = False) -> Message | None:
         """Stream one LLM response and return the accumulated Message.
 
         Uses retry-enabled streaming. On mid-stream errors, attempts to recover
         partial content if available.
 
+        Args:
+            skip_tools: If True, send the request without tool schemas (fallback mode).
+
         Returns:
             The assistant Message, or None if an error occurred.
         """
         messages = self._get_messages_with_system_prompt()
         if self._debug:
             self._debug.log_request(messages, self._config.llm.model)
+        tools = None if skip_tools else self._tools_schema
         t0 = time.monotonic()
         try:
-            chunk_iter = self._client.stream_chat_with_retry(messages, tools=self._tools_schema)
+            chunk_iter = self._client.stream_chat_with_retry(messages, tools=tools)
             result = await self._handler.process_stream(chunk_iter)
             if result and self._debug:
                 elapsed = (time.monotonic() - t0) * 1000

app/models/config.py

@@ -17,6 +17,23 @@ class AgentMode(StrEnum):
     AUTO = "auto"
 
 
+class ModelProfile(BaseModel):
+    """Per-model overrides applied when switching models."""
+
+    max_conversation_tokens: int | None = Field(
+        default=None, description="Token budget override for this model's context window"
+    )
+    thinking: bool | None = Field(
+        default=None, description="Override thinking mode for this model"
+    )
+    temperature: float | None = Field(
+        default=None, description="Override sampling temperature"
+    )
+    max_tokens: int | None = Field(
+        default=None, description="Override max response tokens"
+    )
+
+
 class LLMConfig(BaseModel):
     """LLM backend configuration."""
 
@@ -73,11 +90,19 @@ class ShellToolConfig(BaseModel):
     max_output_bytes: int = Field(default=65536, description="Max output capture size in bytes")
 
 
+class FileCacheConfig(BaseModel):
+    """File cache configuration."""
+
+    enabled: bool = Field(default=True, description="Enable file content caching")
+    max_entries: int = Field(default=128, description="Maximum cached file entries (LRU eviction)")
+
+
 class FilesystemToolConfig(BaseModel):
     """Filesystem tool limits."""
 
     max_file_size_bytes: int = Field(default=1_048_576, description="Max file size for read/write")
     binary_detection: bool = Field(default=True, description="Detect and reject binary files")
+    cache: FileCacheConfig = Field(default_factory=FileCacheConfig, description="File cache settings")
 
 
 class ToolsConfig(BaseModel):
@@ -137,6 +162,10 @@ class AppConfig(BaseModel):
     session: SessionConfig = Field(default_factory=SessionConfig)
     debug: DebugConfig = Field(default_factory=DebugConfig)
     skills: SkillsConfig = Field(default_factory=SkillsConfig)
+    model_profiles: dict[str, ModelProfile] = Field(
+        default_factory=dict,
+        description="Per-model overrides keyed by model name prefix",
+    )
 
     @model_validator(mode="after")
     def resolve_workspace_root(self) -> "AppConfig":
@@ -144,6 +173,39 @@ class AppConfig(BaseModel):
         self.agent.workspace_root = self.agent.workspace_root.resolve()
         return self
 
+    def get_model_profile(self, model: str) -> ModelProfile | None:
+        """Find the best matching model profile by prefix.
+
+        Matches the longest prefix first (e.g., "llama3.1" beats "llama3"
+        for model "llama3.1:latest"). Returns None if no profile matches.
+        """
+        model_lower = model.lower().split(":")[0]  # strip tag
+        best_match: str | None = None
+        for key in self.model_profiles:
+            key_lower = key.lower()
+            if model_lower == key_lower or model_lower.startswith(key_lower):
+                if best_match is None or len(key) > len(best_match):
+                    best_match = key
+        return self.model_profiles.get(best_match) if best_match else None
+
+    def apply_model_profile(self, model: str) -> ModelProfile | None:
+        """Apply the matching model profile overrides to the active config.
+
+        Returns the applied profile, or None if no profile matched.
+        """
+        profile = self.get_model_profile(model)
+        if profile is None:
+            return None
+        if profile.max_conversation_tokens is not None:
+            self.agent.max_conversation_tokens = profile.max_conversation_tokens
+        if profile.thinking is not None:
+            self.llm.thinking = profile.thinking
+        if profile.temperature is not None:
+            self.llm.temperature = profile.temperature
+        if profile.max_tokens is not None:
+            self.llm.max_tokens = profile.max_tokens
+        return profile
+
 
 # Default config file location relative to project root
 _DEFAULT_CONFIG_PATH = Path("config/config.yaml")

app/services/llm.py

@@ -151,8 +151,9 @@ class LLMClient:
         if tools:
             payload["tools"] = tools
 
-        # When thinking is disabled, inject chat_template_kwargs for backends that support it
+        # When thinking is disabled, inject chat_template_kwargs for backends
-        if not self._config.thinking:
+        # that support it (Qwen 3.x thinking models).
+        if not self._config.thinking and self._config.model.lower().startswith(("qwen", "qwq")):
             payload.setdefault("chat_template_kwargs", {})["enable_thinking"] = False
 
         # Merge model-specific extra parameters (e.g., reasoning_effort)
@@ -170,20 +171,32 @@ class LLMClient:
                         status_code=response.status_code,
                     )
 
+                chunk_count = 0
                 async for line in response.aiter_lines():
-                    if not line.startswith("data: "):
+                    line = line.strip()
+                    if not line:
                         continue
 
-                    data = line[6:]  # strip "data: " prefix
+                    # SSE format: "data: {json}" or "data: [DONE]"
-
+                    if line.startswith("data: "):
-                    if data.strip() == "[DONE]":
+                        data = line[6:]
-                        return
+                        if data.strip() == "[DONE]":
+                            break
+                    elif line.startswith("{"):
+                        # Plain NDJSON fallback (some Ollama versions)
+                        data = line
+                    else:
+                        continue
 
                     try:
                         yield json.loads(data)
+                        chunk_count += 1
                     except json.JSONDecodeError:
                         logger.warning("malformed_sse_chunk", data=data[:200])
 
+                if chunk_count == 0:
+                    logger.warning("empty_stream", model=self._config.model)
+
         except httpx.ConnectError as e:
             raise LLMConnectionError(f"Cannot connect to LLM endpoint: {e}") from e
         except httpx.TimeoutException as e:

app/services/session.py

@@ -52,6 +52,10 @@ class SessionManager:
         self._session_dir = workspace_root / config.session_dir
         self._session_id = f"{self._workspace_hash}_{datetime.now(UTC).strftime('%Y%m%d_%H%M%S')}"
 
+    def update_model(self, model: str) -> None:
+        """Update the model name for session metadata."""
+        self._model = model
+
     def save(self, ctx: "SessionContext") -> Path:
         """Save session state to a JSON file via atomic write.
 

app/services/streaming.py

@@ -60,8 +60,10 @@ class StreamHandler:
         """
         thinking_notified = False
         last_update_time = 0.0
+        chunk_count = 0
 
         async for chunk in chunk_iter:
+            chunk_count += 1
             self._process_chunk(chunk)
 
             if not self._display_config.stream_output:
@@ -96,6 +98,14 @@ class StreamHandler:
             self._on_done()
 
         tool_calls = self._build_tool_calls() or None
+
+        if chunk_count > 0 and not self._accumulated_content and not tool_calls:
+            logger.debug(
+                "stream_empty_result",
+                chunks_received=chunk_count,
+                had_reasoning=bool(self._accumulated_reasoning),
+            )
+
         return Message(
             role="assistant",
             content=self._accumulated_content or None,
@@ -183,11 +193,8 @@ class StreamHandler:
         return bool(self._accumulated_reasoning) and not self._accumulated_content and not self._tool_calls
 
     def reset(self) -> None:
-        """Clear all accumulators for the next turn."""
+        """Clear accumulators for the next LLM call, preserving UI callbacks."""
         self._accumulated_content = ""
         self._accumulated_reasoning = ""
         self._tool_calls.clear()
         self._usage = None
-        self._on_content = None
-        self._on_thinking = None
-        self._on_done = None

app/tools/edit.py

@@ -1,13 +1,17 @@
 """Edit tools: str_replace and patch_apply."""
 
+from __future__ import annotations
+
 import subprocess
 from pathlib import Path
 from typing import Any
 
 from pydantic import BaseModel, Field
 
+from app.models.config import AppConfig
 from app.models.tool_call import ToolResult, ToolResultStatus
 from app.tools.base import BaseTool
+from app.utils.file_cache import FileCache, cached_read_file
 from app.utils.file_helpers import (
     FileSizeError,
     PathSecurityError,
@@ -37,6 +41,12 @@ class StrReplaceTool(BaseTool):
     )
     params_model = StrReplaceParams
 
+    def __init__(
+        self, workspace_root: Path, config: AppConfig, file_cache: FileCache | None = None
+    ) -> None:
+        super().__init__(workspace_root, config)
+        self._file_cache = file_cache
+
     def execute(
         self, *, tool_call_id: str, file_path: str, old_str: str, new_str: str, **kwargs: Any
     ) -> ToolResult:
@@ -44,11 +54,12 @@ class StrReplaceTool(BaseTool):
 
         # Read the file
         try:
-            content = safe_read_file(
+            content = cached_read_file(
                 file_path,
                 self.workspace_root,
                 max_size_bytes=fs_config.max_file_size_bytes,
                 check_binary=fs_config.binary_detection,
+                cache=self._file_cache,
             )
         except PathSecurityError as exc:
             return ToolResult(
@@ -117,8 +128,14 @@ class StrReplaceTool(BaseTool):
             safe_path = resolve_safe_path(file_path, self.workspace_root)
             rel_path = safe_path.relative_to(self.workspace_root)
         except (PathSecurityError, ValueError):
+            safe_path = None
             rel_path = Path(file_path)
 
+        # Pre-warm cache with the new content (we already have it in memory).
+        if self._file_cache is not None and safe_path is not None:
+            self._file_cache.invalidate(safe_path)
+            self._file_cache.put(safe_path, new_content)
+
         return ToolResult(
             tool_call_id=tool_call_id,
             tool_name=self.name,
@@ -144,6 +161,12 @@ class PatchApplyTool(BaseTool):
     )
     params_model = PatchApplyParams
 
+    def __init__(
+        self, workspace_root: Path, config: AppConfig, file_cache: FileCache | None = None
+    ) -> None:
+        super().__init__(workspace_root, config)
+        self._file_cache = file_cache
+
     def execute(self, *, tool_call_id: str, file_path: str, patch: str, **kwargs: Any) -> ToolResult:
         try:
             safe_path = resolve_safe_path(file_path, self.workspace_root)
@@ -195,6 +218,9 @@ class PatchApplyTool(BaseTool):
                 error=f"Patch failed (exit {result.returncode}): {result.stderr or result.stdout}",
             )
 
+        if self._file_cache is not None:
+            self._file_cache.invalidate(safe_path)
+
         try:
             rel_path = safe_path.relative_to(self.workspace_root)
         except ValueError:

app/tools/filesystem.py

@@ -1,12 +1,16 @@
 """Filesystem tools: read_file, list_dir, write_file, make_dir, delete_file."""
 
+from __future__ import annotations
+
 from pathlib import Path
 from typing import Any
 
 from pydantic import BaseModel, Field
 
+from app.models.config import AppConfig
 from app.models.tool_call import ToolResult, ToolResultStatus
 from app.tools.base import BaseTool
+from app.utils.file_cache import FileCache, cached_read_file
 from app.utils.file_helpers import (
     BinaryFileError,
     FileSizeError,
@@ -23,6 +27,12 @@ class ReadFileParams(BaseModel):
     file_path: str = Field(description="Path to the file to read (relative to workspace root)")
 
 
+class ReadManyFilesParams(BaseModel):
+    """Parameters for the read_many_files tool."""
+
+    file_paths: list[str] = Field(description="List of file paths to read (relative to workspace root)")
+
+
 class ReadFileTool(BaseTool):
     """Read the contents of a file within the workspace."""
 
@@ -30,14 +40,22 @@ class ReadFileTool(BaseTool):
     description = "Read the full contents of a text file. Returns the file content as a string."
     params_model = ReadFileParams
 
+    def __init__(
+        self, workspace_root: Path, config: AppConfig, file_cache: FileCache | None = None
+    ) -> None:
+        super().__init__(workspace_root, config)
+        self._file_cache = file_cache
+
     def execute(self, *, tool_call_id: str, file_path: str, **kwargs: Any) -> ToolResult:
         fs_config = self.config.tools.filesystem
+        hits_before = self._file_cache.stats.hits if self._file_cache else 0
         try:
-            content = safe_read_file(
+            content = cached_read_file(
                 file_path,
                 self.workspace_root,
                 max_size_bytes=fs_config.max_file_size_bytes,
                 check_binary=fs_config.binary_detection,
+                cache=self._file_cache,
             )
         except PathSecurityError as exc:
             return ToolResult(
@@ -47,11 +65,12 @@ class ReadFileTool(BaseTool):
                 error=str(exc),
             )
         except FileNotFoundError as exc:
+            filename = Path(file_path).name
             return ToolResult(
                 tool_call_id=tool_call_id,
                 tool_name=self.name,
                 status=ToolResultStatus.ERROR,
-                error=str(exc),
+                error=f"{exc}. Use find_files to locate it, e.g. find_files(pattern=\"{filename}\")",
             )
         except FileSizeError as exc:
             return ToolResult(
@@ -68,6 +87,23 @@ class ReadFileTool(BaseTool):
                 error=str(exc),
             )
 
+        # On cache hit the file is unchanged — its content is already in
+        # conversation context from the earlier read, so avoid resending it.
+        was_cache_hit = (
+            self._file_cache is not None
+            and self._file_cache.stats.hits > hits_before
+        )
+        if was_cache_hit:
+            return ToolResult(
+                tool_call_id=tool_call_id,
+                tool_name=self.name,
+                status=ToolResultStatus.SUCCESS,
+                output=(
+                    f"[Cached] {file_path} is unchanged since last read "
+                    f"({len(content):,} chars). Content is already in conversation context."
+                ),
+            )
+
         return ToolResult(
             tool_call_id=tool_call_id,
             tool_name=self.name,
@@ -76,6 +112,76 @@ class ReadFileTool(BaseTool):
         )
 
 
+class ReadManyFilesTool(BaseTool):
+    """Read contents of multiple files at once."""
+
+    name = "read_many_files"
+    description = (
+        "Read contents of multiple files at once. Returns each file's content "
+        "prefixed with its path header."
+    )
+    params_model = ReadManyFilesParams
+
+    def __init__(
+        self, workspace_root: Path, config: AppConfig, file_cache: FileCache | None = None
+    ) -> None:
+        super().__init__(workspace_root, config)
+        self._file_cache = file_cache
+
+    def execute(self, *, tool_call_id: str, file_paths: list[str], **kwargs: Any) -> ToolResult:
+        if not file_paths:
+            return ToolResult(
+                tool_call_id=tool_call_id,
+                tool_name=self.name,
+                status=ToolResultStatus.ERROR,
+                error="file_paths list is empty",
+            )
+
+        fs_config = self.config.tools.filesystem
+        sections: list[str] = []
+        success_count = 0
+
+        for fp in file_paths:
+            hits_before = self._file_cache.stats.hits if self._file_cache else 0
+            try:
+                content = cached_read_file(
+                    fp,
+                    self.workspace_root,
+                    max_size_bytes=fs_config.max_file_size_bytes,
+                    check_binary=fs_config.binary_detection,
+                    cache=self._file_cache,
+                )
+                was_hit = (
+                    self._file_cache is not None
+                    and self._file_cache.stats.hits > hits_before
+                )
+                if was_hit:
+                    sections.append(
+                        f"=== {fp} ===\n[Cached] Unchanged since last read "
+                        f"({len(content):,} chars). Already in conversation context."
+                    )
+                else:
+                    sections.append(f"=== {fp} ===\n{content}")
+                success_count += 1
+            except (PathSecurityError, FileNotFoundError, FileSizeError, BinaryFileError) as exc:
+                sections.append(f"=== {fp} ===\n[ERROR] {exc}")
+
+        if success_count == 0:
+            return ToolResult(
+                tool_call_id=tool_call_id,
+                tool_name=self.name,
+                status=ToolResultStatus.ERROR,
+                error="All files failed to read:\n" + "\n".join(sections),
+            )
+
+        return ToolResult(
+            tool_call_id=tool_call_id,
+            tool_name=self.name,
+            status=ToolResultStatus.SUCCESS,
+            output="\n".join(sections),
+        )
+
+
 class ListDirParams(BaseModel):
     """Parameters for the list_dir tool."""
 
@@ -167,6 +273,12 @@ class WriteFileTool(BaseTool):
     )
     params_model = WriteFileParams
 
+    def __init__(
+        self, workspace_root: Path, config: AppConfig, file_cache: FileCache | None = None
+    ) -> None:
+        super().__init__(workspace_root, config)
+        self._file_cache = file_cache
+
     def execute(self, *, tool_call_id: str, file_path: str, content: str, **kwargs: Any) -> ToolResult:
         fs_config = self.config.tools.filesystem
         try:
@@ -191,6 +303,9 @@ class WriteFileTool(BaseTool):
                 error=str(exc),
             )
 
+        if self._file_cache is not None:
+            self._file_cache.invalidate(safe_path)
+
         try:
             rel_path = safe_path.relative_to(self.workspace_root)
         except ValueError:
@@ -272,6 +387,12 @@ class DeleteFileTool(BaseTool):
     description = "Delete a single file. Does not delete directories."
     params_model = DeleteFileParams
 
+    def __init__(
+        self, workspace_root: Path, config: AppConfig, file_cache: FileCache | None = None
+    ) -> None:
+        super().__init__(workspace_root, config)
+        self._file_cache = file_cache
+
     def execute(self, *, tool_call_id: str, file_path: str, **kwargs: Any) -> ToolResult:
         try:
             safe_path = resolve_safe_path(file_path, self.workspace_root)
@@ -309,6 +430,9 @@ class DeleteFileTool(BaseTool):
                 error=f"Failed to delete file: {exc}",
             )
 
+        if self._file_cache is not None:
+            self._file_cache.invalidate(safe_path)
+
         try:
             rel_path = safe_path.relative_to(self.workspace_root)
         except ValueError:

app/tools/registry.py

@@ -8,6 +8,7 @@ from typing import TYPE_CHECKING, Any
 
 from app.models.config import AppConfig
 from app.tools.base import BaseTool
+from app.utils.file_cache import FileCache
 
 if TYPE_CHECKING:
     from app.services.skills import SkillsManager
@@ -89,6 +90,7 @@ def create_default_registry(
     config: AppConfig,
     skills_manager: SkillsManager | None = None,
     skill_runner: object | None = None,
+    file_cache: FileCache | None = None,
 ) -> ToolRegistry:
     """Create a ToolRegistry populated with all built-in tools.
 
@@ -97,9 +99,10 @@ def create_default_registry(
         config: Application configuration.
         skills_manager: Optional skills manager for skill tools.
         skill_runner: Optional SkillRunner for package skill activation.
+        file_cache: Optional file cache shared across file-reading tools.
     """
     # Read tools
-    from app.tools.filesystem import ListDirTool, ReadFileTool
+    from app.tools.filesystem import ListDirTool, ReadFileTool, ReadManyFilesTool
 
     # Write tools
     from app.tools.filesystem import DeleteFileTool, MakeDirTool, WriteFileTool
@@ -119,7 +122,8 @@ def create_default_registry(
     registry = ToolRegistry()
 
     # Read
-    registry.register(ReadFileTool(workspace_root, config))
+    registry.register(ReadFileTool(workspace_root, config, file_cache=file_cache))
+    registry.register(ReadManyFilesTool(workspace_root, config, file_cache=file_cache))
     registry.register(ListDirTool(workspace_root, config))
 
     # Search
@@ -127,13 +131,13 @@ def create_default_registry(
     registry.register(FindFilesTool(workspace_root, config))
 
     # Write
-    registry.register(WriteFileTool(workspace_root, config))
+    registry.register(WriteFileTool(workspace_root, config, file_cache=file_cache))
     registry.register(MakeDirTool(workspace_root, config))
-    registry.register(DeleteFileTool(workspace_root, config))
+    registry.register(DeleteFileTool(workspace_root, config, file_cache=file_cache))
 
     # Edit
-    registry.register(StrReplaceTool(workspace_root, config))
+    registry.register(StrReplaceTool(workspace_root, config, file_cache=file_cache))
-    registry.register(PatchApplyTool(workspace_root, config))
+    registry.register(PatchApplyTool(workspace_root, config, file_cache=file_cache))
 
     # Shell
     registry.register(RunCommandTool(workspace_root, config))

app/ui/app.py

@@ -75,6 +75,9 @@ class SneakyCodeApp(App):
         """Initialize agent components after the app is mounted."""
         setup_logging_for_tui()
 
+        # Apply model profile for the initial model before creating context
+        self._config.apply_model_profile(self._config.llm.model)
+
         self._ctx = SessionContext(self._config)
 
         # Create long-lived agent dependencies (reused across turns)
@@ -97,11 +100,20 @@ class SneakyCodeApp(App):
                 self._config.skills, self._config.agent.workspace_root
             )
 
+        # Create file cache if enabled
+        self._file_cache = None
+        fs_cache_cfg = self._config.tools.filesystem.cache
+        if fs_cache_cfg.enabled:
+            from app.utils.file_cache import FileCache
+
+            self._file_cache = FileCache(max_entries=fs_cache_cfg.max_entries)
+
         # Create tool registry (SkillRunner wired after registry exists)
         self._tool_registry = create_default_registry(
             self._config.agent.workspace_root,
             self._config,
             skills_manager=self._skills_manager,
+            file_cache=self._file_cache,
         )
 
         # Create SkillRunner and late-bind it to skill tools
@@ -189,8 +201,8 @@ class SneakyCodeApp(App):
             table.add_row("/history", "Show conversation history")
             table.add_row("/save", "Manually save session")
             table.add_row("/session", "Show session info (messages, tokens, start time)")
-            table.add_row("/models", "List available Ollama models")
+            table.add_row("/models, /model", "List available Ollama models")
-            table.add_row("/models <name>", "Switch to a different model")
+            table.add_row("/model <name>", "Switch to a different model")
             table.add_row("/mode", "Show current agent mode")
             table.add_row("/mode normal|plan|auto", "Switch agent mode")
             table.add_row("/skills", "List available skills")
@@ -222,7 +234,7 @@ class SneakyCodeApp(App):
                     f"Started: {self._ctx.start_time.isoformat()}",
                     style="cyan",
                 ))
-        elif cmd.startswith("/models"):
+        elif cmd.split()[0] in ("/models", "/model"):
             parts = command.split(maxsplit=1)
             if len(parts) == 1:
                 # List available models
@@ -246,9 +258,30 @@ class SneakyCodeApp(App):
             else:
                 new_model = parts[1].strip()
                 self._config.llm.model = new_model
+                if self._session_mgr:
+                    self._session_mgr.update_model(new_model)
+                # Apply model-specific profile overrides
+                profile = self._config.apply_model_profile(new_model)
+                if profile and self._ctx:
+                    # Update token budget if the profile overrides it
+                    self._ctx.token_counter.budget = self._config.agent.max_conversation_tokens
                 self.query_one(HeaderPanel).update_model(new_model)
-                log.write(Text(f"Switched to model: {new_model}", style="bold green"))
+                header = self.query_one(HeaderPanel)
-        elif cmd.startswith("/mode"):
+                header.update_tokens(
+                    self._ctx.estimated_tokens if self._ctx else 0,
+                    self._config.agent.max_conversation_tokens,
+                )
+                msg = f"Switched to model: {new_model}"
+                if profile:
+                    overrides = []
+                    if profile.max_conversation_tokens is not None:
+                        overrides.append(f"tokens={profile.max_conversation_tokens:,}")
+                    if profile.thinking is not None:
+                        overrides.append(f"thinking={'on' if profile.thinking else 'off'}")
+                    if overrides:
+                        msg += f" ({', '.join(overrides)})"
+                log.write(Text(msg, style="bold green"))
+        elif cmd.split()[0] == "/mode":
             parts = command.split(maxsplit=1)
             if len(parts) == 1:
                 current = self._permissions.mode

app/ui/widgets.py

@@ -26,7 +26,7 @@ class HeaderPanel(Static):
     HeaderPanel {
         dock: top;
         height: 1;
-        background: $accent;
+        background: darkcyan;
         color: $text;
         padding: 0 2;
     }

app/utils/file_cache.py

@@ -0,0 +1,185 @@
+"""File cache with LRU eviction and mtime-based invalidation."""
+
+from __future__ import annotations
+
+from collections import OrderedDict
+from dataclasses import dataclass, field
+from pathlib import Path
+
+from app.utils.file_helpers import (
+    BinaryFileError,
+    FileSizeError,
+    PathSecurityError,
+    check_file_size,
+    is_binary_file,
+    resolve_safe_path,
+)
+
+
+@dataclass(slots=True)
+class CacheEntry:
+    """A cached file's content and modification timestamp."""
+
+    content: str
+    mtime_ns: int
+
+
+@dataclass
+class CacheStats:
+    """Running statistics for a FileCache instance."""
+
+    hits: int = 0
+    misses: int = 0
+    invalidations: int = 0
+    evictions: int = 0
+
+    @property
+    def hit_rate(self) -> float:
+        """Return cache hit rate as a float between 0.0 and 1.0."""
+        total = self.hits + self.misses
+        if total == 0:
+            return 0.0
+        return self.hits / total
+
+
+class FileCache:
+    """LRU file-content cache with mtime-based invalidation.
+
+    Keyed by resolved absolute ``Path``.  Each lookup performs a cheap
+    ``stat()`` syscall to verify the file hasn't changed on disk — if the
+    nanosecond mtime differs the entry is evicted and the caller gets a
+    cache miss.
+
+    Not thread-safe (single-threaded agent loop).
+    """
+
+    def __init__(self, max_entries: int = 128) -> None:
+        self._max_entries = max_entries
+        self._entries: OrderedDict[Path, CacheEntry] = OrderedDict()
+        self._stats = CacheStats()
+
+    # -- public API --------------------------------------------------
+
+    def get(self, path: Path) -> str | None:
+        """Return cached content if *path* hasn't changed, else ``None``.
+
+        A ``stat()`` call checks ``st_mtime_ns``; on mismatch the stale
+        entry is silently removed.
+        """
+        entry = self._entries.get(path)
+        if entry is None:
+            self._stats.misses += 1
+            return None
+
+        try:
+            current_mtime_ns = path.stat().st_mtime_ns
+        except OSError:
+            # File gone — evict and miss.
+            self._remove(path)
+            self._stats.misses += 1
+            return None
+
+        if current_mtime_ns != entry.mtime_ns:
+            self._remove(path)
+            self._stats.invalidations += 1
+            self._stats.misses += 1
+            return None
+
+        # Cache hit — move to end (most-recently used).
+        self._entries.move_to_end(path)
+        self._stats.hits += 1
+        return entry.content
+
+    def put(self, path: Path, content: str) -> None:
+        """Store *content* for *path* with its current ``st_mtime_ns``.
+
+        Evicts the least-recently-used entry when over capacity.
+        """
+        try:
+            mtime_ns = path.stat().st_mtime_ns
+        except OSError:
+            # Can't stat — don't cache.
+            return
+
+        if path in self._entries:
+            # Update existing; move to end.
+            self._entries[path] = CacheEntry(content=content, mtime_ns=mtime_ns)
+            self._entries.move_to_end(path)
+        else:
+            self._entries[path] = CacheEntry(content=content, mtime_ns=mtime_ns)
+
+        # Evict LRU if over capacity.
+        while len(self._entries) > self._max_entries:
+            self._entries.popitem(last=False)
+            self._stats.evictions += 1
+
+    def invalidate(self, path: Path) -> None:
+        """Remove *path* from the cache if present."""
+        if path in self._entries:
+            del self._entries[path]
+            self._stats.invalidations += 1
+
+    def clear(self) -> None:
+        """Remove all entries."""
+        self._entries.clear()
+
+    @property
+    def stats(self) -> CacheStats:
+        """Return the running cache statistics."""
+        return self._stats
+
+    def __len__(self) -> int:
+        return len(self._entries)
+
+    # -- internals ---------------------------------------------------
+
+    def _remove(self, path: Path) -> None:
+        """Delete an entry without bumping invalidation stats."""
+        self._entries.pop(path, None)
+
+
+def cached_read_file(
+    path: str | Path,
+    workspace_root: Path,
+    max_size_bytes: int = 1_048_576,
+    check_binary: bool = True,
+    cache: FileCache | None = None,
+) -> str:
+    """Read a file with full security checks, using *cache* when available.
+
+    Security checks (path sandboxing, size limit, binary detection) run on
+    **every** call — only the ``Path.read_text()`` I/O is skipped on a cache
+    hit.
+
+    When *cache* is ``None`` this behaves identically to
+    :func:`~app.utils.file_helpers.safe_read_file`.
+
+    Raises:
+        PathSecurityError: If the path escapes the workspace.
+        FileSizeError: If the file is too large.
+        BinaryFileError: If the file is binary and *check_binary* is True.
+        FileNotFoundError: If the file does not exist.
+    """
+    safe_path = resolve_safe_path(path, workspace_root)
+
+    if not safe_path.exists():
+        raise FileNotFoundError(f"File not found: {safe_path}")
+
+    check_file_size(safe_path, max_size_bytes)
+
+    if check_binary and is_binary_file(safe_path):
+        raise BinaryFileError(f"File appears to be binary: {safe_path}")
+
+    # Try cache.
+    if cache is not None:
+        cached = cache.get(safe_path)
+        if cached is not None:
+            return cached
+
+    # Cache miss (or no cache) — read from disk.
+    content = safe_path.read_text(encoding="utf-8")
+
+    if cache is not None:
+        cache.put(safe_path, content)
+
+    return content

app/utils/token_counter.py

@@ -36,6 +36,11 @@ class TokenCounter:
         """The configured token budget."""
         return self._budget
 
+    @budget.setter
+    def budget(self, value: int) -> None:
+        """Update the token budget (e.g., when switching models)."""
+        self._budget = value
+
     @property
     def cumulative_usage(self) -> TokenUsage:
         """Cumulative token usage across all tracked calls."""

config/config.yaml

@@ -18,11 +18,24 @@ llm:
 
 agent:
   max_iterations: 25
-  max_conversation_tokens: 32000
+  max_conversation_tokens: 32000  # Default token budget (overridden by model_profiles)
   workspace_root: "."
   truncation_keep_recent: 10
   truncation_threshold: 0.85
 
+# Per-model overrides — matched by longest model name prefix.
+# Unset fields fall through to the defaults above.
+model_profiles:
+  llama3:
+    max_conversation_tokens: 120000
+    thinking: false
+  qwen:
+    max_conversation_tokens: 32000
+    thinking: false
+  qwq:
+    max_conversation_tokens: 32000
+    thinking: true
+
 permissions:
   auto_approve:
     - read_file
@@ -68,6 +81,9 @@ tools:
   filesystem:
     max_file_size_bytes: 1048576  # 1 MB
     binary_detection: true
+    cache:
+      enabled: true
+      max_entries: 128
 
 session:
   session_dir: ".sneakycode/sessions"

docs/tweaks.md

@@ -1,21 +1,12 @@
-Pressing up should cycle history like claude code.
+# UI Issues
+on /clear we need to reset the token counter in the header panel.
 
-~~Remove the user's input from output "you" - it's not needed.~~ Brought back as a condensed one-liner (first line + line count for multi-line).
+# Bugs
 
-Smart shell auto-approve: auto-approve `run_command` when the base executable is in the `allowed_commands` list and the full command doesn't match any `denied_commands` prefix. Only prompt the user for commands whose base executable is unlisted. Currently all shell commands prompt regardless, which is tedious for safe read-only commands like `git branch` or `ls`. The allow/deny lists in `ShellToolConfig` already define what's safe — the permissions service just needs to be shell-aware.
+# Improvements
+add -p to command line args so that the agent can run the prompt and return data directly via STDOUT
 
-Show a token count or some other display for when the model is "thinking" for a long period of time. I want a way for the user to know the model is working on it.
+# Open questions:
+How might we pass a directory to this app and have it use that directory as it's workspace so I don't have to copy files or do odd things to work in other directories.
 
-/models command to show models available and temporarly change models in the session
+How do we handle huge files not taking up so many tokens?
-
-pass a directory to the tool so that it uses that directory as it's root for commands.
-
-add a skills directory so we can prompt our own skills for the tool to use similar to Claude Code
-
-need not only a session log, but also a log of what the llm is thinking and how it's working somehow. I need a way to see behind the curtain.
-
-# Left of from Phase 7 of old roadmap - finish these first
-- Permission modal auto-approves (TODO: proper modal dialog)
-- Session resume auto-resumes (TODO: modal y/n)
-- LLM client cleanup on unmount not yet wired
-- No automated TUI tests (Textual's AppTest can be added later)

tests/unit/test_file_cache.py

@@ -0,0 +1,314 @@
+"""Tests for the file cache with LRU eviction and mtime invalidation."""
+
+import os
+import time
+from pathlib import Path
+
+import pytest
+
+from app.models.config import AppConfig, load_config
+from app.models.tool_call import ToolResultStatus
+from app.tools.filesystem import ReadFileTool, ReadManyFilesTool
+from app.utils.file_cache import CacheStats, FileCache, cached_read_file
+from app.utils.file_helpers import BinaryFileError, FileSizeError, PathSecurityError
+
+
+# ---------------------------------------------------------------------------
+# FileCache unit tests
+# ---------------------------------------------------------------------------
+
+
+class TestFileCache:
+    def test_put_and_get_roundtrip(self, tmp_path: Path) -> None:
+        cache = FileCache()
+        f = tmp_path / "hello.txt"
+        f.write_text("hello world")
+
+        cache.put(f, "hello world")
+        assert cache.get(f) == "hello world"
+
+    def test_get_returns_none_for_missing_key(self, tmp_path: Path) -> None:
+        cache = FileCache()
+        assert cache.get(tmp_path / "nope.txt") is None
+
+    def test_mtime_change_causes_miss(self, tmp_path: Path) -> None:
+        cache = FileCache()
+        f = tmp_path / "data.txt"
+        f.write_text("v1")
+        cache.put(f, "v1")
+
+        # Mutate the file so mtime changes
+        time.sleep(0.05)  # ensure mtime differs
+        f.write_text("v2")
+
+        assert cache.get(f) is None  # stale → miss
+        assert cache.stats.invalidations == 1
+
+    def test_lru_eviction_at_capacity(self, tmp_path: Path) -> None:
+        cache = FileCache(max_entries=3)
+        files = []
+        for i in range(4):
+            f = tmp_path / f"f{i}.txt"
+            f.write_text(f"content-{i}")
+            files.append(f)
+
+        # Fill cache to capacity
+        for f in files[:3]:
+            cache.put(f, f.read_text())
+        assert len(cache) == 3
+
+        # Adding a 4th evicts the LRU (files[0])
+        cache.put(files[3], files[3].read_text())
+        assert len(cache) == 3
+        assert cache.get(files[0]) is None  # evicted
+        assert cache.stats.evictions == 1
+
+        # files[1..3] still present
+        for f in files[1:]:
+            assert cache.get(f) is not None
+
+    def test_invalidate_removes_entry(self, tmp_path: Path) -> None:
+        cache = FileCache()
+        f = tmp_path / "rm.txt"
+        f.write_text("bye")
+        cache.put(f, "bye")
+        assert len(cache) == 1
+
+        cache.invalidate(f)
+        assert len(cache) == 0
+        assert cache.get(f) is None
+        assert cache.stats.invalidations == 1
+
+    def test_invalidate_noop_for_missing(self) -> None:
+        cache = FileCache()
+        cache.invalidate(Path("/nonexistent"))
+        assert cache.stats.invalidations == 0
+
+    def test_clear_empties_cache(self, tmp_path: Path) -> None:
+        cache = FileCache()
+        for i in range(5):
+            f = tmp_path / f"c{i}.txt"
+            f.write_text(str(i))
+            cache.put(f, str(i))
+        assert len(cache) == 5
+
+        cache.clear()
+        assert len(cache) == 0
+
+    def test_stats_accuracy(self, tmp_path: Path) -> None:
+        cache = FileCache(max_entries=2)
+        a = tmp_path / "a.txt"
+        b = tmp_path / "b.txt"
+        c = tmp_path / "c.txt"
+        a.write_text("a")
+        b.write_text("b")
+        c.write_text("c")
+
+        # Miss
+        cache.get(a)
+        assert cache.stats.misses == 1
+        assert cache.stats.hits == 0
+
+        # Put + hit
+        cache.put(a, "a")
+        cache.get(a)
+        assert cache.stats.hits == 1
+
+        # Fill + evict
+        cache.put(b, "b")
+        cache.put(c, "c")  # evicts a
+        assert cache.stats.evictions == 1
+
+    def test_hit_rate(self) -> None:
+        stats = CacheStats(hits=3, misses=1)
+        assert stats.hit_rate == pytest.approx(0.75)
+
+    def test_hit_rate_zero_total(self) -> None:
+        stats = CacheStats()
+        assert stats.hit_rate == 0.0
+
+    def test_file_deleted_after_caching(self, tmp_path: Path) -> None:
+        cache = FileCache()
+        f = tmp_path / "gone.txt"
+        f.write_text("here")
+        cache.put(f, "here")
+
+        f.unlink()
+        assert cache.get(f) is None  # stat fails → miss
+
+    def test_put_skips_when_stat_fails(self) -> None:
+        cache = FileCache()
+        cache.put(Path("/totally/nonexistent"), "data")
+        assert len(cache) == 0
+
+    def test_get_moves_to_end(self, tmp_path: Path) -> None:
+        """Accessing an entry makes it most-recently-used, protecting from eviction."""
+        cache = FileCache(max_entries=3)
+        files = []
+        for i in range(3):
+            f = tmp_path / f"lru{i}.txt"
+            f.write_text(f"c{i}")
+            files.append(f)
+            cache.put(f, f"c{i}")
+
+        # Touch files[0] to make it MRU
+        cache.get(files[0])
+
+        # Add a new entry — files[1] (LRU) should be evicted, not files[0]
+        extra = tmp_path / "extra.txt"
+        extra.write_text("x")
+        cache.put(extra, "x")
+
+        assert cache.get(files[0]) is not None  # protected by access
+        assert cache.get(files[1]) is None  # evicted
+
+
+# ---------------------------------------------------------------------------
+# cached_read_file tests
+# ---------------------------------------------------------------------------
+
+
+class TestCachedReadFile:
+    def test_without_cache_matches_safe_read(self, tmp_path: Path) -> None:
+        f = tmp_path / "plain.txt"
+        f.write_text("hello")
+        content = cached_read_file(f, tmp_path, cache=None)
+        assert content == "hello"
+
+    def test_populates_on_miss_returns_on_hit(self, tmp_path: Path) -> None:
+        cache = FileCache()
+        f = tmp_path / "cached.txt"
+        f.write_text("data")
+
+        # First call: miss → read from disk → populate cache
+        content1 = cached_read_file(f, tmp_path, cache=cache)
+        assert content1 == "data"
+        assert cache.stats.misses == 1
+        assert cache.stats.hits == 0
+
+        # Second call: hit → from cache
+        content2 = cached_read_file(f, tmp_path, cache=cache)
+        assert content2 == "data"
+        assert cache.stats.hits == 1
+
+    def test_security_checks_run_on_cached_path(self, tmp_path: Path) -> None:
+        cache = FileCache()
+        with pytest.raises(PathSecurityError):
+            cached_read_file("/etc/passwd", tmp_path, cache=cache)
+
+    def test_binary_check_runs_on_cached_path(self, tmp_path: Path) -> None:
+        cache = FileCache()
+        f = tmp_path / "bin.dat"
+        f.write_bytes(b"\x00binary\x00")
+        with pytest.raises(BinaryFileError):
+            cached_read_file(f, tmp_path, cache=cache)
+
+    def test_size_check_runs_on_cached_path(self, tmp_path: Path) -> None:
+        cache = FileCache()
+        f = tmp_path / "big.txt"
+        f.write_text("x" * 200)
+
+        # First read populates cache
+        cached_read_file(f, tmp_path, max_size_bytes=1000, cache=cache)
+
+        # Now make file too big on disk — security check should catch it
+        # even though content is cached
+        f.write_text("x" * 2000)
+        with pytest.raises(FileSizeError):
+            cached_read_file(f, tmp_path, max_size_bytes=1000, cache=cache)
+
+    def test_file_not_found(self, tmp_path: Path) -> None:
+        cache = FileCache()
+        with pytest.raises(FileNotFoundError):
+            cached_read_file(tmp_path / "nope.txt", tmp_path, cache=cache)
+
+
+# ---------------------------------------------------------------------------
+# Tool-level cache-hit dedup tests
+# ---------------------------------------------------------------------------
+
+
+@pytest.fixture
+def config() -> AppConfig:
+    return load_config()
+
+
+@pytest.fixture
+def tmp_workspace(tmp_path: Path, config: AppConfig) -> tuple[Path, AppConfig]:
+    config.agent.workspace_root = tmp_path
+    return tmp_path, config
+
+
+class TestReadFileToolCacheHit:
+    def test_first_read_returns_full_content(self, tmp_workspace: tuple[Path, AppConfig]) -> None:
+        ws, cfg = tmp_workspace
+        cache = FileCache()
+        (ws / "hello.txt").write_text("hello world")
+
+        tool = ReadFileTool(ws, cfg, file_cache=cache)
+        result = tool.run("tc-1", {"file_path": "hello.txt"})
+        assert result.status == ToolResultStatus.SUCCESS
+        assert result.output == "hello world"
+
+    def test_second_read_returns_cached_message(self, tmp_workspace: tuple[Path, AppConfig]) -> None:
+        ws, cfg = tmp_workspace
+        cache = FileCache()
+        (ws / "hello.txt").write_text("hello world")
+
+        tool = ReadFileTool(ws, cfg, file_cache=cache)
+        tool.run("tc-1", {"file_path": "hello.txt"})
+
+        result2 = tool.run("tc-2", {"file_path": "hello.txt"})
+        assert result2.status == ToolResultStatus.SUCCESS
+        assert "[Cached]" in result2.output
+        assert "hello.txt" in result2.output
+        assert "hello world" not in result2.output
+
+    def test_changed_file_returns_full_content_again(self, tmp_workspace: tuple[Path, AppConfig]) -> None:
+        ws, cfg = tmp_workspace
+        cache = FileCache()
+        f = ws / "data.txt"
+        f.write_text("v1")
+
+        tool = ReadFileTool(ws, cfg, file_cache=cache)
+        tool.run("tc-1", {"file_path": "data.txt"})
+
+        # Mutate file so mtime changes
+        time.sleep(0.05)
+        f.write_text("v2")
+
+        result2 = tool.run("tc-2", {"file_path": "data.txt"})
+        assert result2.status == ToolResultStatus.SUCCESS
+        assert result2.output == "v2"
+        assert "[Cached]" not in result2.output
+
+    def test_no_cache_always_returns_content(self, tmp_workspace: tuple[Path, AppConfig]) -> None:
+        ws, cfg = tmp_workspace
+        (ws / "hello.txt").write_text("hello")
+
+        tool = ReadFileTool(ws, cfg, file_cache=None)
+        r1 = tool.run("tc-1", {"file_path": "hello.txt"})
+        r2 = tool.run("tc-2", {"file_path": "hello.txt"})
+        assert r1.output == "hello"
+        assert r2.output == "hello"
+
+
+class TestReadManyFilesToolCacheHit:
+    def test_cached_files_get_short_message(self, tmp_workspace: tuple[Path, AppConfig]) -> None:
+        ws, cfg = tmp_workspace
+        cache = FileCache()
+        (ws / "a.txt").write_text("alpha")
+        (ws / "b.txt").write_text("bravo")
+
+        tool = ReadManyFilesTool(ws, cfg, file_cache=cache)
+
+        # First read — full content
+        r1 = tool.run("tc-1", {"file_paths": ["a.txt", "b.txt"]})
+        assert "alpha" in r1.output
+        assert "bravo" in r1.output
+
+        # Second read — cached messages
+        r2 = tool.run("tc-2", {"file_paths": ["a.txt", "b.txt"]})
+        assert "[Cached]" in r2.output
+        assert "alpha" not in r2.output
+        assert "bravo" not in r2.output

tests/unit/test_filesystem_read_many.py

@@ -0,0 +1,69 @@
+"""Tests for the read_many_files tool."""
+
+from pathlib import Path
+
+import pytest
+
+from app.models.config import AppConfig, load_config
+from app.models.tool_call import ToolResultStatus
+from app.tools.filesystem import ReadManyFilesTool
+
+
+@pytest.fixture
+def config() -> AppConfig:
+    return load_config()
+
+
+@pytest.fixture
+def tmp_workspace(tmp_path: Path, config: AppConfig) -> tuple[Path, AppConfig]:
+    """Create a temporary workspace for read_many_files tests."""
+    config.agent.workspace_root = tmp_path
+    return tmp_path, config
+
+
+class TestReadManyFilesTool:
+    def test_read_multiple_files(self, tmp_workspace: tuple[Path, AppConfig]) -> None:
+        ws, cfg = tmp_workspace
+        (ws / "a.txt").write_text("alpha")
+        (ws / "b.txt").write_text("bravo")
+        tool = ReadManyFilesTool(ws, cfg)
+        result = tool.run("tc-1", {"file_paths": ["a.txt", "b.txt"]})
+        assert result.status == ToolResultStatus.SUCCESS
+        assert "=== a.txt ===" in result.output
+        assert "alpha" in result.output
+        assert "=== b.txt ===" in result.output
+        assert "bravo" in result.output
+
+    def test_partial_failure(self, tmp_workspace: tuple[Path, AppConfig]) -> None:
+        ws, cfg = tmp_workspace
+        (ws / "exists.txt").write_text("hello")
+        tool = ReadManyFilesTool(ws, cfg)
+        result = tool.run("tc-2", {"file_paths": ["exists.txt", "missing.txt"]})
+        assert result.status == ToolResultStatus.SUCCESS
+        assert "hello" in result.output
+        assert "[ERROR]" in result.output
+        assert "=== missing.txt ===" in result.output
+
+    def test_all_files_fail(self, tmp_workspace: tuple[Path, AppConfig]) -> None:
+        ws, cfg = tmp_workspace
+        tool = ReadManyFilesTool(ws, cfg)
+        result = tool.run("tc-3", {"file_paths": ["no1.txt", "no2.txt"]})
+        assert result.status == ToolResultStatus.ERROR
+        assert "All files failed" in (result.error or "")
+
+    def test_empty_file_paths(self, tmp_workspace: tuple[Path, AppConfig]) -> None:
+        ws, cfg = tmp_workspace
+        tool = ReadManyFilesTool(ws, cfg)
+        result = tool.run("tc-4", {"file_paths": []})
+        assert result.status == ToolResultStatus.ERROR
+        assert "empty" in (result.error or "").lower()
+
+    def test_path_security_inline_error(self, tmp_workspace: tuple[Path, AppConfig]) -> None:
+        ws, cfg = tmp_workspace
+        (ws / "safe.txt").write_text("ok")
+        tool = ReadManyFilesTool(ws, cfg)
+        result = tool.run("tc-5", {"file_paths": ["safe.txt", "../../etc/passwd"]})
+        assert result.status == ToolResultStatus.SUCCESS
+        assert "ok" in result.output
+        assert "[ERROR]" in result.output
+        assert "outside" in result.output.lower()

tests/unit/test_tools.py

@@ -108,7 +108,7 @@ class TestToolRegistry:
         registry = create_default_registry(workspace, config)
         names = set(registry.get_all().keys())
         assert names == {
-            "read_file", "list_dir", "grep_files", "find_files",
+            "read_file", "read_many_files", "list_dir", "grep_files", "find_files",
             "write_file", "make_dir", "delete_file",
             "str_replace", "patch_apply",
             "run_command",
@@ -118,7 +118,7 @@ class TestToolRegistry:
     def test_schema_export(self, workspace: Path, config: AppConfig) -> None:
         registry = create_default_registry(workspace, config)
         schemas = registry.get_openai_tools_schema()
-        assert len(schemas) == 11
+        assert len(schemas) == 12
         assert all(s["type"] == "function" for s in schemas)