drydock-cli 2.6.7__tar.gz → 2.6.9__tar.gz

drydock_cli-2.6.9/.auto_release.lock

@@ -0,0 +1 @@
+3928923

{drydock_cli-2.6.7 → drydock_cli-2.6.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drydock-cli
-Version: 2.6.7
+Version: 2.6.9
 Summary: Drydock — nautical CLI coding agent
 Project-URL: Homepage, https://github.com/fbobe321/drydock
 Project-URL: Repository, https://github.com/fbobe321/drydock

{drydock_cli-2.6.7 → drydock_cli-2.6.9}/drydock/cli/textual_ui/app.py

@@ -394,10 +394,18 @@ class VibeApp(App):  # noqa: PLR0904
         input_widget.value = ""
 
         if self._agent_running:
-            # Queue the message instead of interrupting — process after current task
+            # Queue the message — join rapid-fire inputs (multi-line paste)
             if not hasattr(self, "_pending_messages"):
                 self._pending_messages: list[str] = []
-            self._pending_messages.append(value)
+            import time
+            now = time.time()
+            last_queue_time = getattr(self, "_last_queue_time", 0)
+            if now - last_queue_time < 0.5 and self._pending_messages:
+                # Rapid input (< 500ms apart) = likely paste — join with newline
+                self._pending_messages[-1] += "\n" + value
+            else:
+                self._pending_messages.append(value)
+            self._last_queue_time = now
             preview = value[:60] + "..." if len(value) > 60 else value
             self.notify(f"Queued: \"{preview}\" ({len(self._pending_messages)} pending)")
             return

{drydock_cli-2.6.7 → drydock_cli-2.6.9}/drydock/core/agent_loop.py

@@ -567,23 +567,34 @@ class AgentLoop:
                           or "maximum context" in error_str.lower()
                           or "400 bad request" in error_str.lower()
                           or "status: 400" in error_str.lower()):
-                        # Context limit or malformed request — aggressively compact
+                        # Context limit or malformed request — aggressive recovery
                         try:
-                            compacted = 0
+                            # First try: truncate old messages
                             for i, msg in enumerate(self.messages):
                                 if i >= len(self.messages) - 4:
-                                    break  # Keep last 4 messages
+                                    break
                                 if msg.role == Role.tool and hasattr(msg, 'content'):
                                     content = str(msg.content) if msg.content else ""
                                     if len(content) > 200:
-                                        msg.content = content[:100] + "\n[truncated]\n" + content[-50:]
-                                        compacted += 1
+                                        msg.content = content[:100] + "\n[truncated]"
                                 elif msg.role == Role.assistant and hasattr(msg, 'content'):
                                     content = str(msg.content) if msg.content else ""
                                     if len(content) > 500:
                                         msg.content = content[:200] + "\n[truncated]"
-                                        compacted += 1
-                            logger.info("Emergency compaction: truncated %d messages", compacted)
+
+                            # Second try: if messages > 20, keep only last 6
+                            if len(self.messages) > 20:
+                                first_user = None
+                                for msg in self.messages:
+                                    if msg.role == Role.user:
+                                        first_user = msg
+                                        break
+                                kept = []
+                                if first_user:
+                                    kept.append(first_user)
+                                kept.extend(self.messages[-5:])
+                                self.messages.reset(kept)
+                                logger.info("Emergency reset: kept first user + last 5 messages")
                         except Exception:
                             pass
                         error_text = "Context compacted due to API error. Continue with your task."
@@ -592,6 +603,8 @@ class AgentLoop:
                     self._inject_system_note(error_text)
                     continue
 
+                if not self.messages:
+                    continue
                 last_message = self.messages[-1]
 
                 # Track edits — no circuit breakers, just track for has_made_edit
@@ -625,6 +638,8 @@ class AgentLoop:
             if assistant_event.content:
                 yield assistant_event
 
+        if not self.messages:
+            return
         last_message = self.messages[-1]
 
         # Detect repetitive text generation (Gemma 4 sometimes loops text within one response)
@@ -752,7 +767,7 @@ class AgentLoop:
         - Other (bash with commands): block after 3
         """
         args_str = json.dumps(tool_call.args_dict, sort_keys=True, default=str)
-        sig = hashlib.md5(
+        sig = hashlib.sha256(
             f"{tool_call.tool_name}:{args_str}".encode()
         ).hexdigest()
 
@@ -804,7 +819,7 @@ class AgentLoop:
     def _circuit_breaker_record(self, tool_call: ResolvedToolCall, result_text: str) -> None:
         """Record a tool call execution for circuit breaker tracking."""
         args_str = json.dumps(tool_call.args_dict, sort_keys=True, default=str)
-        sig = hashlib.md5(
+        sig = hashlib.sha256(
             f"{tool_call.tool_name}:{args_str}".encode()
         ).hexdigest()
         count, _ = self._tool_call_history.get(sig, (0, ""))
@@ -1147,7 +1162,7 @@ class AgentLoop:
             self.messages.reset(cleaned)
 
         # Fix 2: If last message is assistant, add a user "Continue." prompt
-        if self.messages[-1].role == Role.assistant:
+        if self.messages and self.messages[-1].role == Role.assistant:
             self.messages.append(LLMMessage(role=Role.user, content="Continue."))
 
     async def _chat(self, max_tokens: int | None = None) -> LLMChunk:
@@ -1509,7 +1524,7 @@ class AgentLoop:
                             f"{tc.function.name}:{tc.function.arguments}"
                         )
                         tool_names.append(tc.function.name or "")
-                sig = hashlib.md5(
+                sig = hashlib.sha256(
                     "|".join(sorted(call_parts)).encode()
                 ).hexdigest()[:16]
                 sigs.append(sig)
@@ -1942,7 +1957,7 @@ class AgentLoop:
                             except (json.JSONDecodeError, TypeError):
                                 pass
                         call_parts.append(f"{fn_name}:{fn_args}")
-                sig = hashlib.md5("|".join(sorted(call_parts)).encode()).hexdigest()[:16]
+                sig = hashlib.sha256("|".join(sorted(call_parts)).encode()).hexdigest()[:16]
 
                 if sig in seen_sigs:
                     # Duplicate — mark for removal (assistant msg + next tool result)

{drydock_cli-2.6.7 → drydock_cli-2.6.9}/drydock/core/prompts/gemma4.md

@@ -1,6 +1,6 @@
 You are DryDock, a CLI coding agent. You write code, fix bugs, and build projects.
 
-ACT IMMEDIATELY. Do NOT ask for permission. Do NOT explain what you will do. Just do it.
+ACT IMMEDIATELY. Your FIRST response must be a tool call — not text. Do NOT explain, plan, or ask. Call a tool NOW.
 
 Your tools: read_file, write_file, search_replace, grep, glob, bash.
 

{drydock_cli-2.6.7 → drydock_cli-2.6.9}/drydock/core/tools/builtins/search_replace.py

@@ -241,14 +241,13 @@ class SearchReplace(
                 ">>>>>>> REPLACE"
             )
 
-        # Detect garbled token output (common with some models/parsers)
+        # Detect garbled token output — only flag obvious corruption
+        # (mixed angle brackets + repeated letters like <<t<ttt<t), not normal repeats
         for block in search_replace_blocks:
-            if re.search(r'(.)\1{4,}', block.search) or re.search(r'<{3,}[a-z]', block.search):
+            if re.search(r'<[a-z]<[a-z]{2,}<[a-z]', block.search):
                 raise ToolError(
-                    "Your search text appears garbled (repeated/corrupted characters). "
-                    "This happens when regex or special characters confuse the tool call parser. "
-                    "WORKAROUND: Use read_file to get the exact line numbers, then use write_file "
-                    "to rewrite the entire function instead of search_replace on complex regex patterns."
+                    "Your search text appears garbled (token corruption). "
+                    "Use write_file to rewrite the function instead of search_replace."
                 )
 
         return file_path, search_replace_blocks

{drydock_cli-2.6.7 → drydock_cli-2.6.9}/drydock/core/tools/manager.py

@@ -49,7 +49,7 @@ def _compute_module_name(path: Path) -> str:
         return canonical
 
     resolved = path.resolve()
-    path_hash = hashlib.md5(str(resolved).encode()).hexdigest()[:8]
+    path_hash = hashlib.sha256(str(resolved).encode()).hexdigest()[:8]
     stem = re.sub(r"[^0-9A-Za-z_]", "_", path.stem) or "mod"
     return f"vibe_tools_discovered_{stem}_{path_hash}"
 

{drydock_cli-2.6.7 → drydock_cli-2.6.9}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "drydock-cli"
-version = "2.6.7"
+version = "2.6.9"
 description = "Drydock — nautical CLI coding agent"
 readme = "README.md"
 requires-python = ">=3.12"

drydock_cli-2.6.9/tests/test_user_issues.py

@@ -0,0 +1,129 @@
+"""Tests for user-reported issues (2026-04-07).
+
+Each test should FAIL before the fix and PASS after.
+"""
+import hashlib
+import pytest
+from unittest.mock import patch, MagicMock
+
+
+class TestMD5Crash:
+    """#61: Crash after install — md5 error in manager.py."""
+
+    def test_md5_usedforsecurity(self):
+        """md5 should work with usedforsecurity=False."""
+        h = hashlib.md5(b"test", usedforsecurity=False)
+        assert h.hexdigest() == "098f6bcd4621d373cade4e832627b4f6"
+
+    def test_tool_manager_import(self):
+        """Tool manager should not crash on import."""
+        from drydock.core.tools.manager import ToolManager
+        assert ToolManager is not None
+
+
+class TestSearchReplaceGarbled:
+    """#55: search_replace garbled characters."""
+
+    def test_garbled_detection(self):
+        from drydock.core.tools.builtins.search_replace import SearchReplaceArgs
+        args = SearchReplaceArgs.model_validate({
+            "file_path": "test.py",
+            "content": "<<<<<<< SEARCH\nold text\n=======\nnew text\n>>>>>>> REPLACE",
+        })
+        assert args.file_path == "test.py"
+
+
+class TestTodoEnumQuotes:
+    """#60: Todo enum quoting issues."""
+
+    def test_double_quoted(self):
+        from drydock.core.tools.builtins.todo import TodoItem
+        item = TodoItem.model_validate({
+            "id": "1", "content": "test",
+            "status": '"completed"', "priority": '"high"',
+        })
+        assert item.status.value == "completed"
+
+    def test_single_quoted(self):
+        from drydock.core.tools.builtins.todo import TodoItem
+        item = TodoItem.model_validate({
+            "id": "1", "content": "test",
+            "status": "'completed'", "priority": "'high'",
+        })
+        assert item.status.value == "completed"
+
+
+class TestEmptyState:
+    """#58: IndexError on empty lists."""
+
+    def test_empty_tool_calls(self):
+        from drydock.core.types import LLMMessage, Role
+        msg = LLMMessage(role=Role.assistant, content="hello", tool_calls=None)
+        calls = msg.tool_calls or []
+        assert calls == []
+
+
+if __name__ == "__main__":
+    pytest.main([__file__, "-v"])
+
+
+class TestSearchReplaceGarbledAdvanced:
+    """#55: Garbled detection should only catch real corruption."""
+
+    def test_legitimate_repeats_not_flagged(self):
+        """Normal repeated chars like ===== should NOT be flagged."""
+        from drydock.core.tools.builtins.search_replace import SearchReplace
+        blocks = SearchReplace._parse_search_replace_blocks(
+            "<<<<<<< SEARCH\ndef foo():\n    x = 5\n=======\ndef foo():\n    x = 10\n>>>>>>> REPLACE"
+        )
+        assert len(blocks) == 1
+        assert "x = 5" in blocks[0].search
+
+    def test_real_garble_detected(self):
+        """Token corruption like <<t<ttt<t should be caught."""
+        import re
+        garbled = "(?P<<t<ttt<t<tststssts>\\d+)"
+        assert re.search(r'<[a-z]<[a-z]{2,}<[a-z]', garbled)
+
+    def test_normal_regex_not_garbled(self):
+        """Normal regex should not trigger garble detection."""
+        import re
+        normal = r"(?P<timestamp>\d{4}-\d{2}-\d{2})"
+        assert not re.search(r'<[a-z]<[a-z]{2,}<[a-z]', normal)
+
+
+class TestIndexErrorGuards:
+    """#58: Guard all [-1] accesses on potentially empty lists."""
+
+    def test_empty_messages_safe(self):
+        """Accessing last message on empty list should not crash."""
+        messages = []
+        last = messages[-1] if messages else None
+        assert last is None
+
+    def test_agent_loop_no_md5(self):
+        """Agent loop should use sha256, not md5."""
+        import inspect
+        from drydock.core.agent_loop import AgentLoop
+        source = inspect.getsource(AgentLoop)
+        assert "hashlib.md5" not in source, "Still using md5 in agent_loop"
+        assert "hashlib.sha256" in source or "sha256" not in source  # either uses sha256 or doesn't hash at all
+
+
+class TestAPIRecovery:
+    """#57: API error recovery should auto-compact."""
+
+    def test_message_list_reset_keeps_structure(self):
+        """Reset should produce valid message ordering."""
+        from drydock.core.types import MessageList, LLMMessage, Role
+        ml = MessageList()
+        for i in range(25):
+            ml.append(LLMMessage(role=Role.user, content=f"msg {i}"))
+            ml.append(LLMMessage(role=Role.assistant, content=f"reply {i}"))
+        
+        # Simulate emergency reset: keep first user + last 5
+        first_user = ml[0]
+        kept = [first_user] + list(ml[-5:])
+        ml.reset(kept)
+        assert len(ml) == 6
+        assert ml[0].role == Role.user

drydock_cli-2.6.7/.auto_release.lock

@@ -1 +0,0 @@
-3888135

drydock_cli-2.6.7/tests/test_user_issues.py

@@ -1,265 +0,0 @@
-"""Tests for user-reported issues.
-
-Each test should FAIL before the fix, proving the issue exists.
-After fixing, all should pass.
-"""
-
-from __future__ import annotations
-
-import inspect
-import os
-from pathlib import Path
-from types import SimpleNamespace
-from unittest.mock import patch
-
-import pytest
-
-
-# ============================================================================
-# Issue 1: "DryDock" branding — should be DryDock not Drydock everywhere
-# ============================================================================
-
-class TestBranding:
-    def test_readme_says_drydock_capital(self):
-        readme = Path(__file__).parent.parent / "README.md"
-        content = readme.read_text()
-        # Title should be "# DryDock" not "# Drydock"
-        assert "# DryDock" in content
-
-    def test_system_prompt_says_drydock(self):
-        prompt = Path(__file__).parent.parent / "drydock" / "core" / "prompts" / "cli.md"
-        content = prompt.read_text()
-        # Should say "DryDock" not "Drydock" as the product name
-        assert "DryDock" in content
-
-    def test_loading_widget_says_drydock(self):
-        from drydock.cli.textual_ui.widgets.loading import LoadingWidget
-        widget = LoadingWidget.__new__(LoadingWidget)
-        status = widget._get_default_status()
-        # Anchor emoji is fine, but product references should be DryDock
-        assert status  # Just verify it doesn't crash
-
-
-# ============================================================================
-# Issue 5: What conda env is pip using? Should preserve user's env
-# ============================================================================
-
-class TestCondaEnv:
-    def test_bash_env_preserves_active_conda(self):
-        """If user has CONDA_DEFAULT_ENV set, bash should preserve it."""
-        from drydock.core.tools.builtins.bash import _get_base_env
-        with patch.dict(os.environ, {"CONDA_DEFAULT_ENV": "myproject", "CONDA_EXE": "/opt/conda/bin/conda"}):
-            env = _get_base_env()
-            # BASH_ENV should reference the user's env, not just base
-            if "BASH_ENV" in env:
-                bash_env_content = Path(env["BASH_ENV"]).read_text() if Path(env["BASH_ENV"]).exists() else ""
-                # Should activate the user's env, not just source conda.sh
-                assert "myproject" in bash_env_content or "CONDA_DEFAULT_ENV" in str(env)
-
-
-# ============================================================================
-# Issue 6: Circuit breaker shows message but still repeats
-# ============================================================================
-
-class TestCircuitBreakerEnforcement:
-    def test_circuit_breaker_disabled(self):
-        """Circuit breaker is disabled — even failed commands are never blocked."""
-        from drydock.core.agent_loop import AgentLoop
-        from drydock.core.types import MessageList
-
-        al = object.__new__(AgentLoop)
-        al.messages = MessageList()
-        al._tool_call_history = {}
-
-        tc = SimpleNamespace(tool_name="bash", args_dict={"command": "bad_cmd"})
-        for _ in range(3):
-            al._circuit_breaker_record(tc, "FAILED: command not found")
-
-        # CB is disabled — always returns None
-        assert al._circuit_breaker_check(tc) is None
-
-    def test_circuit_breaker_resets_for_new_conversation(self):
-        """Circuit breaker history is clearable between conversations."""
-        from drydock.core.agent_loop import AgentLoop
-        from drydock.core.types import MessageList
-
-        al = object.__new__(AgentLoop)
-        al.messages = MessageList()
-        al._tool_call_history = {}
-
-        tc = SimpleNamespace(tool_name="bash", args_dict={"command": "bad_cmd"})
-        for _ in range(3):
-            al._circuit_breaker_record(tc, "FAILED: command not found")
-        # CB is disabled — returns None even before clearing
-        assert al._circuit_breaker_check(tc) is None
-
-        # Simulate new conversation — history should be clearable
-        al._tool_call_history.clear()
-        assert al._circuit_breaker_check(tc) is None  # Still None
-
-
-# ============================================================================
-# Issue 7: "Understood" after errors then stops
-# ============================================================================
-
-class TestUnderstoodBug:
-    def test_no_understood_in_agent(self):
-        """The agent should not inject 'Understood.' which causes premature stops."""
-        from drydock.core.agent_loop import AgentLoop
-        src = inspect.getsource(AgentLoop)
-        # "Understood." was causing the model to think conversation ended
-        assert "Understood" not in src, \
-            "'Understood' found in agent_loop — should use 'Continuing...' instead"
-
-
-# ============================================================================
-# Issue 10: Circuit breaker too sensitive — same command later should be OK
-# ============================================================================
-
-class TestCircuitBreakerSensitivity:
-    def test_successful_commands_never_blocked(self):
-        """Successful commands are never blocked."""
-        from drydock.core.agent_loop import AgentLoop
-        from drydock.core.types import MessageList
-
-        al = object.__new__(AgentLoop)
-        al.messages = MessageList()
-        al._tool_call_history = {}
-
-        tc = SimpleNamespace(tool_name="bash", args_dict={"command": "git status"})
-        for _ in range(10):
-            al._circuit_breaker_record(tc, "clean")
-        assert al._circuit_breaker_check(tc) is None  # Never blocked
-
-    def test_failed_commands_not_blocked_cb_disabled(self):
-        """CB disabled — failed commands are NOT blocked even after 3 repeats."""
-        from drydock.core.agent_loop import AgentLoop
-        from drydock.core.types import MessageList
-
-        al = object.__new__(AgentLoop)
-        al.messages = MessageList()
-        al._tool_call_history = {}
-
-        tc = SimpleNamespace(tool_name="bash", args_dict={"command": "git status"})
-        for _ in range(3):
-            al._circuit_breaker_record(tc, "FAILED: not a git repo")
-        assert al._circuit_breaker_check(tc) is None  # CB disabled
-
-
-# ============================================================================
-# Issue 11: /consult needs to show which model is being used
-# ============================================================================
-
-class TestConsultModelDisplay:
-    def test_consult_handler_shows_model_name(self):
-        """The /consult response should include the model name."""
-        from drydock.cli.textual_ui.app import VibeApp
-        src = inspect.getsource(VibeApp._consult_command)
-        # Should display which model is being used
-        assert "consultant_model" in src or "model" in src
-
-    def test_consult_help_shows_config_instructions(self):
-        """Typing /consult with no args should show how to configure."""
-        from drydock.cli.textual_ui.app import VibeApp
-        src = inspect.getsource(VibeApp._consult_command)
-        assert "config.toml" in src or "consultant_model" in src
-
-
-# ============================================================================
-# Issue 12: SSL error on websearch — should use -k / DRYDOCK_INSECURE
-# ============================================================================
-
-class TestWebSearchSSL:
-    def test_websearch_respects_insecure_flag(self):
-        """WebSearch should check DRYDOCK_INSECURE env var."""
-        from drydock.core.tools.builtins.websearch import WebSearch
-        src = inspect.getsource(WebSearch)
-        assert "DRYDOCK_INSECURE" in src or "insecure" in src.lower() or "verify" in src
-
-    def test_webfetch_respects_insecure_flag(self):
-        """WebFetch should check DRYDOCK_INSECURE env var."""
-        from drydock.core.tools.builtins.webfetch import WebFetch
-        src = inspect.getsource(WebFetch)
-        assert "DRYDOCK_INSECURE" in src
-
-
-# ============================================================================
-# Issue 14: Config.toml consultant_model setup
-# ============================================================================
-
-class TestConsultantConfig:
-    def test_config_has_consultant_model_field(self):
-        """VibeConfig should have consultant_model field."""
-        from drydock.core.config import VibeConfig
-        assert hasattr(VibeConfig, "model_fields")
-        assert "consultant_model" in VibeConfig.model_fields
-
-    def test_consultant_model_documented_in_readme(self):
-        """README should mention how to set consultant_model."""
-        readme = Path(__file__).parent.parent / "README.md"
-        content = readme.read_text()
-        assert "consultant_model" in content or "consultant" in content.lower()
-
-
-# ============================================================================
-# Issue 15: Auto-consult agent — model should ask consultant when stuck
-# ============================================================================
-
-class TestAutoConsult:
-    def test_circuit_breaker_disabled_returns_none(self):
-        """CB is disabled — returns None even for repeatedly failed commands."""
-        from drydock.core.agent_loop import AgentLoop
-        from drydock.core.types import MessageList
-
-        al = object.__new__(AgentLoop)
-        al.messages = MessageList()
-        al._tool_call_history = {}
-
-        tc = SimpleNamespace(tool_name="bash", args_dict={"command": "bad_cmd"})
-        for _ in range(3):
-            al._circuit_breaker_record(tc, "FAILED: command not found")
-        assert al._circuit_breaker_check(tc) is None
-
-
-# ============================================================================
-# Issue 16: /consult should include conversation history
-# ============================================================================
-
-class TestConsultHistory:
-    def test_consult_sends_context(self):
-        """/consult handler should send conversation context to consultant."""
-        from drydock.core.consultant import ask_consultant
-        src = inspect.getsource(ask_consultant)
-        # Should accept and use context/conversation history
-        assert "context" in src or "messages" in src or "history" in src
-
-
-# ============================================================================
-# Issue 17: Deep research skill should exist
-# ============================================================================
-
-class TestDeepResearchSkill:
-    def test_research_skill_exists(self):
-        """A deep-research skill should be bundled."""
-        skills_dir = Path(__file__).parent.parent / "drydock" / "skills"
-        skill_dirs = [d.name for d in skills_dir.iterdir() if d.is_dir()] if skills_dir.exists() else []
-        has_research = any("research" in d for d in skill_dirs)
-        assert has_research, f"No research skill found. Available: {skill_dirs}"
-
-
-# ============================================================================
-# Issue 18: Not using agents — explore/diagnostic agents should be utilized
-# ============================================================================
-
-class TestAgentsAvailable:
-    def test_explore_agent_registered(self):
-        """The 'explore' agent should be registered."""
-        from drydock.core.agents.models import BuiltinAgentName
-        assert hasattr(BuiltinAgentName, "EXPLORE") or "explore" in dir(BuiltinAgentName)
-
-    def test_system_prompt_mentions_agents(self):
-        """System prompt should tell the model about available agents."""
-        prompt = Path(__file__).parent.parent / "drydock" / "core" / "prompts" / "cli.md"
-        content = prompt.read_text()
-        has_agent_ref = "agent" in content.lower() or "subagent" in content.lower() or "task" in content.lower()
-        assert has_agent_ref