drydock-cli 0.6.0__tar.gz → 0.6.1__tar.gz

{drydock_cli-0.6.0 → drydock_cli-0.6.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drydock-cli
-Version: 0.6.0
+Version: 0.6.1
 Summary: Drydock — nautical CLI coding agent
 Project-URL: Homepage, https://github.com/fbobe321/drydock
 Project-URL: Repository, https://github.com/fbobe321/drydock

{drydock_cli-0.6.0 → drydock_cli-0.6.1}/drydock/core/agent_loop.py

@@ -835,20 +835,57 @@ class AgentLoop:
                 )
             )
 
-    def _circuit_breaker_check(self, tool_call: ResolvedToolCall) -> str | None:
-        """Block exact-duplicate tool calls. Returns cached result or None."""
+    def _get_attempted_summary(self) -> str:
+        """Build a summary of what the agent has already tried."""
+        if not self._tool_call_history:
+            return ""
+        attempts = []
+        for sig, (count, result) in self._tool_call_history.items():
+            if count >= 2:
+                attempts.append(f"  - Ran {count}x, result: {result[:80]}")
+        if attempts:
+            return "ALREADY ATTEMPTED (do NOT repeat):\n" + "\n".join(attempts[:10])
+        return ""
+
+    async def _circuit_breaker_check(self, tool_call: ResolvedToolCall) -> str | None:
+        """Block exact-duplicate tool calls. Returns cached result or None.
+        On 3rd attempt, auto-consults the smarter model if available."""
         sig = hashlib.md5(
             f"{tool_call.tool_name}:{tool_call.raw_arguments}".encode()
         ).hexdigest()
 
         count, last_result = self._tool_call_history.get(sig, (0, ""))
         if count >= 2:
-            return (
-                f"CIRCUIT BREAKER: You already ran this exact command {count} times "
-                f"and got the same result. Previous result: {last_result[:300]}\n\n"
-                f"Do NOT repeat. Try a DIFFERENT approach, different arguments, "
-                f"or a different tool."
+            attempted = self._get_attempted_summary()
+            msg = (
+                f"CIRCUIT BREAKER: You already ran `{tool_call.tool_name}` with these "
+                f"exact arguments {count} times and got the same result each time.\n\n"
+                f"Previous result: {last_result[:200]}\n\n"
+                f"{attempted}\n\n"
+                f"STOP repeating. You MUST try something DIFFERENT:\n"
+                f"- Different arguments or search terms\n"
+                f"- A completely different tool\n"
+                f"- Ask the user for clarification"
             )
+
+            # Auto-consult smarter model if available
+            if count == 2:
+                try:
+                    from drydock.core.consultant import is_consultant_available, ask_consultant
+                    if is_consultant_available():
+                        question = (
+                            f"I'm stuck in a loop. I keep running `{tool_call.tool_name}` "
+                            f"with args: {tool_call.raw_arguments[:100]} "
+                            f"and getting: {last_result[:200]}. "
+                            f"What should I do differently?"
+                        )
+                        advice = await ask_consultant(question)
+                        if advice:
+                            msg += f"\n\nCONSULTANT ADVICE: {advice}"
+                except Exception as e:
+                    logger.debug("Consultant call failed: %s", e)
+
+            return msg
         return None
 
     def _circuit_breaker_record(self, tool_call: ResolvedToolCall, result_text: str) -> None:
@@ -863,7 +900,7 @@ class AgentLoop:
         self, tool_call: ResolvedToolCall
     ) -> AsyncGenerator[ToolResultEvent | ToolStreamEvent]:
         # Circuit breaker: block exact duplicate calls after 2 attempts
-        if blocked := self._circuit_breaker_check(tool_call):
+        if blocked := await self._circuit_breaker_check(tool_call):
             yield ToolResultEvent(
                 tool_name=tool_call.tool_name,
                 tool_class=tool_call.tool_class,

drydock_cli-0.6.1/drydock/core/consultant.py

@@ -0,0 +1,141 @@
+"""Consultant agent — asks a smarter model for single-turn advice.
+
+The consultant is a READ-ONLY advisor:
+- The local model stays in control of ALL tool calls
+- The consultant NEVER calls tools, writes files, or runs commands
+- It only receives a question and returns reasoning/advice
+- Used when the local model is stuck, uncertain, or looping
+
+Usage from agent_loop:
+    advice = await ask_consultant("I found 3 files that match. Which one likely contains the bug?")
+
+The consultant model is configured via:
+- CLI: drydock --consultant gemini-2.5-pro
+- Env: DRYDOCK_CONSULTANT_MODEL=gemini-2.5-pro
+"""
+
+from __future__ import annotations
+
+import logging
+import os
+from typing import Any
+
+import httpx
+
+logger = logging.getLogger(__name__)
+
+
+async def ask_consultant(
+    question: str,
+    context: str = "",
+    model: str | None = None,
+    api_key: str | None = None,
+    api_base: str | None = None,
+) -> str:
+    """Ask the consultant model a question and get text advice back.
+
+    Parameters
+    ----------
+    question : str
+        The specific question to ask.
+    context : str
+        Optional context (e.g., file contents, error messages).
+    model : str
+        Model name. Defaults to DRYDOCK_CONSULTANT_MODEL env var.
+    api_key : str
+        API key. Auto-detected from common env vars if not provided.
+    api_base : str
+        API base URL. Auto-detected based on model name if not provided.
+
+    Returns
+    -------
+    str
+        The consultant's advice text, or an error message.
+    """
+    model = model or os.environ.get("DRYDOCK_CONSULTANT_MODEL", "")
+    if not model:
+        return ""
+
+    # Auto-detect API key and base URL from model name
+    api_key = api_key or _detect_api_key(model)
+    api_base = api_base or _detect_api_base(model)
+
+    if not api_key:
+        logger.warning("No API key found for consultant model '%s'", model)
+        return ""
+
+    prompt = f"""You are a senior software engineering consultant. A junior AI coding agent is asking for your advice.
+Give concise, actionable guidance. Do NOT provide code — just tell them WHAT to do and WHERE to look.
+
+QUESTION:
+{question}
+"""
+    if context:
+        prompt += f"\nCONTEXT:\n{context[:2000]}\n"
+
+    try:
+        # Use OpenAI-compatible API (works with Gemini, OpenAI, Mistral, local models)
+        verify_ssl = os.environ.get("DRYDOCK_INSECURE") != "1"
+        async with httpx.AsyncClient(verify=verify_ssl, timeout=30.0) as client:
+            response = await client.post(
+                f"{api_base}/chat/completions",
+                headers={
+                    "Authorization": f"Bearer {api_key}",
+                    "Content-Type": "application/json",
+                },
+                json={
+                    "model": model,
+                    "messages": [{"role": "user", "content": prompt}],
+                    "max_tokens": 500,
+                    "temperature": 0.3,
+                    # NO tools — consultant is read-only
+                },
+            )
+            response.raise_for_status()
+            data = response.json()
+            advice = data["choices"][0]["message"]["content"]
+            logger.info("Consultant (%s) responded: %s", model, advice[:100])
+            return advice
+
+    except Exception as e:
+        logger.warning("Consultant call failed: %s", e)
+        return f"(Consultant unavailable: {e})"
+
+
+def _detect_api_key(model: str) -> str:
+    """Auto-detect API key from environment based on model name."""
+    model_lower = model.lower()
+
+    if "gemini" in model_lower:
+        return os.environ.get("GEMINI_API_KEY", os.environ.get("GOOGLE_API_KEY", ""))
+    if "gpt" in model_lower or "o1" in model_lower or "o3" in model_lower:
+        return os.environ.get("OPENAI_API_KEY", "")
+    if "claude" in model_lower:
+        return os.environ.get("ANTHROPIC_API_KEY", "")
+    if "mistral" in model_lower or "devstral" in model_lower:
+        return os.environ.get("MISTRAL_API_KEY", "")
+
+    # Generic fallback
+    return os.environ.get("DRYDOCK_CONSULTANT_API_KEY", "")
+
+
+def _detect_api_base(model: str) -> str:
+    """Auto-detect API base URL from model name."""
+    model_lower = model.lower()
+
+    if "gemini" in model_lower:
+        return "https://generativelanguage.googleapis.com/v1beta/openai"
+    if "gpt" in model_lower or "o1" in model_lower or "o3" in model_lower:
+        return "https://api.openai.com/v1"
+    if "claude" in model_lower:
+        return "https://api.anthropic.com/v1"
+    if "mistral" in model_lower or "devstral" in model_lower:
+        return os.environ.get("MISTRAL_API_BASE", "https://api.mistral.ai/v1")
+
+    # Try local vLLM
+    return os.environ.get("DRYDOCK_CONSULTANT_API_BASE", "http://localhost:8000/v1")
+
+
+def is_consultant_available() -> bool:
+    """Check if a consultant model is configured."""
+    return bool(os.environ.get("DRYDOCK_CONSULTANT_MODEL"))

{drydock_cli-0.6.0 → drydock_cli-0.6.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "drydock-cli"
-version = "0.6.0"
+version = "0.6.1"
 description = "Drydock — nautical CLI coding agent"
 readme = "README.md"
 requires-python = ">=3.12"

{drydock_cli-0.6.0 → drydock_cli-0.6.1}/tests/test_drydock_tasks.py

@@ -391,6 +391,7 @@ class TestCircuitBreaker:
     """Prevents exact duplicate tool calls."""
 
     def test_circuit_breaker_blocks_after_2(self):
+        import asyncio
         from drydock.core.agent_loop import AgentLoop
         from drydock.core.types import MessageList
         from types import SimpleNamespace
@@ -399,21 +400,19 @@ class TestCircuitBreaker:
         al.messages = MessageList()
         al._tool_call_history = {}
 
-        # Mock tool call with same name+args
         tc = SimpleNamespace(tool_name="bash", raw_arguments='{"command": "ls -ltr"}')
 
-        # First two calls: no block
-        assert al._circuit_breaker_check(tc) is None
+        assert asyncio.run(al._circuit_breaker_check(tc)) is None
         al._circuit_breaker_record(tc, "file1.py\nfile2.py")
-        assert al._circuit_breaker_check(tc) is None
+        assert asyncio.run(al._circuit_breaker_check(tc)) is None
         al._circuit_breaker_record(tc, "file1.py\nfile2.py")
 
-        # Third call: BLOCKED
-        result = al._circuit_breaker_check(tc)
+        result = asyncio.run(al._circuit_breaker_check(tc))
         assert result is not None
         assert "CIRCUIT BREAKER" in result
 
     def test_different_args_not_blocked(self):
+        import asyncio
         from drydock.core.agent_loop import AgentLoop
         from drydock.core.types import MessageList
         from types import SimpleNamespace
@@ -427,9 +426,8 @@ class TestCircuitBreaker:
 
         al._circuit_breaker_record(tc1, "result1")
         al._circuit_breaker_record(tc1, "result1")
-        # tc1 blocked, tc2 still works
-        assert al._circuit_breaker_check(tc1) is not None
-        assert al._circuit_breaker_check(tc2) is None
+        assert asyncio.run(al._circuit_breaker_check(tc1)) is not None
+        assert asyncio.run(al._circuit_breaker_check(tc2)) is None
 
 
 # ============================================================================