drydock-cli 1.5.1__tar.gz → 1.5.3__tar.gz

{drydock_cli-1.5.1 → drydock_cli-1.5.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drydock-cli
-Version: 1.5.1
+Version: 1.5.3
 Summary: Drydock — nautical CLI coding agent
 Project-URL: Homepage, https://github.com/fbobe321/drydock
 Project-URL: Repository, https://github.com/fbobe321/drydock

{drydock_cli-1.5.1 → drydock_cli-1.5.3}/drydock/core/agent_loop.py

@@ -193,6 +193,7 @@ class AgentLoop:
         self._tool_call_history: dict[str, tuple[int, str]] = {}
         self._consecutive_circuit_breaker_fires: int = 0
         self._empty_responses: int = 0
+        self._successful_test_runs: int = 0
 
         system_prompt = get_universal_system_prompt(
             self.tool_manager, self.config, self.skill_manager, self.agent_manager
@@ -473,6 +474,13 @@ class AgentLoop:
 
         yield UserMessageEvent(content=user_msg, message_id=user_message.message_id)
 
+        # === AUTO-DELEGATION ===
+        # Inject context automatically so the model doesn't have to choose to delegate.
+        # This replaces relying on the model to call task()/invoke_skill().
+        auto_context = self._build_auto_context(user_msg)
+        if auto_context:
+            self._inject_system_note(auto_context)
+
         try:
             should_break_loop = False
             tool_turns = 0
@@ -564,6 +572,10 @@ class AgentLoop:
                                     first_edit = not has_made_edit
                                     has_made_edit = True
                                     bash_count = 0  # Reset on successful edit
+                                    # Reset circuit breaker — the world changed after an edit,
+                                    # so previously-failing commands may now succeed
+                                    self._tool_call_history.clear()
+                                    self._consecutive_circuit_breaker_fires = 0
                                     # Track blast radius
                                     try:
                                         edit_args = json.loads(tc.function.arguments or "{}")
@@ -637,13 +649,13 @@ class AgentLoop:
                             consecutive_bash += 1
                         else:
                             break
-                    if consecutive_bash >= 4:
+                    if consecutive_bash >= 3:
                         self._inject_system_note(
-                            "You have run 4+ bash commands in a row after editing files. "
-                            "Each one failed or gave the same result. STOP running bash. "
-                            "Read the error message carefully, then use search_replace "
-                            "to fix the actual code. Running the same command again will "
-                            "not give a different result."
+                            "You have run 3+ bash commands in a row. "
+                            "If your project is WORKING, you are DONE — tell the user the project "
+                            "is complete and what commands they can use. "
+                            "If it is FAILING, STOP running bash and use read_file + search_replace "
+                            "to fix the code. Do NOT keep running the same test."
                         )
 
                 # Detect bash file creation (touch, echo >, cat <<)
@@ -1162,6 +1174,17 @@ class AgentLoop:
 
             result_dict = result_model.model_dump()
             text = "\n".join(f"{k}: {v}" for k, v in result_dict.items())
+
+            # After a successful bash test of built code, nudge to wrap up
+            if tool_call.tool_name in ("bash", "run_command"):
+                self._successful_test_runs += 1
+                if self._successful_test_runs >= 3:
+                    self._inject_system_note(
+                        "Your project is WORKING. You have verified it successfully. "
+                        "STOP testing. Summarize what you built and tell the user "
+                        "how to use it. Do NOT run any more bash commands."
+                    )
+
             # Record for circuit breaker
             self._circuit_breaker_record(tool_call, text)
             self._handle_tool_response(
@@ -1635,6 +1658,93 @@ class AgentLoop:
 
         return None
 
+    def _build_auto_context(self, user_msg: str) -> str | None:
+        """Build auto-delegation context based on the prompt and project state.
+
+        Instead of hoping the model calls task()/invoke_skill(), we inject:
+        1. Project file listing (if files exist) — replaces explore subagent
+        2. Skill content (if prompt matches a skill) — replaces invoke_skill
+        3. Planning prompt (if complex build task) — replaces planner subagent
+        """
+        parts: list[str] = []
+        msg_lower = user_msg.lower()
+
+        # 1. Auto-explore: list project files so model doesn't have to
+        try:
+            cwd = Path.cwd()
+            py_files = sorted(cwd.rglob("*.py"))
+            py_files = [f for f in py_files if ".logs" not in str(f) and ".venv" not in str(f)]
+            if len(py_files) >= 3:
+                listing = "\n".join(f"  {f.relative_to(cwd)}" for f in py_files[:30])
+                parts.append(f"PROJECT FILES ({len(py_files)} Python files):\n{listing}")
+                if len(py_files) > 30:
+                    parts.append(f"  ... and {len(py_files) - 30} more")
+        except Exception:
+            pass
+
+        # Also list non-Python files of interest
+        try:
+            for pattern in ("*.md", "*.txt", "*.json", "*.yaml", "*.yml", "*.toml", "*.csv"):
+                for f in sorted(cwd.glob(pattern))[:5]:
+                    if f.name not in (".logs",):
+                        parts.append(f"  {f.name} ({f.stat().st_size} bytes)")
+        except Exception:
+            pass
+
+        # 2. Skill routing: match prompt to skill and inject content
+        skill_name = None
+        if any(kw in msg_lower for kw in ("review this", "review the", "code review", "audit")):
+            skill_name = "review"
+        elif any(kw in msg_lower for kw in ("investigate", "debug", "why does", "why is", "broken", "diagnose")):
+            skill_name = "investigate"
+        elif any(kw in msg_lower for kw in ("ship it", "commit and push", "create a pr", "open a pr")):
+            skill_name = "ship"
+        elif any(kw in msg_lower for kw in ("simplify", "clean up", "reduce complexity")):
+            skill_name = "simplify"
+
+        if skill_name:
+            skill_info = self.skill_manager.get_skill(skill_name)
+            if skill_info and skill_info.skill_path:
+                try:
+                    content = skill_info.skill_path.read_text(encoding="utf-8")
+                    # Strip frontmatter
+                    if content.startswith("---"):
+                        end = content.find("---", 3)
+                        if end > 0:
+                            content = content[end + 3:].strip()
+                    parts.append(f"FOLLOW THIS WORKFLOW ({skill_name} skill):\n{content[:2000]}")
+                except Exception:
+                    pass
+
+        # 3. Planning nudge for complex build tasks
+        is_build = any(kw in msg_lower for kw in (
+            "build", "create", "implement", "make a", "write a",
+            "set up", "scaffold", "get started", "prd",
+        ))
+        is_complex = len(user_msg) > 100 or any(kw in msg_lower for kw in (
+            "multiple", "modules", "package", "api", "database",
+            "features", "cli", "commands",
+        ))
+        if is_build and is_complex:
+            parts.append(
+                "PLANNING REQUIRED — This is a complex build task. Before writing code:\n"
+                "1. List ALL files you will create (with brief purpose)\n"
+                "2. Define the package structure\n"
+                "3. Use ABSOLUTE imports (from pkg.module import X), NOT relative imports\n"
+                "4. Create files in dependency order (utilities first, then modules that import them)\n"
+                "5. After creating all files, test with: python3 -m package_name\n"
+                "6. If tests fail, READ the error, then FIX with search_replace"
+            )
+        elif is_build:
+            parts.append(
+                "REMINDER: Use write_file to create files. Use absolute imports. "
+                "Test with python3 -m package_name (not python3 package/file.py)."
+            )
+
+        if not parts:
+            return None
+        return "\n\n".join(parts)
+
     def _recent_tool_names(self, limit: int = 10) -> list[str]:
         """Return recent tool names from message history (most recent last)."""
         names: list[str] = []

{drydock_cli-1.5.1 → drydock_cli-1.5.3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "drydock-cli"
-version = "1.5.1"
+version = "1.5.3"
 description = "Drydock — nautical CLI coding agent"
 readme = "README.md"
 requires-python = ">=3.12"

drydock_cli-1.5.3/scripts/monitor_test_battery.sh

@@ -0,0 +1,115 @@
+#!/bin/bash
+# Monitor the overnight PRD test battery.
+# Restarts if stuck, logs results, and stops on repeated failures.
+# Cron: */10 * * * * /data3/drydock/scripts/monitor_test_battery.sh
+
+set -euo pipefail
+
+export PATH="/home/bobef/miniconda3/bin:$PATH"
+PYTHON="/home/bobef/miniconda3/bin/python3"
+DRYDOCK="/data3/drydock"
+LOGDIR="$DRYDOCK/test_bank_results"
+LOGFILE="$LOGDIR/overnight_run2.log"
+MONITOR_LOG="$LOGDIR/monitor.log"
+LOCKFILE="/tmp/drydock_test_monitor.lock"
+
+log() { echo "[$(date '+%Y-%m-%d %H:%M:%S')] $*" >> "$MONITOR_LOG"; }
+
+# Prevent concurrent monitors
+if [ -f "$LOCKFILE" ]; then
+    pid=$(cat "$LOCKFILE" 2>/dev/null)
+    if kill -0 "$pid" 2>/dev/null; then
+        exit 0
+    fi
+fi
+echo $$ > "$LOCKFILE"
+trap "rm -f $LOCKFILE" EXIT
+
+cd "$DRYDOCK"
+
+# Check if test process is running
+TEST_PID=$(pgrep -f "pytest.*test_bank_prd" 2>/dev/null | head -1 || echo "")
+
+if [ -z "$TEST_PID" ]; then
+    # Check if it completed
+    if [ -f "$LOGFILE" ] && grep -q "passed" "$LOGFILE" 2>/dev/null; then
+        PASSED=$(grep -c "PASSED" "$LOGFILE" 2>/dev/null || echo 0)
+        FAILED=$(grep -c "FAILED" "$LOGFILE" 2>/dev/null || echo 0)
+        TOTAL=$((PASSED + FAILED))
+
+        if [ "$TOTAL" -ge 65 ]; then
+            log "TEST BATTERY COMPLETE: $PASSED passed, $FAILED failed out of $TOTAL"
+            # Log failures
+            grep "FAILED" "$LOGFILE" >> "$MONITOR_LOG" 2>/dev/null || true
+            exit 0
+        fi
+    fi
+
+    # Not running and not complete — restart
+    log "Test battery not running. Restarting..."
+
+    # Check how many tests already completed
+    DONE=$(grep -c "PASSED\|FAILED" "$LOGFILE" 2>/dev/null || echo 0)
+
+    if [ "$DONE" -gt 0 ]; then
+        # Save progress
+        cp "$LOGFILE" "$LOGDIR/overnight_run2_partial_$(date +%H%M).log" 2>/dev/null || true
+
+        # Check for repeated failures (same test failing 3+ times across runs)
+        FAIL_LOG="$LOGDIR/failure_tracking.log"
+        grep "FAILED" "$LOGFILE" >> "$FAIL_LOG" 2>/dev/null || true
+
+        # If a specific test has failed 3+ times, skip it
+        SKIP_TESTS=""
+        if [ -f "$FAIL_LOG" ]; then
+            while IFS= read -r test_name; do
+                COUNT=$(grep -c "$test_name" "$FAIL_LOG" 2>/dev/null || echo 0)
+                if [ "$COUNT" -ge 3 ]; then
+                    log "SKIP: $test_name failed $COUNT times, skipping"
+                    SKIP_TESTS="$SKIP_TESTS and not ${test_name##*::}"
+                fi
+            done < <(grep "FAILED" "$FAIL_LOG" | sed 's/ .*//' | sort -u)
+        fi
+    fi
+
+    # Restart
+    DESELECT=""
+    if [ -n "${SKIP_TESTS:-}" ]; then
+        DESELECT="-k 'not _PLACEHOLDER_ $SKIP_TESTS'"
+    fi
+
+    nohup $PYTHON -m pytest tests/test_bank_prd.py tests/test_bank_prd_extended.py \
+        -v -p no:xdist -p no:cov --override-ini="addopts=" --tb=short \
+        $DESELECT \
+        > "$LOGFILE" 2>&1 &
+
+    log "Restarted with PID $! (skipping: ${SKIP_TESTS:-none})"
+    exit 0
+fi
+
+# Process is running — check if stuck
+LAST_MODIFIED=$(stat -c %Y "$LOGFILE" 2>/dev/null || echo 0)
+NOW=$(date +%s)
+STALE=$((NOW - LAST_MODIFIED))
+
+if [ "$STALE" -gt 600 ]; then
+    # Log file hasn't been updated in 10 minutes — test might be stuck
+    CURRENT_TEST=$(tail -1 "$LOGFILE" | grep -o "test_[a-z_]*" | head -1 || echo "unknown")
+    log "WARNING: Test battery stale for ${STALE}s. Current test: $CURRENT_TEST"
+
+    if [ "$STALE" -gt 1200 ]; then
+        # 20 minutes stuck — kill and restart
+        log "STUCK: Killing PID $TEST_PID (stale ${STALE}s on $CURRENT_TEST)"
+        kill "$TEST_PID" 2>/dev/null || true
+        sleep 5
+        kill -9 "$TEST_PID" 2>/dev/null || true
+
+        # Will restart on next cron run
+    fi
+else
+    # Running fine — log progress
+    PASSED=$(grep -c "PASSED" "$LOGFILE" 2>/dev/null || echo 0)
+    FAILED=$(grep -c "FAILED" "$LOGFILE" 2>/dev/null || echo 0)
+    TOTAL=$((PASSED + FAILED))
+    log "Running: $TOTAL/65 done ($PASSED passed, $FAILED failed)"
+fi