drydock-cli 2.6.30__tar.gz → 2.6.32__tar.gz

{drydock_cli-2.6.30 → drydock_cli-2.6.32}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drydock-cli
-Version: 2.6.30
+Version: 2.6.32
 Summary: Local-first CLI coding agent — tested with Gemma 4 26B via vLLM
 Project-URL: Homepage, https://github.com/fbobe321/drydock
 Project-URL: Repository, https://github.com/fbobe321/drydock
@@ -60,11 +60,11 @@ Description-Content-Type: text/markdown
 [![License](https://img.shields.io/github/license/fbobe321/drydock)](https://github.com/fbobe321/drydock/blob/main/LICENSE)
 
 ```
-     ____                 _             _    
-    |  _ \ _ __ _   _  __| | ___   ___ | | __
-    | | | | '__| | | |/ _` |/ _ \ / __|| |/ /
-    | |_| | |  | |_| | (_| | (_) | (__ |   < 
-    |____/|_|   \__, |\__,_|\___/ \___||_|\_\
+     ____              ____             _    
+    |  _ \ _ __ _   _ |  _ \  ___   ___| | __
+    | | | | '__| | | || | | |/ _ \ / __| |/ /
+    | |_| | |  | |_| || |_| | (_) | (__|   < 
+    |____/|_|   \__, ||____/ \___/ \___|_|\_\
                 |___/
 ```
 

{drydock_cli-2.6.30 → drydock_cli-2.6.32}/README.md

@@ -4,11 +4,11 @@
 [![License](https://img.shields.io/github/license/fbobe321/drydock)](https://github.com/fbobe321/drydock/blob/main/LICENSE)
 
 ```
-     ____                 _             _    
-    |  _ \ _ __ _   _  __| | ___   ___ | | __
-    | | | | '__| | | |/ _` |/ _ \ / __|| |/ /
-    | |_| | |  | |_| | (_| | (_) | (__ |   < 
-    |____/|_|   \__, |\__,_|\___/ \___||_|\_\
+     ____              ____             _    
+    |  _ \ _ __ _   _ |  _ \  ___   ___| | __
+    | | | | '__| | | || | | |/ _ \ / __| |/ /
+    | |_| | |  | |_| || |_| | (_) | (__|   < 
+    |____/|_|   \__, ||____/ \___/ \___|_|\_\
                 |___/
 ```
 

{drydock_cli-2.6.30 → drydock_cli-2.6.32}/drydock/core/prompts/gemma4.md

@@ -29,10 +29,11 @@ For multi-module changes that need a plan first:
 
 DO NOT delegate trivial work. Most PRDs are small — build them directly.
 Rules of thumb:
-- 1-5 files, no subdirs → BUILD INLINE. Do not call task.
-- 6+ files OR subdirectories OR multiple subpackages → DELEGATE to builder.
+- 1-8 files → BUILD INLINE. Do not call task.
+- 9+ files with multiple subdirectories → DELEGATE to builder.
 - Editing an existing file or fixing a known bug → BUILD INLINE.
 - "Where does function X live?" → DELEGATE to explore.
+- If the user asks you to PLAN or EXPLAIN → respond with text. Do not delegate.
 
 When in doubt, build inline. A wasted delegation costs 60-90 seconds of
 extra context loading.
@@ -61,4 +62,7 @@ Rules:
 - Follow the EXACT CLI interface specified in the PRD. Match argument names, subcommands, and flags exactly.
 - Every subcommand in the PRD must have a working handler — not just argparse registration.
 - After creating files, run python3 -m package_name [subcommand] to verify each one works.
+- If you have a todo list, update it after completing each major step
+  (e.g. after building all files, after tests pass). Use todo(action="write")
+  to mark items as done.
 - If a write_file result says "BLOCKED:" you've called it 3+ times with identical content. STOP that path. Write a DIFFERENT file or run bash. Never retry the blocked write.

{drydock_cli-2.6.30 → drydock_cli-2.6.32}/drydock/core/tools/builtins/search_replace.py

@@ -286,12 +286,14 @@ class SearchReplace(
         if not search_replace_blocks:
             raise ToolError(
                 "No valid SEARCH/REPLACE blocks found in content.\n"
-                "Expected format:\n"
-                "<<<<<<< SEARCH\n"
-                "[exact content to find]\n"
-                "=======\n"
-                "[new content to replace with]\n"
-                ">>>>>>> REPLACE"
+                "You sent raw code without SEARCH/REPLACE markers.\n"
+                "Either use this format:\n"
+                "  <<<<<<< SEARCH\n"
+                "  [exact text to find]\n"
+                "  =======\n"
+                "  [replacement text]\n"
+                "  >>>>>>> REPLACE\n"
+                "Or use write_file to replace the entire file."
             )
 
         # Detect garbled token output — only flag obvious corruption

{drydock_cli-2.6.30 → drydock_cli-2.6.32}/drydock/core/tools/builtins/task.py

@@ -194,7 +194,9 @@ class Task(
             )
             return
 
-        # Foreground execution: block and collect results
+        # Foreground execution: block and collect results.
+        # Stream both tool CALLS and tool RESULTS so the user sees
+        # progress in the TUI instead of silence for 10 minutes.
         accumulated_response: list[str] = []
         completed = True
         try:
@@ -203,13 +205,24 @@ class Task(
                     accumulated_response.append(event.content)
                     if event.stopped_by_middleware:
                         completed = False
+                elif isinstance(event, ToolCallEvent):
+                    # Show what the subagent is about to do
+                    if event.tool_class:
+                        adapter = ToolUIDataAdapter(event.tool_class)
+                        display = adapter.get_call_display(event)
+                        message = f"→ {display.summary}"
+                        yield ToolStreamEvent(
+                            tool_name=self.get_name(),
+                            message=message,
+                            tool_call_id=ctx.tool_call_id,
+                        )
                 elif isinstance(event, ToolResultEvent):
                     if event.skipped:
                         completed = False
                     elif event.result and event.tool_class:
                         adapter = ToolUIDataAdapter(event.tool_class)
                         display = adapter.get_result_display(event)
-                        message = f"{event.tool_name}: {display.message}"
+                        message = f"  {event.tool_name}: {display.message}"
                         yield ToolStreamEvent(
                             tool_name=self.get_name(),
                             message=message,

{drydock_cli-2.6.30 → drydock_cli-2.6.32}/drydock/core/tools/builtins/todo.py

@@ -2,7 +2,7 @@ from __future__ import annotations
 
 from collections.abc import AsyncGenerator
 from enum import StrEnum, auto
-from typing import ClassVar
+from typing import Any, ClassVar
 
 from pydantic import BaseModel, Field, field_validator
 
@@ -58,6 +58,40 @@ class TodoArgs(BaseModel):
         default=None, description="Complete list of todos when writing."
     )
 
+    @field_validator("todos", mode="before")
+    @classmethod
+    def coerce_todos(cls, v: Any) -> Any:
+        """Gemma 4 sends todos in various malformed formats.
+
+        Common patterns:
+          - A single string: "1. Do this 2. Do that"
+          - A list of strings: ["Do this", "Do that"]
+          - A list of dicts missing required fields
+
+        Coerce all of these into list[TodoItem]-compatible dicts.
+        """
+        if v is None:
+            return v
+        if isinstance(v, str):
+            # Split numbered or newline-separated items
+            import re
+            items = re.split(r'\n|(?:\d+\.\s)', v)
+            return [
+                {"content": item.strip(), "status": "pending"}
+                for item in items if item.strip()
+            ]
+        if isinstance(v, list):
+            result = []
+            for item in v:
+                if isinstance(item, str):
+                    result.append({"content": item.strip(), "status": "pending"})
+                elif isinstance(item, dict):
+                    result.append(item)
+                else:
+                    result.append({"content": str(item), "status": "pending"})
+            return result
+        return v
+
 
 class TodoResult(BaseModel):
     message: str

{drydock_cli-2.6.30 → drydock_cli-2.6.32}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "drydock-cli"
-version = "2.6.30"
+version = "2.6.32"
 description = "Local-first CLI coding agent — tested with Gemma 4 26B via vLLM"
 readme = "README.md"
 requires-python = ">=3.12"

{drydock_cli-2.6.30 → drydock_cli-2.6.32}/scripts/shakedown_interactive.py

@@ -32,16 +32,20 @@ SESSION_ROOT = Path.home() / ".vibe" / "logs" / "session"
 # wait_condition: "msgs>=N" or "writes>=N" or "done"
 
 SCRIPTS: dict[str, list[tuple[str, str, int]]] = {
-    # ── Tool Agent: build, test, extend, troubleshoot ──
+    # ── Tool Agent: plan, todo, build, test, extend, troubleshoot ──
     "tool_agent": [
         ("Review the PRD and create a plan. List what files you will create "
          "and in what order. Do NOT write any code yet — just plan.",
          "msgs>=4", 120),
-        ("Good. Now start building — write __init__.py, __main__.py, and tools.py first.",
-         "writes>=3", 120),
-        ("Continue — write agent.py and cli.py.",
-         "writes>=5", 120),
-        ("Run python3 -m {pkg} --help to verify the package works. If it fails, fix it.",
+        ("Good plan. Now create a todo list with all the steps needed to "
+         "build this package. Include: create files, test, verify each tool. "
+         "Then start working through the todo list — build the first 3 files.",
+         "writes>=3", 150),
+        ("Keep going through your todo list. Write the remaining files. "
+         "Update each todo item as you complete it.",
+         "writes>=5", 150),
+        ("Run python3 -m {pkg} --help to verify the package works. If it "
+         "fails, fix it. Mark the test todo as done when it passes.",
          "msgs>={prev}+3", 120),
         ("Explain how the agent loop works. Walk me through what happens when "
          "a user types a query.",
@@ -52,18 +56,22 @@ SCRIPTS: dict[str, list[tuple[str, str, int]]] = {
         ("Test it: run python3 -m {pkg} \"What is the date today?\"",
          "msgs>={prev}+2", 90),
         ("Now test all the original tools still work: "
-         "\"What is 23 * 47?\", \"Count words in PRD.md\", \"Search for def in *.py\"",
+         "\"What is 23 * 47?\", \"Count words in PRD.md\", \"Search for def in *.py\". "
+         "Show me the final todo list status.",
          "done", 180),
     ],
 
-    # ── Stock Screener: build, create data, test filters, add feature ──
+    # ── Stock Screener: todo-driven build + test + feature add ──
     "stock_screener": [
         ("Review the PRD. What modules will you need? Plan the architecture "
          "before writing any code.",
          "msgs>=4", 120),
-        ("Start building. Create the data loading module and the screener logic first.",
-         "writes>=3", 120),
-        ("Create the CLI, formatter, and __main__.py. Make sure the package can run.",
+        ("Create a todo list for the full build: 1) data loader, 2) screener "
+         "logic, 3) CLI, 4) formatter, 5) sample data, 6) test all commands, "
+         "7) export test. Then start working — build the data loader and screener.",
+         "writes>=3", 150),
+        ("Continue through the todo list. Create the CLI, formatter, and "
+         "__main__.py. Make sure the package can run. Update todos as you go.",
          "writes>=6", 150),
         ("Create a sample portfolio.csv with at least 8 stocks for testing.",
          "msgs>={prev}+2", 90),
@@ -74,19 +82,21 @@ SCRIPTS: dict[str, list[tuple[str, str, int]]] = {
          "msgs>={prev}+4", 120),
         ("Test the rank command: python3 -m {pkg} rank --data portfolio.csv --top 5",
          "msgs>={prev}+2", 90),
-        ("Export results to JSON and show me the file contents.",
+        ("Export results to JSON and show me the file contents. "
+         "Show me the final todo list — everything should be done.",
          "done", 120),
     ],
 
-    # ── Eval Harness: build, create dataset, run eval, fix issues ──
+    # ── Eval Harness: todo-driven build + dataset + eval + improve ──
     "eval_harness": [
         ("Review the PRD and plan. How will you implement the code evaluator "
          "safely without using eval()? Think through the design.",
          "msgs>=4", 120),
-        ("Build the evaluators module first — exact match, fuzzy match, "
-         "and contains evaluators.",
-         "writes>=2", 120),
-        ("Now build runner.py, report.py, cli.py, and __main__.py.",
+        ("Create a todo list for the full build. Then start — build the "
+         "evaluators module first with exact match, fuzzy match, and contains.",
+         "writes>=2", 150),
+        ("Continue through the todo list. Build runner.py, report.py, cli.py, "
+         "and __main__.py. Update todos as you finish each file.",
          "writes>=6", 150),
         ("Create a sample tasks.json with 10 test cases — mix of math, "
          "text, and logic questions.",
@@ -99,18 +109,22 @@ SCRIPTS: dict[str, list[tuple[str, str, int]]] = {
          "msgs>={prev}+4", 120),
         ("Run the eval again and show me the report. Are the results better?",
          "msgs>={prev}+3", 120),
-        ("What's the overall accuracy? Which tasks are failing and why?",
+        ("What's the overall accuracy? Which tasks are failing and why? "
+         "Show me the final todo list status.",
          "done", 120),
     ],
 
-    # ── Doc QA: build, ingest, query, improve retrieval ──
+    # ── Doc QA: todo-driven build + ingest + query + debug ──
     "doc_qa": [
         ("Review the PRD. This is a TF-IDF retrieval system with no external "
          "deps. How will you implement cosine similarity with just stdlib? Plan first.",
          "msgs>=4", 120),
-        ("Build the ingestion module — document loading and chunking.",
-         "writes>=2", 120),
-        ("Build the TF-IDF index module and query processor.",
+        ("Create a todo list for the build: 1) ingestion, 2) chunking, "
+         "3) TF-IDF index, 4) query engine, 5) CLI, 6) test data, 7) test "
+         "ingest, 8) test query. Then start — build the ingestion module.",
+         "writes>=2", 150),
+        ("Continue through the todo list. Build the TF-IDF index module "
+         "and query processor. Update todos as you complete each.",
          "writes>=4", 150),
         ("Build the CLI and __main__.py. Make sure python3 -m {pkg} --help works.",
          "writes>=6", 120),
@@ -121,18 +135,21 @@ SCRIPTS: dict[str, list[tuple[str, str, int]]] = {
         ("Query: python3 -m {pkg} query \"What is the main topic?\"",
          "msgs>={prev}+2", 90),
         ("The retrieval results look off. Read the index.py file and explain "
-         "the TF-IDF calculation. Is the IDF formula correct?",
+         "the TF-IDF calculation. Is the IDF formula correct? "
+         "Show me the final todo list.",
          "done", 120),
     ],
 
-    # ── Prompt Optimizer: build, create dataset, run optimization ──
+    # ── Prompt Optimizer: todo-driven build + dataset + optimize + improve ──
     "prompt_optimizer": [
         ("Review the PRD. The default executor uses keyword matching — explain "
          "how you'll implement that without an LLM. Plan the architecture.",
          "msgs>=4", 120),
-        ("Build the template generation and mutation module first.",
-         "writes>=2", 120),
-        ("Build the executor, scorer, and optimizer loop.",
+        ("Create a todo list for the full build. Then start — build the "
+         "template generation and mutation module first.",
+         "writes>=2", 150),
+        ("Continue through the todo list. Build the executor, scorer, and "
+         "optimizer loop. Update todos as you go.",
          "writes>=5", 150),
         ("Build the CLI and create a sample sentiment dataset with 15 examples.",
          "writes>=7", 120),
@@ -143,7 +160,8 @@ SCRIPTS: dict[str, list[tuple[str, str, int]]] = {
         ("The keyword matching is too simple. Add support for negation words "
          "like 'not good' = negative. Use search_replace on executor.py.",
          "msgs>={prev}+4", 120),
-        ("Run the optimization again and compare results to the previous run.",
+        ("Run the optimization again and compare results to the previous run. "
+         "Show me the final todo list — everything should be marked done.",
          "done", 120),
     ],
 }
@@ -320,7 +338,6 @@ def run_interactive(cwd: Path, pkg: str) -> int:
     try:
         for step_idx, (prompt_template, condition, max_wait) in enumerate(script):
             prompt = prompt_template.replace("{pkg}", pkg)
-            prev_msgs = len(watcher.messages)
 
             print(f"\n--- Step {step_idx + 1}/{len(script)} ---")
             print(f"  PROMPT: {prompt[:80]}{'...' if len(prompt) > 80 else ''}")
@@ -329,6 +346,22 @@ def run_interactive(cwd: Path, pkg: str) -> int:
             type_message(child, prompt)
             time.sleep(1)
 
+            # On first step, wait for the session to actually appear
+            # before counting messages. The TUI takes a few seconds to
+            # create the session directory.
+            if step_idx == 0:
+                print("  Waiting for session to appear...", end="", flush=True)
+                for _ in range(30):  # up to 30s
+                    drain_pty(child)
+                    if watcher.find_session():
+                        break
+                if watcher.find_session():
+                    print(f" found: {watcher.session_dir.name}")
+                else:
+                    print(" not found (continuing anyway)")
+
+            prev_msgs = len(watcher.messages)
+
             # Wait for condition
             step_start = time.time()
             met = False

{drydock_cli-2.6.30 → drydock_cli-2.6.32}/uv.lock

@@ -326,7 +326,7 @@ wheels = [
 
 [[package]]
 name = "drydock-cli"
-version = "2.6.29"
+version = "2.6.31"
 source = { editable = "." }
 dependencies = [
     { name = "agent-client-protocol" },