drydock-cli 3.0.87__tar.gz → 3.0.89__tar.gz

{drydock_cli-3.0.87/drydock_cli.egg-info → drydock_cli-3.0.89}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drydock-cli
-Version: 3.0.87
+Version: 3.0.89
 Summary: Drydock — a local, provider-agnostic terminal coding agent for local LLMs
 Author: Frank Bobe III
 License-Expression: Apache-2.0

{drydock_cli-3.0.87 → drydock_cli-3.0.89}/drydock/tools/__init__.py

@@ -85,7 +85,7 @@ SCHEMAS = [
     },
     {
         "name": "ViewImage",
-        "description": "Look at an image file with your vision — use this when you need to SEE a screenshot, mockup, diagram, photo, or rendered output (.png/.jpg/.jpeg/.gif/.webp/.bmp). The image is shown to you so you can describe it, read text from it, or debug it. (Reading an image with the Read tool gives binary garbage; use ViewImage instead.)",
+        "description": "Look at an image file with your vision — use this when you need to SEE a screenshot, mockup, diagram, photo, or rendered output (.png/.jpg/.jpeg/.gif/.webp/.bmp). The image is shown to you so you can describe it, read text from it, or debug it. To READ TEXT OR DATA out of an image — a scanned document, invoice, receipt, form, or a screenshot with text — use ViewImage FIRST (you can read it directly); prefer it over OCR tools like tesseract/pdftotext, which are often unreliable or not installed. (Reading an image with the Read tool gives binary garbage; use ViewImage instead.)",
         "input_schema": {
             "type": "object",
             "properties": {
@@ -936,6 +936,7 @@ def tool_bash(params: dict, config: dict) -> str:
         reader = threading.Thread(target=_drain, daemon=True)
         reader.start()
         start = time.monotonic()
+        backgrounded = False
         while reader.is_alive():
             reader.join(0.3)
             if capped.is_set():
@@ -945,6 +946,17 @@ def tool_bash(params: dict, config: dict) -> str:
                 kill_process_group(proc)
                 proc.wait()
                 return "[stopped by user]"
+            # The SHELL has exited but the pipe is still open → the command
+            # backgrounded a child (`cmd &`, a server the task wants to keep
+            # running) that inherited stdout. Don't wait for it (that would hang
+            # until the timeout) and DON'T kill it — return what we have so the
+            # background process survives. (Redirecting its output, `cmd >log &`,
+            # closes the pipe and never reaches here.)
+            if proc.poll() is not None:
+                reader.join(0.5)  # brief grace for any final buffered output
+                if reader.is_alive():
+                    backgrounded = True
+                    break
             if time.monotonic() - start > timeout:
                 kill_process_group(proc)
                 proc.wait()
@@ -958,11 +970,16 @@ def tool_bash(params: dict, config: dict) -> str:
                 if _is_network_command(cmd):
                     msg += _OFFLINE_HINT
                 return msg
-        proc.wait()
+        if not backgrounded:
+            proc.wait()
         # Collapse repetitive runs FIRST (turns 256 KB of "y\n" into ~2 lines),
         # then note if we hit the byte cap. Bounds both RAM (the cap) and context
-        # tokens (the collapse).
-        output = _collapse_repeated_lines("".join(chunks))
+        # tokens (the collapse). Snapshot chunks (list()) in case the reader
+        # daemon is still appending for a backgrounded child.
+        output = _collapse_repeated_lines("".join(list(chunks)))
+        if backgrounded:
+            return (output.rstrip() + "\n[a process was left running in the background; "
+                    "the command returned. Check it with a follow-up command.]").lstrip("\n")
         if capped.is_set():
             return (
                 output.rstrip()

{drydock_cli-3.0.87 → drydock_cli-3.0.89}/drydock/tuning.py

@@ -95,7 +95,9 @@ _DEFAULT_SYSTEM_PROMPT = (
     ".jpeg/.gif/.webp/.bmp on disk — a screenshot, mockup, diagram, or photo), it "
     "is attached and you can SEE it; you can also call the `ViewImage` tool to "
     "look at any image you find yourself (Read gives binary garbage for images — "
-    "use ViewImage). Describe it, read text from it, or debug with it — never "
+    "use ViewImage). To read text/data out of a scanned document, invoice, "
+    "receipt, or screenshot, use ViewImage directly rather than OCR tools like "
+    "tesseract. Describe it, read text from it, or debug with it — never "
     "claim you cannot view images."
 )
 
@@ -123,8 +125,10 @@ _GEMMA_SYSTEM_PROMPT = (
     "(.png/.jpg/.jpeg/.gif/.webp/.bmp that exists on disk — a screenshot, mockup, "
     "diagram, or photo), that image is attached and you can SEE it. You can also "
     "call the `ViewImage` tool to look at any image you find yourself (Read gives "
-    "binary garbage for images — use ViewImage). Describe it, read text from it, "
-    "or use it to debug — never claim you cannot view images."
+    "binary garbage for images — use ViewImage). To read TEXT or DATA out of a "
+    "scanned document, invoice, receipt, form, or a screenshot, use ViewImage "
+    "directly — don't reach for OCR tools like tesseract first. Describe it, read "
+    "text from it, or use it to debug — never claim you cannot view images."
 )
 
 

{drydock_cli-3.0.87 → drydock_cli-3.0.89/drydock_cli.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drydock-cli
-Version: 3.0.87
+Version: 3.0.89
 Summary: Drydock — a local, provider-agnostic terminal coding agent for local LLMs
 Author: Frank Bobe III
 License-Expression: Apache-2.0

{drydock_cli-3.0.87 → drydock_cli-3.0.89}/drydock_cli.egg-info/SOURCES.txt

@@ -49,6 +49,7 @@ drydock_cli.egg-info/top_level.txt
 tests/test_advisor.py
 tests/test_approval.py
 tests/test_back_command.py
+tests/test_bash_background.py
 tests/test_bash_output_bounding.py
 tests/test_bash_process_group.py
 tests/test_bash_safety.py

{drydock_cli-3.0.87 → drydock_cli-3.0.89}/pyproject.toml

@@ -7,7 +7,7 @@ build-backend = "setuptools.build_meta"
 # PyPI distribution name is drydock-cli (the established install name, continued
 # from the retired v2 fork); the import package + CLI command stay `drydock`.
 name = "drydock-cli"
-version = "3.0.87"
+version = "3.0.89"
 description = "Drydock — a local, provider-agnostic terminal coding agent for local LLMs"
 readme = "README.md"
 requires-python = ">=3.11"

drydock_cli-3.0.89/tests/test_bash_background.py

@@ -0,0 +1,44 @@
+"""tool_bash must support backgrounded processes (`cmd &`) — a server/daemon the
+task wants to keep running — instead of hanging to the timeout then killing it.
+Found via the kv-store-grpc tbench task (needs a background gRPC server)."""
+from __future__ import annotations
+
+import subprocess
+import time
+
+from drydock.tools import tool_bash
+
+
+def _timed(cmd, timeout=8):
+    t0 = time.monotonic()
+    out = tool_bash({"command": cmd, "timeout": timeout}, {"cwd": "/tmp"})
+    return time.monotonic() - t0, out
+
+
+def test_backgrounded_command_returns_fast_and_survives():
+    try:
+        dt, out = _timed("sleep 12 & echo bg-started")
+        assert dt < 3.0, f"backgrounded cmd hung ({dt:.1f}s) — should return promptly"
+        assert "background" in out.lower()
+        time.sleep(0.5)
+        n = subprocess.run("pgrep -f 'sleep 12' | wc -l", shell=True,
+                           capture_output=True, text=True).stdout.strip()
+        assert int(n) >= 1, "backgrounded process was killed (should survive)"
+    finally:
+        subprocess.run("pkill -f 'sleep 12'", shell=True)
+
+
+def test_normal_command_unaffected():
+    dt, out = _timed("echo hello && echo world")
+    assert "hello" in out and "world" in out and dt < 2.0
+    assert "background" not in out.lower()
+
+
+def test_foreground_hang_still_times_out():
+    dt, out = _timed("sleep 30", timeout=3)
+    assert "timed out" in out and dt < 6.0
+
+
+def test_failing_command_still_reports_exit_code():
+    _, out = _timed("false")
+    assert "[exit code: 1]" in out

{drydock_cli-3.0.87 → drydock_cli-3.0.89}/tests/test_viewimage.py

@@ -74,3 +74,15 @@ def test_image_load_error_classifier():
     assert is_image_load_error("invalid_request_error: could not decode image")
     assert not is_image_load_error("rate limit exceeded")
     assert not is_image_load_error("context_length_exceeded")
+
+
+def test_viewimage_nudges_over_ocr_for_documents():
+    """The tool description + both system prompts steer the model to ViewImage
+    (its own vision) for reading text off invoices/receipts/scans, instead of
+    reaching for OCR tools first (observed on the financial-doc tbench task)."""
+    from drydock.tuning import system_prompt_for_model
+    desc = reg.get("ViewImage").schema["description"]
+    assert "OCR" in desc and ("invoice" in desc or "receipt" in desc)
+    for model in ("gemma4", "some-other-model"):
+        p = system_prompt_for_model(model)
+        assert "OCR" in p and "ViewImage" in p