drydock-cli 3.0.60__tar.gz → 3.0.62__tar.gz

{drydock_cli-3.0.60/drydock_cli.egg-info → drydock_cli-3.0.62}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drydock-cli
-Version: 3.0.60
+Version: 3.0.62
 Summary: Drydock — a local, provider-agnostic terminal coding agent for local LLMs
 Author: Frank Bobe III
 License-Expression: Apache-2.0
@@ -18,6 +18,8 @@ Requires-Dist: pytest; extra == "dev"
 Requires-Dist: pytest-timeout; extra == "dev"
 Requires-Dist: ruff; extra == "dev"
 Requires-Dist: pyright; extra == "dev"
+Provides-Extra: pdf
+Requires-Dist: pypdf>=4; extra == "pdf"
 Dynamic: license-file
 
 # ⚓ Drydock
@@ -71,6 +73,53 @@ A full agentic CLI harness — every tool below is clean-room and dependency-fre
   `~/.drydock/skills/` (or `<project>/.drydock/skills/`); `$ARGS` substitution.
 - **Loops** — `/loop <count> <prompt>` runs a prompt iteratively (Esc stops).
 
+## Slash commands
+
+Typed into the prompt. The agent also knows these, so you can just **ask it**
+("how do I add my own docs?") and it'll point you to the right one.
+
+| Command | What it does |
+| --- | --- |
+| `/graphrag build <path>` | Build a knowledge base from a file or folder of docs/code |
+| `/graphrag add <path>` | Incrementally add more documents to the base |
+| `/graphrag query <q>` | Test what the base returns (no model) |
+| `/graphrag status` · `clear` | List indexed sources · wipe the base |
+| `/skills` | List your skills |
+| `/skills new <name> <prompt>` | Create a reusable `/<name>` skill (use `$ARGS` for input) |
+| `/<name>` | Run a skill |
+| `/loop <count> <prompt>` | Repeat a prompt N times (Esc stops) |
+| `/mcp` | List connected MCP servers + their tools |
+| `/model` · `/cwd` | Show/set model & endpoint · working directory |
+| `/undo` · `/back` | Revert the last write · rewind the last turn |
+| `/compact` · `/status` · `/clear` | Shrink context · session stats · reset |
+| `/help` · `/quit` | Help · exit |
+
+### Knowledge base (GraphRAG) — ingesting your documents
+
+```
+/graphrag build ./docs        # index a file or a whole folder
+/graphrag add ./more_docs     # add more later, incrementally
+/graphrag query "how are refunds handled?"   # check retrieval
+/graphrag status              # what's indexed
+```
+
+Once built, the agent **automatically** retrieves from it (read-only `Knowledge`
+tool) when a question touches your material. Ingests text formats
+(`.md .txt .py .js .json .yaml .sql …`) **plus PDF and Word (`.docx`)**. `.docx`
+needs nothing extra; PDF uses the `pdftotext` binary (poppler) if present, else
+`pip install drydock-cli[pdf]` (pypdf). The index is a single JSON at
+`<project>/.drydock/graphrag.json` — clean-room, no embeddings.
+
+### Custom skills
+
+```
+/skills new commitmsg  Write a concise conventional-commit message for: $ARGS
+/commitmsg the staged auth changes      # runs the skill with $ARGS substituted
+```
+
+Skills are markdown files in `~/.drydock/skills/` (personal) or
+`<project>/.drydock/skills/` (project); `/skills new` writes one for you.
+
 ## Install
 
 ```bash

{drydock_cli-3.0.60 → drydock_cli-3.0.62}/README.md

@@ -49,6 +49,53 @@ A full agentic CLI harness — every tool below is clean-room and dependency-fre
   `~/.drydock/skills/` (or `<project>/.drydock/skills/`); `$ARGS` substitution.
 - **Loops** — `/loop <count> <prompt>` runs a prompt iteratively (Esc stops).
 
+## Slash commands
+
+Typed into the prompt. The agent also knows these, so you can just **ask it**
+("how do I add my own docs?") and it'll point you to the right one.
+
+| Command | What it does |
+| --- | --- |
+| `/graphrag build <path>` | Build a knowledge base from a file or folder of docs/code |
+| `/graphrag add <path>` | Incrementally add more documents to the base |
+| `/graphrag query <q>` | Test what the base returns (no model) |
+| `/graphrag status` · `clear` | List indexed sources · wipe the base |
+| `/skills` | List your skills |
+| `/skills new <name> <prompt>` | Create a reusable `/<name>` skill (use `$ARGS` for input) |
+| `/<name>` | Run a skill |
+| `/loop <count> <prompt>` | Repeat a prompt N times (Esc stops) |
+| `/mcp` | List connected MCP servers + their tools |
+| `/model` · `/cwd` | Show/set model & endpoint · working directory |
+| `/undo` · `/back` | Revert the last write · rewind the last turn |
+| `/compact` · `/status` · `/clear` | Shrink context · session stats · reset |
+| `/help` · `/quit` | Help · exit |
+
+### Knowledge base (GraphRAG) — ingesting your documents
+
+```
+/graphrag build ./docs        # index a file or a whole folder
+/graphrag add ./more_docs     # add more later, incrementally
+/graphrag query "how are refunds handled?"   # check retrieval
+/graphrag status              # what's indexed
+```
+
+Once built, the agent **automatically** retrieves from it (read-only `Knowledge`
+tool) when a question touches your material. Ingests text formats
+(`.md .txt .py .js .json .yaml .sql …`) **plus PDF and Word (`.docx`)**. `.docx`
+needs nothing extra; PDF uses the `pdftotext` binary (poppler) if present, else
+`pip install drydock-cli[pdf]` (pypdf). The index is a single JSON at
+`<project>/.drydock/graphrag.json` — clean-room, no embeddings.
+
+### Custom skills
+
+```
+/skills new commitmsg  Write a concise conventional-commit message for: $ARGS
+/commitmsg the staged auth changes      # runs the skill with $ARGS substituted
+```
+
+Skills are markdown files in `~/.drydock/skills/` (personal) or
+`<project>/.drydock/skills/` (project); `/skills new` writes one for you.
+
 ## Install
 
 ```bash

drydock_cli-3.0.62/drydock/extract.py

@@ -0,0 +1,94 @@
+"""Document text extraction for GraphRAG ingestion — PDF + Word (.docx).
+
+Kept clean-room and dependency-light:
+  • .docx is a zip of XML, so we read it with STDLIB only (zipfile + ElementTree)
+    — no dependency.
+  • .pdf uses whatever the system already has: the `pdftotext` binary (poppler)
+    if present, else the optional `pypdf` package (`pip install drydock-cli[pdf]`).
+
+Returns None when a document can't be read (unsupported, corrupt, or no PDF
+backend available) so ingestion skips it cleanly instead of failing.
+
+All logic original to Drydock.
+"""
+from __future__ import annotations
+
+import subprocess
+import xml.etree.ElementTree as ET
+import zipfile
+from pathlib import Path
+
+# Extensions we can pull text out of (beyond the plain-text formats GraphRAG
+# already reads directly).
+EXTRACTABLE_EXT = {".pdf", ".docx"}
+
+_W = "{http://schemas.openxmlformats.org/wordprocessingml/2006/main}"
+
+
+def _extract_docx(path: str | Path) -> str | None:
+    """Pull paragraph text from a .docx (stdlib only). Each <w:p> is a paragraph;
+    its text lives in <w:t> runs."""
+    try:
+        with zipfile.ZipFile(path) as z:
+            xml = z.read("word/document.xml")
+    except (zipfile.BadZipFile, KeyError, OSError):
+        return None
+    try:
+        root = ET.fromstring(xml)
+    except ET.ParseError:
+        return None
+    paras: list[str] = []
+    for p in root.iter(f"{_W}p"):
+        runs = [t.text for t in p.iter(f"{_W}t") if t.text]
+        if runs:
+            paras.append("".join(runs))
+    text = "\n\n".join(paras).strip()
+    return text or None
+
+
+def _extract_pdf(path: str | Path) -> str | None:
+    """Extract text from a PDF: try the `pdftotext` binary first, then `pypdf`.
+    Returns None if neither is available or the PDF has no extractable text."""
+    # 1) pdftotext (poppler) — fast and common on Linux/macOS.
+    try:
+        r = subprocess.run(
+            ["pdftotext", "-q", str(path), "-"],
+            capture_output=True, text=True, timeout=120,
+        )
+        if r.returncode == 0 and r.stdout.strip():
+            return r.stdout
+    except (FileNotFoundError, subprocess.TimeoutExpired, OSError):
+        pass
+    # 2) pypdf (optional dependency: drydock-cli[pdf]).
+    try:
+        import pypdf
+
+        reader = pypdf.PdfReader(str(path))
+        text = "\n\n".join((pg.extract_text() or "") for pg in reader.pages).strip()
+        return text or None
+    except Exception:  # noqa: BLE001 — optional dep missing / unreadable PDF
+        return None
+
+
+def extract_document(path: str | Path) -> str | None:
+    """Extract text from a supported document (.pdf/.docx), or None."""
+    ext = Path(path).suffix.lower()
+    if ext == ".docx":
+        return _extract_docx(path)
+    if ext == ".pdf":
+        return _extract_pdf(path)
+    return None
+
+
+def pdf_backend_available() -> bool:
+    """Whether a PDF can be extracted at all (pdftotext binary or pypdf)."""
+    from shutil import which
+
+    if which("pdftotext"):
+        return True
+    try:
+        import pypdf  # noqa: F401
+
+        return True
+    except ImportError:
+        return False

{drydock_cli-3.0.60 → drydock_cli-3.0.62}/drydock/graphrag.py

@@ -26,6 +26,8 @@ import json
 import os
 import re
 from collections import Counter, defaultdict
+
+from drydock import extract
 from pathlib import Path
 
 # Files we ingest as text. Everything else (binaries, images) is skipped.
@@ -106,7 +108,8 @@ def _iter_text_files(paths: list[str]):
                 dirs[:] = [d for d in dirs if d not in _SKIP_DIRS]
                 for f in files:
                     fp = Path(root) / f
-                    if fp.suffix.lower() in _TEXT_EXT or fp.suffix == "":
+                    ext = fp.suffix.lower()
+                    if ext in _TEXT_EXT or ext in extract.EXTRACTABLE_EXT or fp.suffix == "":
                         yield fp
 
 
@@ -119,10 +122,15 @@ def _ingest_files(paths, cwd, chunks, entity_chunks, edges, skip_sources):
         rel = os.path.relpath(str(fp), cwd)
         if rel in skip_sources:
             continue
-        try:
-            text = fp.read_text("utf-8", "ignore")
-        except OSError:
-            continue
+        if fp.suffix.lower() in extract.EXTRACTABLE_EXT:
+            text = extract.extract_document(fp)  # PDF/Word → text (or None)
+            if not text:
+                continue  # unreadable / no PDF backend — skip cleanly
+        else:
+            try:
+                text = fp.read_text("utf-8", "ignore")
+            except OSError:
+                continue
         if not text.strip():
             continue
         added += 1

{drydock_cli-3.0.60 → drydock_cli-3.0.62}/drydock/skills.py

@@ -77,6 +77,25 @@ def skills_dirs(cwd: str) -> list[Path]:
     return [Path.home() / ".drydock" / "skills", Path(cwd) / ".drydock" / "skills"]
 
 
+def create_skill(name: str, body: str, *, description: str = "", scope: str = "user",
+                 cwd: str = ".") -> Path:
+    """Author a new skill markdown file and return its path. scope='user' writes
+    to ~/.drydock/skills (available everywhere); scope='project' writes to
+    <cwd>/.drydock/skills. Raises ValueError on a bad name / empty body."""
+    name = (name or "").strip().lower()
+    body = (body or "").strip()
+    if not _NAME_RE.match(name):
+        raise ValueError("skill name must be letters/digits/-/_ (e.g. 'review')")
+    if not body:
+        raise ValueError("a skill needs a prompt body")
+    d = (Path(cwd) if scope == "project" else Path.home()) / ".drydock" / "skills"
+    d.mkdir(parents=True, exist_ok=True)
+    path = d / f"{name}.md"
+    front = f"---\nname: {name}\ndescription: {description or 'user skill'}\n---\n"
+    path.write_text(front + body + "\n", encoding="utf-8")
+    return path
+
+
 def load_skills(cwd: str) -> dict[str, Skill]:
     """Load all skills; later dirs (project) override earlier (user) by name."""
     skills: dict[str, Skill] = {}

{drydock_cli-3.0.60 → drydock_cli-3.0.62}/drydock/tui/app.py

@@ -492,7 +492,7 @@ class DrydockApp(App):
                 "  /compact         shrink old context to free up the window\n"
                 "  /graphrag        ingest docs into a knowledge base the agent can use\n"
                 "                   build <path> · add <path> · query <q> · status · clear\n"
-                "  /skills          list your reusable /<name> skills\n"
+                "  /skills          list skills · /skills new <name> <prompt> to create one\n"
                 "  /loop            /loop <count> <prompt> — repeat a prompt (Esc stops)\n"
                 "  /mcp             list connected MCP servers and their tools\n"
                 "  /clear           reset the conversation\n"
@@ -504,7 +504,7 @@ class DrydockApp(App):
         elif cmd == "/loop":
             self._cmd_loop(arg)
         elif cmd == "/skills":
-            self._cmd_skills()
+            self._cmd_skills(arg)
         elif cmd[1:] in self._skills:
             self._run_skill(self._skills[cmd[1:]], arg)
         else:
@@ -554,19 +554,42 @@ class DrydockApp(App):
         self._mount(UserMessage(prompt))
         self._begin(prompt)
 
-    def _cmd_skills(self) -> None:
+    def _cmd_skills(self, arg: str = "") -> None:
+        """List skills, or create one: /skills new <name> <prompt…> (use $ARGS in
+        the prompt for trailing input). Created skills are usable as /<name>
+        immediately."""
+        from drydock import skills as skillsmod
+
+        parts = arg.split(maxsplit=2)
+        if parts and parts[0].lower() == "new":
+            if len(parts) < 3:
+                self._info(
+                    "usage: /skills new <name> <prompt text>\n"
+                    "  e.g.  /skills new review  Run GitDiff, then review the changes "
+                    "for bugs.\n  (use $ARGS in the prompt for trailing input)"
+                )
+                return
+            name, body = parts[1], parts[2]
+            try:
+                path = skillsmod.create_skill(name, body, cwd=self.config.get("cwd") or ".")
+            except ValueError as e:
+                self._info(f"Couldn't create skill: {e}")
+                return
+            self._skills = skillsmod.load_skills(self.config.get("cwd") or ".")  # reload
+            self._info(f"✓ Created skill /{name.lower()} ({path}). Invoke it as /{name.lower()}.")
+            return
         if not self._skills:
             self._info(
-                "No skills yet. Create one as a markdown file in "
-                "~/.drydock/skills/<name>.md (or <project>/.drydock/skills/), "
-                "then invoke it as /<name>. The body is the prompt; use $ARGS for "
-                "trailing input."
+                "No skills yet. Create one with:  /skills new <name> <prompt text>\n"
+                "(or drop a markdown file in ~/.drydock/skills/<name>.md). Invoke as "
+                "/<name>; use $ARGS for trailing input."
             )
             return
         lines = ["Skills (invoke as /<name>):"]
         for name in sorted(self._skills):
             sk = self._skills[name]
             lines.append(f"  /{name}" + (f"  — {sk.description}" if sk.description else ""))
+        lines.append("Create one:  /skills new <name> <prompt text>")
         self._info("\n".join(lines))
 
     def _run_skill(self, skill, arg: str) -> None:

{drydock_cli-3.0.60 → drydock_cli-3.0.62}/drydock/tuning.py

@@ -206,9 +206,33 @@ def filter_tool_schemas(tool_schemas: list, model: str | None) -> list:
     return [t for t in tool_schemas if t.get("name") not in GEMMA_DISABLED_TOOLS]
 
 
+# Reference for the model so it can answer "how do I …" questions about Drydock
+# itself. The user TYPES these slash commands; you (the model) do NOT run them —
+# you just explain the right one when asked. Appended to every system prompt.
+_DRYDOCK_COMMANDS_HELP = (
+    "\n\nAbout Drydock (the tool you run inside). If the user asks how to do "
+    "something with Drydock, tell them the slash command — they type it, you do "
+    "NOT run it:\n"
+    "- Knowledge base from their docs: `/graphrag build <path>` (a file or "
+    "folder), `/graphrag add <path>` to add more, `/graphrag query <q>` to test "
+    "it, `/graphrag status`, `/graphrag clear`. Once built, you automatically use "
+    "the `Knowledge` tool to draw on it. It ingests text formats (md/txt/code/"
+    "json/yaml/…) plus PDF and Word (.docx) directly.\n"
+    "- Custom skills (reusable `/<name>` prompts): `/skills new <name> <prompt "
+    "text>` creates one (use $ARGS in the prompt for trailing input); `/skills` "
+    "lists them; then they run it as `/<name>`.\n"
+    "- Other: `/model` (model/endpoint), `/cwd`, `/undo` (revert last write), "
+    "`/back` (rewind a turn), `/compact` (shrink context), `/loop <n> <prompt>` "
+    "(repeat a prompt), `/mcp` (connected MCP servers), `/status`, `/clear`, "
+    "`/help`, `/quit`. Internet + git are tools you call yourself "
+    "(WebSearch/WebFetch, GitStatus/GitDiff/GitLog/GitCommit)."
+)
+
+
 def system_prompt_for_model(model: str | None) -> str:
     """Return the system prompt best suited to the model."""
-    return _GEMMA_SYSTEM_PROMPT if is_gemma(model) else _DEFAULT_SYSTEM_PROMPT
+    base = _GEMMA_SYSTEM_PROMPT if is_gemma(model) else _DEFAULT_SYSTEM_PROMPT
+    return base + _DRYDOCK_COMMANDS_HELP
 
 
 def thinking_level_for_turn(turn_count: int, is_user_turn: bool) -> str:

{drydock_cli-3.0.60 → drydock_cli-3.0.62/drydock_cli.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drydock-cli
-Version: 3.0.60
+Version: 3.0.62
 Summary: Drydock — a local, provider-agnostic terminal coding agent for local LLMs
 Author: Frank Bobe III
 License-Expression: Apache-2.0
@@ -18,6 +18,8 @@ Requires-Dist: pytest; extra == "dev"
 Requires-Dist: pytest-timeout; extra == "dev"
 Requires-Dist: ruff; extra == "dev"
 Requires-Dist: pyright; extra == "dev"
+Provides-Extra: pdf
+Requires-Dist: pypdf>=4; extra == "pdf"
 Dynamic: license-file
 
 # ⚓ Drydock
@@ -71,6 +73,53 @@ A full agentic CLI harness — every tool below is clean-room and dependency-fre
   `~/.drydock/skills/` (or `<project>/.drydock/skills/`); `$ARGS` substitution.
 - **Loops** — `/loop <count> <prompt>` runs a prompt iteratively (Esc stops).
 
+## Slash commands
+
+Typed into the prompt. The agent also knows these, so you can just **ask it**
+("how do I add my own docs?") and it'll point you to the right one.
+
+| Command | What it does |
+| --- | --- |
+| `/graphrag build <path>` | Build a knowledge base from a file or folder of docs/code |
+| `/graphrag add <path>` | Incrementally add more documents to the base |
+| `/graphrag query <q>` | Test what the base returns (no model) |
+| `/graphrag status` · `clear` | List indexed sources · wipe the base |
+| `/skills` | List your skills |
+| `/skills new <name> <prompt>` | Create a reusable `/<name>` skill (use `$ARGS` for input) |
+| `/<name>` | Run a skill |
+| `/loop <count> <prompt>` | Repeat a prompt N times (Esc stops) |
+| `/mcp` | List connected MCP servers + their tools |
+| `/model` · `/cwd` | Show/set model & endpoint · working directory |
+| `/undo` · `/back` | Revert the last write · rewind the last turn |
+| `/compact` · `/status` · `/clear` | Shrink context · session stats · reset |
+| `/help` · `/quit` | Help · exit |
+
+### Knowledge base (GraphRAG) — ingesting your documents
+
+```
+/graphrag build ./docs        # index a file or a whole folder
+/graphrag add ./more_docs     # add more later, incrementally
+/graphrag query "how are refunds handled?"   # check retrieval
+/graphrag status              # what's indexed
+```
+
+Once built, the agent **automatically** retrieves from it (read-only `Knowledge`
+tool) when a question touches your material. Ingests text formats
+(`.md .txt .py .js .json .yaml .sql …`) **plus PDF and Word (`.docx`)**. `.docx`
+needs nothing extra; PDF uses the `pdftotext` binary (poppler) if present, else
+`pip install drydock-cli[pdf]` (pypdf). The index is a single JSON at
+`<project>/.drydock/graphrag.json` — clean-room, no embeddings.
+
+### Custom skills
+
+```
+/skills new commitmsg  Write a concise conventional-commit message for: $ARGS
+/commitmsg the staged auth changes      # runs the skill with $ARGS substituted
+```
+
+Skills are markdown files in `~/.drydock/skills/` (personal) or
+`<project>/.drydock/skills/` (project); `/skills new` writes one for you.
+
 ## Install
 
 ```bash

{drydock_cli-3.0.60 → drydock_cli-3.0.62}/drydock_cli.egg-info/SOURCES.txt

@@ -10,6 +10,7 @@ drydock/cli.py
 drydock/compaction.py
 drydock/config.py
 drydock/detect.py
+drydock/extract.py
 drydock/gittools.py
 drydock/graphrag.py
 drydock/guards.py
@@ -47,6 +48,7 @@ tests/test_context_limit_config.py
 tests/test_detect.py
 tests/test_dispatch.py
 tests/test_empty_response.py
+tests/test_extract.py
 tests/test_failure_loop.py
 tests/test_first_run_setup.py
 tests/test_gittools.py
@@ -65,6 +67,7 @@ tests/test_skills.py
 tests/test_stop.py
 tests/test_streaming_newlines.py
 tests/test_subagent.py
+tests/test_system_prompt_help.py
 tests/test_todo.py
 tests/test_tool_arg_parsing.py
 tests/test_tools_undo.py

{drydock_cli-3.0.60 → drydock_cli-3.0.62}/drydock_cli.egg-info/requires.txt

@@ -6,3 +6,6 @@ pytest
 pytest-timeout
 ruff
 pyright
+
+[pdf]
+pypdf>=4

{drydock_cli-3.0.60 → drydock_cli-3.0.62}/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 = "3.0.60"
+version = "3.0.62"
 description = "Drydock — a local, provider-agnostic terminal coding agent for local LLMs"
 readme = "README.md"
 requires-python = ">=3.11"
@@ -29,6 +29,9 @@ drydock = "drydock.cli:main"
 
 [project.optional-dependencies]
 dev = ["pytest", "pytest-timeout", "ruff", "pyright"]
+# PDF ingestion fallback when the `pdftotext` binary (poppler) isn't installed.
+# .docx needs nothing extra (stdlib zip/xml). `pip install drydock-cli[pdf]`.
+pdf = ["pypdf>=4"]
 
 [tool.ruff]
 target-version = "py312"

drydock_cli-3.0.62/tests/test_extract.py

@@ -0,0 +1,71 @@
+"""PDF + Word (.docx) document extraction for GraphRAG ingestion."""
+from __future__ import annotations
+
+import zipfile
+
+import pytest
+
+from drydock import extract, graphrag
+
+
+def _make_docx(path, text):
+    doc = ('<?xml version="1.0"?><w:document xmlns:w='
+           '"http://schemas.openxmlformats.org/wordprocessingml/2006/main">'
+           f'<w:body><w:p><w:r><w:t>{text}</w:t></w:r></w:p></w:body></w:document>')
+    with zipfile.ZipFile(path, "w") as z:
+        z.writestr("word/document.xml", doc)
+
+
+def _make_pdf(path, text):
+    objs = [b"<< /Type /Catalog /Pages 2 0 R >>",
+            b"<< /Type /Pages /Kids [3 0 R] /Count 1 >>",
+            b"<< /Type /Page /Parent 2 0 R /MediaBox [0 0 612 792] /Contents 4 0 R "
+            b"/Resources << /Font << /F1 5 0 R >> >> >>"]
+    stream = b"BT /F1 18 Tf 72 700 Td (" + text.encode() + b") Tj ET"
+    objs.append(b"<< /Length " + str(len(stream)).encode() + b" >>\nstream\n" + stream + b"\nendstream")
+    objs.append(b"<< /Type /Font /Subtype /Type1 /BaseFont /Helvetica >>")
+    out = b"%PDF-1.4\n"; offs = []
+    for i, o in enumerate(objs, 1):
+        offs.append(len(out)); out += str(i).encode() + b" 0 obj\n" + o + b"\nendobj\n"
+    x = len(out); out += b"xref\n0 " + str(len(objs) + 1).encode() + b"\n0000000000 65535 f \n"
+    for o in offs:
+        out += ("%010d 00000 n \n" % o).encode()
+    out += (b"trailer\n<< /Size " + str(len(objs) + 1).encode() + b" /Root 1 0 R >>\n"
+            b"startxref\n" + str(x).encode() + b"\n%%EOF")
+    path.write_bytes(out)
+
+
+def test_docx_extraction(tmp_path):
+    p = tmp_path / "d.docx"
+    _make_docx(p, "The Nimbus service uses the Falcon-Token header.")
+    assert extract.extract_document(p) == "The Nimbus service uses the Falcon-Token header."
+
+
+def test_pdf_extraction(tmp_path):
+    if not extract.pdf_backend_available():
+        pytest.skip("no PDF backend (pdftotext/pypdf) available")
+    p = tmp_path / "m.pdf"
+    _make_pdf(p, "Refunds via the Zephyr gateway.")
+    out = extract.extract_document(p)
+    assert out and "Zephyr gateway" in out
+
+
+def test_unsupported_and_corrupt_return_none(tmp_path):
+    assert extract.extract_document(tmp_path / "x.png") is None
+    bad = tmp_path / "bad.docx"; bad.write_text("not a zip")
+    assert extract.extract_document(bad) is None
+
+
+def test_graphrag_ingests_docx_and_pdf(tmp_path):
+    _make_docx(tmp_path / "spec.docx", "The Nimbus API uses the Falcon-Token header for auth.")
+    if extract.pdf_backend_available():
+        _make_pdf(tmp_path / "manual.pdf", "Refunds are issued via the Zephyr gateway.")
+    store = tmp_path / ".drydock" / "graphrag.json"
+    graphrag.build_index(["."], store, cwd=str(tmp_path))
+    idx = graphrag.load_index(store)
+    srcs = graphrag.sources(idx)
+    assert "spec.docx" in srcs
+    assert graphrag.query_index(idx, "Falcon-Token header")["chunks"][0]["source"] == "spec.docx"
+    if extract.pdf_backend_available():
+        assert "manual.pdf" in srcs
+        assert graphrag.query_index(idx, "refunds Zephyr gateway")["chunks"][0]["source"] == "manual.pdf"

{drydock_cli-3.0.60 → drydock_cli-3.0.62}/tests/test_skills.py

@@ -54,3 +54,28 @@ def test_empty_body_skipped(tmp_path):
 
 def test_no_skills_dir_is_empty(tmp_path):
     assert skills.load_skills(str(tmp_path)) == {}
+
+
+def test_create_skill_user_scope(tmp_path, monkeypatch):
+    monkeypatch.setattr(skills.Path, "home", lambda: tmp_path)
+    path = skills.create_skill("review", "Run GitDiff then review for bugs.",
+                               description="Review the diff")
+    assert path.exists()
+    loaded = skills.load_skills(str(tmp_path / "proj"))  # user-scope visible anywhere
+    assert "review" in loaded and "review for bugs" in loaded["review"].body
+    assert loaded["review"].description == "Review the diff"
+
+
+def test_create_skill_rejects_bad_name_and_empty_body(tmp_path, monkeypatch):
+    monkeypatch.setattr(skills.Path, "home", lambda: tmp_path)
+    import pytest
+    with pytest.raises(ValueError):
+        skills.create_skill("bad name!", "body")
+    with pytest.raises(ValueError):
+        skills.create_skill("ok", "   ")
+
+
+def test_create_skill_project_scope(tmp_path):
+    path = skills.create_skill("explain", "Explain $ARGS.", scope="project", cwd=str(tmp_path))
+    assert str(tmp_path) in str(path)
+    assert skills.load_skills(str(tmp_path))["explain"].render("auth.py") == "Explain auth.py."

drydock_cli-3.0.62/tests/test_system_prompt_help.py

@@ -0,0 +1,15 @@
+"""The system prompt must teach the model Drydock's own slash commands so it can
+answer 'how do I add documents / make a skill' when a user asks."""
+from __future__ import annotations
+
+from drydock.tuning import system_prompt_for_model
+
+
+def test_prompt_documents_graphrag_and_skills():
+    for model in ("gemma4", "some-other-model"):
+        p = system_prompt_for_model(model)
+        assert "/graphrag build" in p and "/graphrag add" in p
+        assert "/skills new" in p
+        assert "Knowledge" in p
+        # framed so the model explains, not runs, the command
+        assert "you do" in p.lower() and "not run" in p.lower()