drydock-cli 3.0.111__tar.gz → 3.0.113__tar.gz

{drydock_cli-3.0.111/drydock_cli.egg-info → drydock_cli-3.0.113}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drydock-cli
-Version: 3.0.111
+Version: 3.0.113
 Summary: Drydock — a local, provider-agnostic terminal coding agent for local LLMs
 Author: Frank Bobe III
 License-Expression: Apache-2.0

{drydock_cli-3.0.111 → drydock_cli-3.0.113}/drydock/graphrag.py

@@ -101,9 +101,20 @@ def _chunk_text(text: str, max_chars: int = 1200) -> list[str]:
     return chunks
 
 
+def _unquote(p: str) -> str:
+    """Strip surrounding quotes a user wrapped a path in — common on Windows:
+    `/graphrag build "C:\\Users\\me\\Documents"`. Without this the quote chars
+    become part of the path, isabs() is False, it's joined under cwd, and nothing
+    is found ("No text found. Nothing was indexed.")."""
+    p = p.strip()
+    if len(p) >= 2 and p[0] == p[-1] and p[0] in "\"'":
+        p = p[1:-1]
+    return p
+
+
 def _iter_text_files(paths: list[str]):
     for p in paths:
-        path = Path(p)
+        path = Path(_unquote(p))
         if path.is_file():
             yield path
         elif path.is_dir():
@@ -120,27 +131,31 @@ def _ingest_files(paths, cwd, chunks, entity_chunks, edges, skip_sources):
     """Chunk + extract + graph every text file under paths into the (mutable)
     accumulators, skipping any source already present. Returns files added."""
     added = 0
+    cleaned = [_unquote(p) for p in paths]
     for fp in _iter_text_files([str(Path(cwd) / p) if not os.path.isabs(p) else p
-                                for p in paths]):
-        rel = os.path.relpath(str(fp), cwd)
-        if rel in skip_sources:
-            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:
+                                for p in cleaned]):
+        # Per-file isolation: ONE unreadable/odd file (a locked Word doc, a weird
+        # encoding, a PDF backend quirk) must never crash a whole-folder build —
+        # skip it and keep indexing the rest.
+        try:
+            rel = os.path.relpath(str(fp), cwd)
+            if rel in skip_sources:
                 continue
-        if not text.strip():
+            if fp.suffix.lower() in extract.EXTRACTABLE_EXT:
+                text = extract.extract_document(fp)  # PDF/Word → text (or None)
+            else:
+                text = fp.read_text("utf-8", "ignore")
+            if not text or not text.strip():
+                continue  # unreadable / empty — skip cleanly
+            # Build this file's chunks locally; commit only if it ALL succeeds so a
+            # mid-file failure leaves no partial state and isn't miscounted.
+            local = [(body, extract_entities(body)) for body in _chunk_text(text)]
+        except Exception:  # noqa: BLE001 — isolate a bad file, never abort the build
             continue
         added += 1
         skip_sources.add(rel)  # don't double-ingest the same file in one call
-        for body in _chunk_text(text):
+        for body, ents in local:
             cid = len(chunks)
-            ents = extract_entities(body)
             chunks.append({"id": cid, "source": rel, "text": body, "entities": ents})
             for e in ents:
                 entity_chunks[e].add(cid)

{drydock_cli-3.0.111 → drydock_cli-3.0.113}/drydock/tools/__init__.py

@@ -607,6 +607,34 @@ SCHEMAS = [
             "required": ["query"],
         },
     },
+    {
+        "name": "BuildKnowledge",
+        "description": (
+            "BUILD or extend the user's GraphRAG knowledge base FROM THEIR documents "
+            "or code. Call this YOURSELF when the user asks you to index / ingest / "
+            "'make a knowledge base from' a folder or file (e.g. 'create a knowledge "
+            "base from my Documents'). Give `path` (a file or directory). mode 'build' "
+            "= (re)build from scratch; 'add' = merge into an existing base. When it "
+            "succeeds you can search it with the Knowledge tool. Do NOT tell the user "
+            "to type a '/graphrag' slash command — actually do it with this tool."
+        ),
+        "input_schema": {
+            "type": "object",
+            "properties": {
+                "path": {
+                    "type": "string",
+                    "description": "File or directory to index. A normal path — on "
+                                   "Windows e.g. C:\\Users\\me\\Documents (quotes optional).",
+                },
+                "mode": {
+                    "type": "string",
+                    "description": "'build' (rebuild from scratch, default) or 'add' "
+                                   "(merge into the existing base).",
+                },
+            },
+            "required": ["path"],
+        },
+    },
 ]
 
 # ── Tool implementations ──────────────────────────────────────────────────
@@ -1741,6 +1769,34 @@ def tool_knowledge(params: dict, config: dict) -> str:
     return graphrag.format_results(result, query)
 
 
+def tool_build_knowledge(params: dict, config: dict) -> str:
+    """Build/extend the GraphRAG knowledge base from a file or directory of docs —
+    the model-callable equivalent of `/graphrag build|add`, so the agent can do it
+    itself when asked instead of telling the user to run a slash command."""
+    from drydock import graphrag
+
+    path = _as_str_arg(params.get("path")).strip()
+    if not path:
+        return ("Error: `BuildKnowledge` needs a `path` — a file or directory of "
+                "documents/code to index.")
+    mode = (_as_str_arg(params.get("mode")) or "build").strip().lower()
+    if mode not in ("build", "add"):
+        mode = "build"
+    cwd = config.get("cwd") or os.getcwd()
+    store = config.get("graphrag_store") or graphrag.default_store_path(cwd)
+    try:
+        fn = graphrag.build_index if mode == "build" else graphrag.add_to_index
+        stats = fn([path], store, cwd=cwd)
+    except Exception as e:  # noqa: BLE001 — report, never crash the agent loop
+        return f"Error building the knowledge base from {path}: {e}"
+    if not stats.get("chunks"):
+        return (f"No indexable text found under {path}. Nothing was added. (Supported: "
+                "text/code files, .docx, .pdf, .md, etc. Check the path is correct.)")
+    verb = "Built" if mode == "build" else "Updated"
+    return (f"{verb} the knowledge base: {stats['files']} files, {stats['chunks']} "
+            f"chunks, {stats['entities']} entities. Search it with the Knowledge tool.")
+
+
 # ── Register all tools ────────────────────────────────────────────────────
 
 _TOOLS = [
@@ -1754,6 +1810,7 @@ _TOOLS = [
     ("task", tool_task, True),
     ("Dispatch", tool_dispatch, True),
     ("Knowledge", tool_knowledge, True),
+    ("BuildKnowledge", tool_build_knowledge, False),
     ("GraphQuery", tool_graphquery, True),
     ("GraphAdd", tool_graphadd, False),
     ("StigRules", tool_stigrules, True),
@@ -1776,6 +1833,7 @@ def register_all():
             "Bash": tool_bash, "Glob": tool_glob, "Grep": tool_grep,
             "todo": tool_todo, "task": tool_task, "Dispatch": tool_dispatch,
             "Consult": tool_consult, "Knowledge": tool_knowledge,
+            "BuildKnowledge": tool_build_knowledge,
             "GraphQuery": tool_graphquery, "GraphAdd": tool_graphadd,
             "StigRules": tool_stigrules, "StigRule": tool_stigrule, "StigSet": tool_stigset,
             "WebSearch": tool_websearch, "WebFetch": tool_webfetch,

{drydock_cli-3.0.111 → drydock_cli-3.0.113/drydock_cli.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drydock-cli
-Version: 3.0.111
+Version: 3.0.113
 Summary: Drydock — a local, provider-agnostic terminal coding agent for local LLMs
 Author: Frank Bobe III
 License-Expression: Apache-2.0

{drydock_cli-3.0.111 → drydock_cli-3.0.113}/drydock_cli.egg-info/SOURCES.txt

@@ -80,6 +80,7 @@ tests/test_gittools.py
 tests/test_glob_edit_edges.py
 tests/test_graphify_example.py
 tests/test_graphrag.py
+tests/test_graphrag_quoted_path.py
 tests/test_grep_and_read_robust.py
 tests/test_guards_and_tools.py
 tests/test_hallucinated_tools.py

{drydock_cli-3.0.111 → drydock_cli-3.0.113}/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 = "3.0.111"
+version = "3.0.113"
 description = "Drydock — a local, provider-agnostic terminal coding agent for local LLMs"
 readme = "README.md"
 requires-python = ">=3.11"

drydock_cli-3.0.113/tests/test_graphrag_quoted_path.py

@@ -0,0 +1,76 @@
+"""A user wrapping the path in quotes — common on Windows, e.g.
+`/graphrag build "C:\\Users\\me\\Documents"` — must still index. Before the fix
+the quote chars became part of the path, isabs() was False, it was joined under
+cwd, and nothing was found ('No text found. Nothing was indexed.')."""
+from __future__ import annotations
+
+import os
+import tempfile
+
+from drydock import graphrag
+
+
+def _make_docs():
+    src = tempfile.mkdtemp()
+    open(os.path.join(src, "notes.txt"), "w").write("Apollo landed on the Moon in 1969.")
+    open(os.path.join(src, "readme.md"), "w").write("# Project\nDrydock is a coding agent.")
+    store = os.path.join(tempfile.mkdtemp(), "kb.json")  # store OUTSIDE the scanned dir
+    return src, store
+
+
+def test_double_quoted_path_indexes():
+    src, store = _make_docs()
+    stats = graphrag.build_index([f'"{src}"'], store)
+    assert stats["files"] == 2 and stats["chunks"] >= 2
+
+
+def test_single_quoted_path_indexes():
+    src, store = _make_docs()
+    stats = graphrag.build_index([f"'{src}'"], store)
+    assert stats["files"] == 2
+
+
+def test_unquoted_path_still_works():
+    src, store = _make_docs()
+    assert graphrag.build_index([src], store)["files"] == 2
+
+
+def test_unquote_helper():
+    assert graphrag._unquote('"C:\\Users\\x"') == "C:\\Users\\x"
+    assert graphrag._unquote("'/a/b'") == "/a/b"
+    assert graphrag._unquote("/plain/path") == "/plain/path"
+    assert graphrag._unquote('  "spaced"  ') == "spaced"
+    assert graphrag._unquote('"') == '"'  # lone quote unchanged
+
+
+def test_one_bad_file_does_not_abort_build():
+    """A single file that raises mid-processing must be skipped, not crash the
+    whole folder build (the Windows 'NoneType.strip' report)."""
+    from unittest import mock
+    src, store = _make_docs()
+    open(os.path.join(src, "boom.txt"), "w").write("EXPLODE here")
+    real = graphrag.extract_entities
+
+    def boom(text):
+        if "EXPLODE" in text:
+            raise AttributeError("'NoneType' object has no attribute 'strip'")
+        return real(text)
+
+    with mock.patch.object(graphrag, "extract_entities", boom):
+        stats = graphrag.build_index([src], store)
+    assert stats["files"] == 2  # the 2 good files indexed, boom.txt skipped (not counted)
+
+
+def test_build_knowledge_tool_builds_and_reports():
+    from drydock import tools
+    src, store = _make_docs()
+    out = tools.tool_build_knowledge({"path": f'"{src}"', "mode": "build"},
+                                     {"cwd": os.getcwd(), "graphrag_store": store})
+    assert "Built the knowledge base" in out and "2 files" in out
+
+
+def test_build_knowledge_available_to_gemma():
+    from drydock import tool_registry
+    from drydock.tuning import filter_tool_schemas
+    names = [s["name"] for s in filter_tool_schemas(tool_registry.schemas(), "gemma4")]
+    assert "BuildKnowledge" in names