rlmgrep 0.1.8__tar.gz → 0.1.17__tar.gz

{rlmgrep-0.1.8 → rlmgrep-0.1.17}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: rlmgrep
-Version: 0.1.8
+Version: 0.1.17
 Summary: Grep-shaped CLI search powered by DSPy RLM
 Author: rlmgrep
 License: MIT
@@ -8,11 +8,12 @@ Requires-Python: >=3.11
 Description-Content-Type: text/markdown
 Requires-Dist: dspy>=3.1.1
 Requires-Dist: markitdown[all]>=0.1.4
+Requires-Dist: pathspec>=0.12.1
 Requires-Dist: pypdf>=4.0.0
 
 # rlmgrep
 
-Grep-shaped search powered by DSPy RLM. It accepts a natural-language query, scans the files you point at, and prints matching lines in a grep-like format.
+Grep-shaped search powered by DSPy RLM. It accepts a natural-language query, scans the files you point at, and prints matching lines in a grep-like format. Use `--answer` to get a narrative response grounded in the selected files/directories.
 
 ## Quickstart
 
@@ -22,9 +23,20 @@ uv tool install rlmgrep
 # uv tool install git+https://github.com/halfprice06/rlmgrep.git
 
 export OPENAI_API_KEY=...  # or set keys in ~/.rlmgrep
-rlmgrep "where are API keys read" rlmgrep/
 ```
 
+```sh
+rlmgrep --answer "What does this repo do and where are the entry points?" .
+```
+
+![Quickstart answer mode](docs/images/quickstart-answer.png)
+
+```sh
+rlmgrep -C 2 "Where is retry/backoff configured and what are the defaults?" .
+```
+
+![Quickstart context mode](docs/images/quickstart-context.png)
+
 ## Requirements
 
 - Python 3.11+
@@ -38,8 +50,8 @@ One of rlmgrep’s most useful features is that it can “grep” **PDFs and Off
 How it works:
 - **PDFs** are parsed with `pypdf`. Each page gets a marker line like `===== Page N =====`, and output lines include a `page=N` suffix. Line numbers refer to the extracted text (not PDF coordinates).
 - **Office & binary docs** (`.docx`, `.pptx`, `.xlsx`, `.html`, `.zip`, etc.) are converted to Markdown via **MarkItDown**. This happens during ingestion, so rlmgrep can search them like any other text file.
-- **Images** can be described by a vision model through MarkItDown (OpenAI/Anthropic/Gemini).
-- **Audio** transcription is supported through OpenAI when enabled.
+- **Images** can be described by a vision model and then searched through MarkItDown (OpenAI/Anthropic/Gemini), enable and configure in config.toml.
+- **Audio** transcription is supported through OpenAI when enabled, configure in config.toml.
 
 Sidecar caching:
 - For images/audio, converted text is cached next to the original file as `<original>.<ext>.md` and reused on later runs.
@@ -47,7 +59,7 @@ Sidecar caching:
 
 ## Install Deno
 
-DSPy requires the Deno runtime. Install it with the official scripts:
+DSPy's default implementation of RLM requires the Deno runtime. Install it with the official scripts:
 
 macOS/Linux:
 
@@ -75,12 +87,15 @@ rlmgrep [options] "query" [paths...]
 
 Common options:
 
+- `--answer` return a narrative answer before the grep output
 - `-C N` context lines before/after (grep-style)
 - `-A N` context lines after
 - `-B N` context lines before
 - `-m N` max matching lines per file
 - `-g GLOB` include files matching glob (repeatable, comma-separated)
 - `--type T` include file types (repeatable, comma-separated)
+- `--hidden` include hidden files and directories
+- `--no-ignore` do not respect `.gitignore`
 - `--no-recursive` do not recurse directories
 - `-a`, `--text` treat binary files as text
 - `-y`, `--yes` skip file count confirmation
@@ -95,7 +110,7 @@ Examples:
 
 ```sh
 # Natural-language query over a repo
-rlmgrep -n -C 2 "Where is retry/backoff configured and what are the defaults?" .
+rlmgrep -C 2 "Where is retry/backoff configured and what are the defaults?" .
 
 # Restrict to Python files
 rlmgrep "Where do we parse JWTs and enforce expiration?" --type py .
@@ -113,6 +128,7 @@ rg -l "token" . | rlmgrep --files-from-stdin --answer "What does this token cont
 ## Input selection
 
 - Directories are searched recursively by default. Use `--no-recursive` to stop recursion.
+- Hidden files and ignore files (`.gitignore`, `.ignore`, `.rgignore`) are respected by default. Use `--hidden` or `--no-ignore` to include them.
 - `--type` uses built-in type mappings (e.g., `py`, `js`, `md`); unknown values are treated as file extensions.
 - `-g/--glob` matches path globs against normalized paths (forward slashes).
 - Paths are printed relative to the current working directory when possible.
@@ -125,7 +141,7 @@ rg -l "token" . | rlmgrep --files-from-stdin --answer "What does this token cont
 - Output uses rg-style headings by default:
   - A file header line like `./path/to/file`
   - Then `line:\ttext` for matches, `line-\ttext` for context lines
-- Line numbers are 1-based.
+- Line numbers are always included and are 1-based.
 - When context ranges are disjoint, a `--` line separates groups.
 - Exit codes:
   - `0` = at least one match
@@ -140,7 +156,7 @@ rlmgrep can interpret traditional regex-style patterns inside a natural-language
 Example (best-effort regex semantics + extra context):
 
 ```sh
-rlmgrep -n "Find Python functions that look like `def test_\\w+` and are marked as slow or flaky in nearby comments." .
+rlmgrep "Find Python functions that look like `def test_\\w+` and are marked as slow or flaky in nearby comments." .
 ```
 
 If you need strict, deterministic regex behavior, use `rg`/`grep`.

{rlmgrep-0.1.8 → rlmgrep-0.1.17}/README.md

@@ -1,6 +1,6 @@
 # rlmgrep
 
-Grep-shaped search powered by DSPy RLM. It accepts a natural-language query, scans the files you point at, and prints matching lines in a grep-like format.
+Grep-shaped search powered by DSPy RLM. It accepts a natural-language query, scans the files you point at, and prints matching lines in a grep-like format. Use `--answer` to get a narrative response grounded in the selected files/directories.
 
 ## Quickstart
 
@@ -10,9 +10,20 @@ uv tool install rlmgrep
 # uv tool install git+https://github.com/halfprice06/rlmgrep.git
 
 export OPENAI_API_KEY=...  # or set keys in ~/.rlmgrep
-rlmgrep "where are API keys read" rlmgrep/
 ```
 
+```sh
+rlmgrep --answer "What does this repo do and where are the entry points?" .
+```
+
+![Quickstart answer mode](docs/images/quickstart-answer.png)
+
+```sh
+rlmgrep -C 2 "Where is retry/backoff configured and what are the defaults?" .
+```
+
+![Quickstart context mode](docs/images/quickstart-context.png)
+
 ## Requirements
 
 - Python 3.11+
@@ -26,8 +37,8 @@ One of rlmgrep’s most useful features is that it can “grep” **PDFs and Off
 How it works:
 - **PDFs** are parsed with `pypdf`. Each page gets a marker line like `===== Page N =====`, and output lines include a `page=N` suffix. Line numbers refer to the extracted text (not PDF coordinates).
 - **Office & binary docs** (`.docx`, `.pptx`, `.xlsx`, `.html`, `.zip`, etc.) are converted to Markdown via **MarkItDown**. This happens during ingestion, so rlmgrep can search them like any other text file.
-- **Images** can be described by a vision model through MarkItDown (OpenAI/Anthropic/Gemini).
-- **Audio** transcription is supported through OpenAI when enabled.
+- **Images** can be described by a vision model and then searched through MarkItDown (OpenAI/Anthropic/Gemini), enable and configure in config.toml.
+- **Audio** transcription is supported through OpenAI when enabled, configure in config.toml.
 
 Sidecar caching:
 - For images/audio, converted text is cached next to the original file as `<original>.<ext>.md` and reused on later runs.
@@ -35,7 +46,7 @@ Sidecar caching:
 
 ## Install Deno
 
-DSPy requires the Deno runtime. Install it with the official scripts:
+DSPy's default implementation of RLM requires the Deno runtime. Install it with the official scripts:
 
 macOS/Linux:
 
@@ -63,12 +74,15 @@ rlmgrep [options] "query" [paths...]
 
 Common options:
 
+- `--answer` return a narrative answer before the grep output
 - `-C N` context lines before/after (grep-style)
 - `-A N` context lines after
 - `-B N` context lines before
 - `-m N` max matching lines per file
 - `-g GLOB` include files matching glob (repeatable, comma-separated)
 - `--type T` include file types (repeatable, comma-separated)
+- `--hidden` include hidden files and directories
+- `--no-ignore` do not respect `.gitignore`
 - `--no-recursive` do not recurse directories
 - `-a`, `--text` treat binary files as text
 - `-y`, `--yes` skip file count confirmation
@@ -83,7 +97,7 @@ Examples:
 
 ```sh
 # Natural-language query over a repo
-rlmgrep -n -C 2 "Where is retry/backoff configured and what are the defaults?" .
+rlmgrep -C 2 "Where is retry/backoff configured and what are the defaults?" .
 
 # Restrict to Python files
 rlmgrep "Where do we parse JWTs and enforce expiration?" --type py .
@@ -101,6 +115,7 @@ rg -l "token" . | rlmgrep --files-from-stdin --answer "What does this token cont
 ## Input selection
 
 - Directories are searched recursively by default. Use `--no-recursive` to stop recursion.
+- Hidden files and ignore files (`.gitignore`, `.ignore`, `.rgignore`) are respected by default. Use `--hidden` or `--no-ignore` to include them.
 - `--type` uses built-in type mappings (e.g., `py`, `js`, `md`); unknown values are treated as file extensions.
 - `-g/--glob` matches path globs against normalized paths (forward slashes).
 - Paths are printed relative to the current working directory when possible.
@@ -113,7 +128,7 @@ rg -l "token" . | rlmgrep --files-from-stdin --answer "What does this token cont
 - Output uses rg-style headings by default:
   - A file header line like `./path/to/file`
   - Then `line:\ttext` for matches, `line-\ttext` for context lines
-- Line numbers are 1-based.
+- Line numbers are always included and are 1-based.
 - When context ranges are disjoint, a `--` line separates groups.
 - Exit codes:
   - `0` = at least one match
@@ -128,7 +143,7 @@ rlmgrep can interpret traditional regex-style patterns inside a natural-language
 Example (best-effort regex semantics + extra context):
 
 ```sh
-rlmgrep -n "Find Python functions that look like `def test_\\w+` and are marked as slow or flaky in nearby comments." .
+rlmgrep "Find Python functions that look like `def test_\\w+` and are marked as slow or flaky in nearby comments." .
 ```
 
 If you need strict, deterministic regex behavior, use `rg`/`grep`.

{rlmgrep-0.1.8 → rlmgrep-0.1.17}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "rlmgrep"
-version = "0.1.8"
+version = "0.1.17"
 description = "Grep-shaped CLI search powered by DSPy RLM"
 readme = "README.md"
 requires-python = ">=3.11"
@@ -9,6 +9,7 @@ license = { text = "MIT" }
 dependencies = [
   "dspy>=3.1.1",
   "markitdown[all]>=0.1.4",
+  "pathspec>=0.12.1",
   "pypdf>=4.0.0",
 ]
 

{rlmgrep-0.1.8 → rlmgrep-0.1.17}/rlmgrep/__init__.py

@@ -1,2 +1,2 @@
 __all__ = ["__version__"]
-__version__ = "0.1.8"
+__version__ = "0.1.17"

{rlmgrep-0.1.8 → rlmgrep-0.1.17}/rlmgrep/cli.py

@@ -3,13 +3,21 @@ from __future__ import annotations
 import argparse
 import os
 import sys
+import shutil
+import subprocess
 from pathlib import Path
 
 import dspy
 from . import __version__
 from .config import ensure_default_config, load_config
 from .file_map import build_file_map
-from .ingest import FileRecord, collect_candidates, load_files, resolve_type_exts
+from .ingest import (
+    FileRecord,
+    build_ignore_spec,
+    collect_candidates,
+    load_files,
+    resolve_type_exts,
+)
 from .rlm import Match, build_lm, run_rlm
 from .render import render_matches
 
@@ -72,16 +80,22 @@ def _parse_args(argv: list[str]) -> argparse.Namespace:
     parser.add_argument("pattern", nargs="?", help="Query string (interpreted by RLM)")
     parser.add_argument("paths", nargs="*", help="Files or directories")
 
-    parser.add_argument("-n", dest="line_numbers", action="store_true", help="Show line numbers (default)")
     parser.add_argument("-r", dest="recursive", action="store_true", help="Recursive (directories are searched recursively by default)")
     parser.add_argument("--no-recursive", dest="recursive", action="store_false", help="Do not recurse directories")
-    parser.set_defaults(recursive=True, line_numbers=True)
+    parser.set_defaults(recursive=True)
 
     parser.add_argument("-C", dest="context", type=int, default=0, help="Context lines before/after")
     parser.add_argument("-A", dest="after", type=int, default=None, help="Context lines after")
     parser.add_argument("-B", dest="before", type=int, default=None, help="Context lines before")
     parser.add_argument("-m", dest="max_count", type=int, default=None, help="Max matching lines per file")
     parser.add_argument("-a", "--text", dest="binary_as_text", action="store_true", help="Search binary files as text")
+    parser.add_argument("--hidden", action="store_true", help="Include hidden files and directories")
+    parser.add_argument(
+        "--no-ignore",
+        dest="no_ignore",
+        action="store_true",
+        help="Do not respect ignore files (.gitignore/.ignore/.rgignore)",
+    )
     parser.add_argument("--answer", action="store_true", help="Print a narrative answer before grep output")
     parser.add_argument("-y", "--yes", action="store_true", help="Skip file count confirmation")
     parser.add_argument(
@@ -140,6 +154,67 @@ def _pick(cli_value, config: dict, key: str, default=None):
     return default
 
 
+def _find_git_root(start: Path) -> tuple[Path | None, Path | None]:
+    for p in [start, *start.parents]:
+        git_path = p / ".git"
+        if git_path.is_dir():
+            return p, git_path
+        if git_path.is_file():
+            try:
+                raw = git_path.read_text(encoding="utf-8", errors="ignore").strip()
+            except Exception:
+                raw = ""
+            if raw.startswith("gitdir:"):
+                git_dir = raw.split(":", 1)[1].strip()
+                git_dir_path = Path(git_dir)
+                if not git_dir_path.is_absolute():
+                    git_dir_path = (p / git_dir_path).resolve()
+                return p, git_dir_path
+            return p, None
+    return None, None
+
+
+def _global_ignore_paths(cwd: Path | None = None) -> list[Path]:
+    paths: list[Path] = []
+    cwd = cwd or Path.cwd()
+    if shutil.which("git"):
+        try:
+            result = subprocess.run(
+                ["git", "config", "--get", "--path", "core.excludesfile"],
+                cwd=cwd,
+                capture_output=True,
+                text=True,
+                check=False,
+            )
+            value = (result.stdout or "").strip()
+        except Exception:
+            value = ""
+        if value:
+            candidate = Path(value).expanduser()
+            if candidate.exists():
+                paths.append(candidate)
+
+    xdg_config = os.getenv("XDG_CONFIG_HOME")
+    if xdg_config:
+        default_path = Path(xdg_config) / "git" / "ignore"
+    else:
+        default_path = Path.home() / ".config" / "git" / "ignore"
+    if default_path.exists():
+        paths.append(default_path)
+
+    legacy = Path.home() / ".gitignore_global"
+    if legacy.exists():
+        paths.append(legacy)
+
+    seen: set[Path] = set()
+    unique: list[Path] = []
+    for p in paths:
+        if p not in seen:
+            seen.add(p)
+            unique.append(p)
+    return unique
+
+
 def _env_value(name: str) -> str | None:
     val = os.getenv(name)
     if val is None:
@@ -425,12 +500,26 @@ def main(argv: list[str] | None = None) -> int:
         if hard_max is not None and hard_max <= 0:
             hard_max = None
 
+        ignore_spec = None
+        ignore_root = None
+        if not args.no_ignore:
+            git_root, git_dir = _find_git_root(cwd)
+            ignore_root = git_root or cwd
+            extra_ignores: list[Path] = []
+            if git_dir is not None:
+                extra_ignores.append(git_dir / "info" / "exclude")
+            extra_ignores.extend(_global_ignore_paths(ignore_root))
+            ignore_spec = build_ignore_spec(ignore_root, extra_paths=extra_ignores)
+
         candidates = collect_candidates(
             input_paths,
             cwd=cwd,
             recursive=args.recursive,
             include_globs=globs,
             type_exts=type_exts,
+            include_hidden=args.hidden,
+            ignore_spec=ignore_spec,
+            ignore_root=ignore_root,
         )
         candidate_count = len(candidates)
         if hard_max is not None and candidate_count > hard_max:
@@ -565,7 +654,6 @@ def main(argv: list[str] | None = None) -> int:
     output_lines = render_matches(
         files=files,
         matches=verified,
-        show_line_numbers=args.line_numbers,
         before=before,
         after=after,
         use_color=use_color,

{rlmgrep-0.1.8 → rlmgrep-0.1.17}/rlmgrep/ingest.py

@@ -2,8 +2,11 @@ from __future__ import annotations
 
 from dataclasses import dataclass
 from fnmatch import fnmatch
+import os
 from pathlib import Path, PurePosixPath
-from typing import Iterable, Any, Callable
+from typing import Any, Callable, Iterable
+
+import pathspec
 
 from pypdf import PdfReader
 
@@ -161,6 +164,97 @@ def collect_files(paths: Iterable[str], recursive: bool = True) -> list[Path]:
     return files
 
 
+IGNORE_FILENAMES = {".gitignore", ".ignore", ".rgignore"}
+
+
+def build_ignore_spec(
+    root: Path, extra_paths: Iterable[Path] | None = None
+) -> "pathspec.PathSpec | None":
+    root = root.resolve()
+    ignore_paths: list[Path] = []
+    extra_paths = list(extra_paths or [])
+
+    for dirpath, dirnames, filenames in os.walk(root):
+        if ".git" in dirnames:
+            dirnames.remove(".git")
+        for name in filenames:
+            if name in IGNORE_FILENAMES:
+                ignore_paths.append(Path(dirpath) / name)
+
+    for extra in extra_paths:
+        if extra.exists():
+            ignore_paths.append(extra)
+
+    if not ignore_paths:
+        return None
+
+    def _sort_key(p: Path) -> tuple[int, str]:
+        try:
+            rel = p.parent.relative_to(root)
+            depth = len(rel.parts)
+            return depth, rel.as_posix()
+        except ValueError:
+            return 0, p.as_posix()
+
+    ignore_paths.sort(key=_sort_key)
+
+    patterns: list[str] = []
+    for gi in ignore_paths:
+        try:
+            rel_dir = gi.parent.relative_to(root).as_posix()
+        except ValueError:
+            rel_dir = ""
+        if rel_dir in {".", ""}:
+            rel_dir = ""
+        try:
+            raw_lines = gi.read_text(encoding="utf-8", errors="ignore").splitlines()
+        except Exception:
+            continue
+        for raw in raw_lines:
+            line = raw.rstrip("\n")
+            if not line:
+                continue
+            escaped = False
+            if line.startswith("\\#") or line.startswith("\\!"):
+                line = line[1:]
+                escaped = True
+            if not escaped and line.startswith("#"):
+                continue
+
+            negated = False
+            if not escaped and line.startswith("!"):
+                negated = True
+                line = line[1:]
+            if not line:
+                continue
+
+            anchored = False
+            if line.startswith("/"):
+                anchored = True
+                line = line[1:]
+            if not line:
+                continue
+
+            if rel_dir:
+                if anchored:
+                    line = f"{rel_dir}/{line}"
+                elif "/" in line:
+                    line = f"{rel_dir}/{line}"
+                else:
+                    line = f"{rel_dir}/**/{line}"
+            else:
+                if anchored:
+                    line = f"/{line}"
+
+            if negated:
+                line = "!" + line
+            patterns.append(line)
+
+    if not patterns:
+        return None
+    return pathspec.PathSpec.from_lines("gitwildmatch", patterns)
+
+
 TYPE_EXTS = {
     "bash": {".bash"},
     "c": {".c", ".h"},
@@ -237,21 +331,46 @@ def _matches_globs(path: str, globs: list[str]) -> bool:
     return False
 
 
+def _is_hidden_path(path: Path) -> bool:
+    return any(part.startswith(".") for part in path.parts if part)
+
+
 def collect_candidates(
     paths: Iterable[str],
     cwd: Path,
     recursive: bool = True,
     include_globs: list[str] | None = None,
     type_exts: set[str] | None = None,
+    include_hidden: bool = False,
+    ignore_spec: "pathspec.PathSpec | None" = None,
+    ignore_root: Path | None = None,
 ) -> list[Path]:
     files = collect_files(paths, recursive=recursive)
+    explicit_files: set[Path] = set()
+    for raw in paths:
+        p = Path(raw)
+        if p.exists() and p.is_file():
+            explicit_files.add(p.resolve())
     candidates: list[Path] = []
     for fp in files:
+        fp_resolved = fp.resolve()
+        is_explicit = fp_resolved in explicit_files
+        if not include_hidden and not is_explicit and _is_hidden_path(fp):
+            continue
+
         try:
             key = fp.relative_to(cwd).as_posix()
         except ValueError:
             key = fp.as_posix()
 
+        if ignore_spec is not None and ignore_root is not None and not is_explicit:
+            try:
+                rel = fp.relative_to(ignore_root).as_posix()
+            except ValueError:
+                rel = None
+            if rel and ignore_spec.match_file(rel):
+                continue
+
         if include_globs and not _matches_globs(key, include_globs):
             continue
 

{rlmgrep-0.1.8 → rlmgrep-0.1.17}/rlmgrep/render.py

@@ -23,13 +23,10 @@ def _format_line(
     line_no: int,
     text: str,
     is_match: bool,
-    show_line_numbers: bool,
     use_color: bool,
     heading: bool,
 ) -> str:
     delim = ":" if is_match else "-"
-    if not show_line_numbers:
-        return text
     prefix = _colorize(str(line_no), COLOR_LINE_NO, use_color)
     sep = "\t" if heading else ""
     return f"{prefix}{delim}{sep}{text}"
@@ -52,7 +49,6 @@ def _merge_ranges(ranges: list[tuple[int, int]]) -> list[tuple[int, int]]:
 def render_matches(
     files: dict[str, FileRecord],
     matches: dict[str, list[int]],
-    show_line_numbers: bool,
     before: int,
     after: int,
     use_color: bool = False,
@@ -86,7 +82,6 @@ def render_matches(
                             line_no,
                             text,
                             True,
-                            show_line_numbers,
                             use_color,
                             heading,
                         )
@@ -111,7 +106,6 @@ def render_matches(
                         line_no,
                         text,
                         is_match,
-                        show_line_numbers,
                         use_color,
                         heading,
                     )

{rlmgrep-0.1.8 → rlmgrep-0.1.17}/rlmgrep.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: rlmgrep
-Version: 0.1.8
+Version: 0.1.17
 Summary: Grep-shaped CLI search powered by DSPy RLM
 Author: rlmgrep
 License: MIT
@@ -8,11 +8,12 @@ Requires-Python: >=3.11
 Description-Content-Type: text/markdown
 Requires-Dist: dspy>=3.1.1
 Requires-Dist: markitdown[all]>=0.1.4
+Requires-Dist: pathspec>=0.12.1
 Requires-Dist: pypdf>=4.0.0
 
 # rlmgrep
 
-Grep-shaped search powered by DSPy RLM. It accepts a natural-language query, scans the files you point at, and prints matching lines in a grep-like format.
+Grep-shaped search powered by DSPy RLM. It accepts a natural-language query, scans the files you point at, and prints matching lines in a grep-like format. Use `--answer` to get a narrative response grounded in the selected files/directories.
 
 ## Quickstart
 
@@ -22,9 +23,20 @@ uv tool install rlmgrep
 # uv tool install git+https://github.com/halfprice06/rlmgrep.git
 
 export OPENAI_API_KEY=...  # or set keys in ~/.rlmgrep
-rlmgrep "where are API keys read" rlmgrep/
 ```
 
+```sh
+rlmgrep --answer "What does this repo do and where are the entry points?" .
+```
+
+![Quickstart answer mode](docs/images/quickstart-answer.png)
+
+```sh
+rlmgrep -C 2 "Where is retry/backoff configured and what are the defaults?" .
+```
+
+![Quickstart context mode](docs/images/quickstart-context.png)
+
 ## Requirements
 
 - Python 3.11+
@@ -38,8 +50,8 @@ One of rlmgrep’s most useful features is that it can “grep” **PDFs and Off
 How it works:
 - **PDFs** are parsed with `pypdf`. Each page gets a marker line like `===== Page N =====`, and output lines include a `page=N` suffix. Line numbers refer to the extracted text (not PDF coordinates).
 - **Office & binary docs** (`.docx`, `.pptx`, `.xlsx`, `.html`, `.zip`, etc.) are converted to Markdown via **MarkItDown**. This happens during ingestion, so rlmgrep can search them like any other text file.
-- **Images** can be described by a vision model through MarkItDown (OpenAI/Anthropic/Gemini).
-- **Audio** transcription is supported through OpenAI when enabled.
+- **Images** can be described by a vision model and then searched through MarkItDown (OpenAI/Anthropic/Gemini), enable and configure in config.toml.
+- **Audio** transcription is supported through OpenAI when enabled, configure in config.toml.
 
 Sidecar caching:
 - For images/audio, converted text is cached next to the original file as `<original>.<ext>.md` and reused on later runs.
@@ -47,7 +59,7 @@ Sidecar caching:
 
 ## Install Deno
 
-DSPy requires the Deno runtime. Install it with the official scripts:
+DSPy's default implementation of RLM requires the Deno runtime. Install it with the official scripts:
 
 macOS/Linux:
 
@@ -75,12 +87,15 @@ rlmgrep [options] "query" [paths...]
 
 Common options:
 
+- `--answer` return a narrative answer before the grep output
 - `-C N` context lines before/after (grep-style)
 - `-A N` context lines after
 - `-B N` context lines before
 - `-m N` max matching lines per file
 - `-g GLOB` include files matching glob (repeatable, comma-separated)
 - `--type T` include file types (repeatable, comma-separated)
+- `--hidden` include hidden files and directories
+- `--no-ignore` do not respect `.gitignore`
 - `--no-recursive` do not recurse directories
 - `-a`, `--text` treat binary files as text
 - `-y`, `--yes` skip file count confirmation
@@ -95,7 +110,7 @@ Examples:
 
 ```sh
 # Natural-language query over a repo
-rlmgrep -n -C 2 "Where is retry/backoff configured and what are the defaults?" .
+rlmgrep -C 2 "Where is retry/backoff configured and what are the defaults?" .
 
 # Restrict to Python files
 rlmgrep "Where do we parse JWTs and enforce expiration?" --type py .
@@ -113,6 +128,7 @@ rg -l "token" . | rlmgrep --files-from-stdin --answer "What does this token cont
 ## Input selection
 
 - Directories are searched recursively by default. Use `--no-recursive` to stop recursion.
+- Hidden files and ignore files (`.gitignore`, `.ignore`, `.rgignore`) are respected by default. Use `--hidden` or `--no-ignore` to include them.
 - `--type` uses built-in type mappings (e.g., `py`, `js`, `md`); unknown values are treated as file extensions.
 - `-g/--glob` matches path globs against normalized paths (forward slashes).
 - Paths are printed relative to the current working directory when possible.
@@ -125,7 +141,7 @@ rg -l "token" . | rlmgrep --files-from-stdin --answer "What does this token cont
 - Output uses rg-style headings by default:
   - A file header line like `./path/to/file`
   - Then `line:\ttext` for matches, `line-\ttext` for context lines
-- Line numbers are 1-based.
+- Line numbers are always included and are 1-based.
 - When context ranges are disjoint, a `--` line separates groups.
 - Exit codes:
   - `0` = at least one match
@@ -140,7 +156,7 @@ rlmgrep can interpret traditional regex-style patterns inside a natural-language
 Example (best-effort regex semantics + extra context):
 
 ```sh
-rlmgrep -n "Find Python functions that look like `def test_\\w+` and are marked as slow or flaky in nearby comments." .
+rlmgrep "Find Python functions that look like `def test_\\w+` and are marked as slow or flaky in nearby comments." .
 ```
 
 If you need strict, deterministic regex behavior, use `rg`/`grep`.

{rlmgrep-0.1.8 → rlmgrep-0.1.17}/rlmgrep.egg-info/requires.txt

@@ -1,3 +1,4 @@
 dspy>=3.1.1
 markitdown[all]>=0.1.4
+pathspec>=0.12.1
 pypdf>=4.0.0