git-commit-message 0.8.0__tar.gz → 0.8.2__tar.gz

{git_commit_message-0.8.0 → git_commit_message-0.8.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: git-commit-message
-Version: 0.8.0
+Version: 0.8.2
 Summary: Generate Git commit messages from staged changes using LLM
 Maintainer-email: Mina Her <minacle@live.com>
 License: This is free and unencumbered software released into the public domain.
@@ -51,14 +51,14 @@ Requires-Dist: tiktoken>=0.12.0
 
 # git-commit-message
 
-Generate a commit message from your staged changes using OpenAI, Google Gemini, or Ollama.
+Generate a commit message from your staged changes using OpenAI, Google Gemini, Ollama, or llama.cpp.
 
 [![asciicast](https://asciinema.org/a/jk0phFqNnc5vaCiIZEYBwZOyN.svg)](https://asciinema.org/a/jk0phFqNnc5vaCiIZEYBwZOyN)
 
 ## Requirements
 
 - Python 3.13+
-- A Git repo with staged changes (`git add ...`)
+- A Git repo with staged changes (`git add ...`) (or use `--amend` even if nothing is staged)
 
 ## Install
 
@@ -120,6 +120,23 @@ export GIT_COMMIT_MESSAGE_PROVIDER=ollama
 export OLLAMA_MODEL=mistral
 ```
 
+### llama.cpp (local models)
+
+1. Build and run llama.cpp server with your model:
+
+```sh
+llama-server -hf ggml-org/gpt-oss-20b-GGUF --host 0.0.0.0 --port 8080
+```
+
+2. The server runs on `http://localhost:8080` by default.
+
+Optional: set defaults:
+
+```sh
+export GIT_COMMIT_MESSAGE_PROVIDER=llamacpp
+export LLAMACPP_HOST=http://localhost:8080
+```
+
 Note (fish):
 
 ```fish
@@ -141,10 +158,13 @@ git add -A
 git-commit-message "optional extra context about the change"
 ```
 
-Generate a single-line subject only:
+Generate a single-line subject only (when no trailers are appended):
 
 ```sh
 git-commit-message --one-line "optional context"
+
+# with trailers, output is subject plus trailer lines
+git-commit-message --one-line --co-author 'John Doe <john.doe@example.com>'
 ```
 
 Select provider:
@@ -158,6 +178,9 @@ git-commit-message --provider google
 
 # Ollama
 git-commit-message --provider ollama
+
+# llama.cpp
+git-commit-message --provider llamacpp
 ```
 
 Commit immediately (optionally open editor):
@@ -165,6 +188,24 @@ Commit immediately (optionally open editor):
 ```sh
 git-commit-message --commit "refactor parser for speed"
 git-commit-message --commit --edit "refactor parser for speed"
+
+# add co-author trailers
+git-commit-message --commit --co-author 'John Doe <john.doe@example.com>'
+git-commit-message --commit --co-author 'John Doe <john.doe@example.com>' --co-author 'Jane Doe <jane.doe@example.com>'
+git-commit-message --commit --co-author copilot
+```
+
+Amend the previous commit:
+
+```sh
+# print only (useful for pasting into a GUI editor)
+git-commit-message --amend "optional context"
+
+# amend immediately
+git-commit-message --commit --amend "optional context"
+
+# amend immediately, but open editor for final tweaks
+git-commit-message --commit --amend --edit "optional context"
 ```
 
 Limit subject length:
@@ -206,18 +247,26 @@ Configure Ollama host (if running on a different machine):
 git-commit-message --provider ollama --host http://192.168.1.100:11434
 ```
 
+Configure llama.cpp host:
+
+```sh
+git-commit-message --provider llamacpp --host http://192.168.1.100:8080
+```
+
 ## Options
 
-- `--provider {openai,google,ollama}`: provider to use (default: `openai`)
-- `--model MODEL`: model override (provider-specific)
+- `--provider {openai,google,ollama,llamacpp}`: provider to use (default: `openai`)
+- `--model MODEL`: model override (provider-specific; ignored for llama.cpp)
 - `--language TAG`: output language/locale (default: `en-GB`)
-- `--one-line`: output subject only
+- `--one-line`: output subject only when no trailers are appended; with `--co-author`, output is a single-line subject plus `Co-authored-by:` trailer lines
 - `--max-length N`: max subject length (default: 72)
 - `--chunk-tokens N`: token budget per diff chunk (`0` = single summary pass, `-1` disables summarisation)
 - `--debug`: print request/response details
 - `--commit`: run `git commit -m <message>`
+- `--amend`: generate a message suitable for amending the previous commit (diff is from the amended commit's parent to the staged index; if nothing is staged, this effectively becomes the diff introduced by `HEAD`)
 - `--edit`: with `--commit`, open editor for final message
-- `--host URL`: host URL for providers like Ollama (default: `http://localhost:11434`)
+- `--host URL`: host URL for providers like Ollama or llama.cpp (default: `http://localhost:11434` for Ollama, `http://localhost:8080` for llama.cpp)
+- `--co-author VALUE`: append `Co-authored-by:` trailer(s). Repeat to add multiple values. Accepted forms: `Name <email@example.com>` or `copilot` (alias, case-insensitive).
 
 ## Environment variables
 
@@ -233,6 +282,7 @@ Optional:
 - `OPENAI_MODEL`: OpenAI-only model override (used if `--model`/`GIT_COMMIT_MESSAGE_MODEL` are not set)
 - `OLLAMA_MODEL`: Ollama-only model override (used if `--model`/`GIT_COMMIT_MESSAGE_MODEL` are not set)
 - `OLLAMA_HOST`: Ollama server URL (default: `http://localhost:11434`)
+- `LLAMACPP_HOST`: llama.cpp server URL (default: `http://localhost:8080`)
 - `GIT_COMMIT_MESSAGE_LANGUAGE`: default language/locale (default: `en-GB`)
 - `GIT_COMMIT_MESSAGE_CHUNK_TOKENS`: default chunk token budget (default: `0`)
 
@@ -241,6 +291,7 @@ Default models (if not overridden):
 - OpenAI: `gpt-5-mini`
 - Google: `gemini-2.5-flash`
 - Ollama: `gpt-oss:20b`
+- llama.cpp: uses pre-loaded model (model parameter is ignored)
 
 ## AI-generated code notice
 

{git_commit_message-0.8.0 → git_commit_message-0.8.2}/README.md

@@ -1,13 +1,13 @@
 # git-commit-message
 
-Generate a commit message from your staged changes using OpenAI, Google Gemini, or Ollama.
+Generate a commit message from your staged changes using OpenAI, Google Gemini, Ollama, or llama.cpp.
 
 [![asciicast](https://asciinema.org/a/jk0phFqNnc5vaCiIZEYBwZOyN.svg)](https://asciinema.org/a/jk0phFqNnc5vaCiIZEYBwZOyN)
 
 ## Requirements
 
 - Python 3.13+
-- A Git repo with staged changes (`git add ...`)
+- A Git repo with staged changes (`git add ...`) (or use `--amend` even if nothing is staged)
 
 ## Install
 
@@ -69,6 +69,23 @@ export GIT_COMMIT_MESSAGE_PROVIDER=ollama
 export OLLAMA_MODEL=mistral
 ```
 
+### llama.cpp (local models)
+
+1. Build and run llama.cpp server with your model:
+
+```sh
+llama-server -hf ggml-org/gpt-oss-20b-GGUF --host 0.0.0.0 --port 8080
+```
+
+2. The server runs on `http://localhost:8080` by default.
+
+Optional: set defaults:
+
+```sh
+export GIT_COMMIT_MESSAGE_PROVIDER=llamacpp
+export LLAMACPP_HOST=http://localhost:8080
+```
+
 Note (fish):
 
 ```fish
@@ -90,10 +107,13 @@ git add -A
 git-commit-message "optional extra context about the change"
 ```
 
-Generate a single-line subject only:
+Generate a single-line subject only (when no trailers are appended):
 
 ```sh
 git-commit-message --one-line "optional context"
+
+# with trailers, output is subject plus trailer lines
+git-commit-message --one-line --co-author 'John Doe <john.doe@example.com>'
 ```
 
 Select provider:
@@ -107,6 +127,9 @@ git-commit-message --provider google
 
 # Ollama
 git-commit-message --provider ollama
+
+# llama.cpp
+git-commit-message --provider llamacpp
 ```
 
 Commit immediately (optionally open editor):
@@ -114,6 +137,24 @@ Commit immediately (optionally open editor):
 ```sh
 git-commit-message --commit "refactor parser for speed"
 git-commit-message --commit --edit "refactor parser for speed"
+
+# add co-author trailers
+git-commit-message --commit --co-author 'John Doe <john.doe@example.com>'
+git-commit-message --commit --co-author 'John Doe <john.doe@example.com>' --co-author 'Jane Doe <jane.doe@example.com>'
+git-commit-message --commit --co-author copilot
+```
+
+Amend the previous commit:
+
+```sh
+# print only (useful for pasting into a GUI editor)
+git-commit-message --amend "optional context"
+
+# amend immediately
+git-commit-message --commit --amend "optional context"
+
+# amend immediately, but open editor for final tweaks
+git-commit-message --commit --amend --edit "optional context"
 ```
 
 Limit subject length:
@@ -155,18 +196,26 @@ Configure Ollama host (if running on a different machine):
 git-commit-message --provider ollama --host http://192.168.1.100:11434
 ```
 
+Configure llama.cpp host:
+
+```sh
+git-commit-message --provider llamacpp --host http://192.168.1.100:8080
+```
+
 ## Options
 
-- `--provider {openai,google,ollama}`: provider to use (default: `openai`)
-- `--model MODEL`: model override (provider-specific)
+- `--provider {openai,google,ollama,llamacpp}`: provider to use (default: `openai`)
+- `--model MODEL`: model override (provider-specific; ignored for llama.cpp)
 - `--language TAG`: output language/locale (default: `en-GB`)
-- `--one-line`: output subject only
+- `--one-line`: output subject only when no trailers are appended; with `--co-author`, output is a single-line subject plus `Co-authored-by:` trailer lines
 - `--max-length N`: max subject length (default: 72)
 - `--chunk-tokens N`: token budget per diff chunk (`0` = single summary pass, `-1` disables summarisation)
 - `--debug`: print request/response details
 - `--commit`: run `git commit -m <message>`
+- `--amend`: generate a message suitable for amending the previous commit (diff is from the amended commit's parent to the staged index; if nothing is staged, this effectively becomes the diff introduced by `HEAD`)
 - `--edit`: with `--commit`, open editor for final message
-- `--host URL`: host URL for providers like Ollama (default: `http://localhost:11434`)
+- `--host URL`: host URL for providers like Ollama or llama.cpp (default: `http://localhost:11434` for Ollama, `http://localhost:8080` for llama.cpp)
+- `--co-author VALUE`: append `Co-authored-by:` trailer(s). Repeat to add multiple values. Accepted forms: `Name <email@example.com>` or `copilot` (alias, case-insensitive).
 
 ## Environment variables
 
@@ -182,6 +231,7 @@ Optional:
 - `OPENAI_MODEL`: OpenAI-only model override (used if `--model`/`GIT_COMMIT_MESSAGE_MODEL` are not set)
 - `OLLAMA_MODEL`: Ollama-only model override (used if `--model`/`GIT_COMMIT_MESSAGE_MODEL` are not set)
 - `OLLAMA_HOST`: Ollama server URL (default: `http://localhost:11434`)
+- `LLAMACPP_HOST`: llama.cpp server URL (default: `http://localhost:8080`)
 - `GIT_COMMIT_MESSAGE_LANGUAGE`: default language/locale (default: `en-GB`)
 - `GIT_COMMIT_MESSAGE_CHUNK_TOKENS`: default chunk token budget (default: `0`)
 
@@ -190,6 +240,7 @@ Default models (if not overridden):
 - OpenAI: `gpt-5-mini`
 - Google: `gemini-2.5-flash`
 - Ollama: `gpt-oss:20b`
+- llama.cpp: uses pre-loaded model (model parameter is ignored)
 
 ## AI-generated code notice
 

{git_commit_message-0.8.0 → git_commit_message-0.8.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "git-commit-message"
-version = "0.8.0"
+version = "0.8.2"
 description = "Generate Git commit messages from staged changes using LLM"
 readme = "README.md"
 requires-python = ">=3.13"

{git_commit_message-0.8.0 → git_commit_message-0.8.2}/src/git_commit_message/_cli.py

@@ -9,6 +9,8 @@ from __future__ import annotations
 from argparse import ArgumentParser, Namespace
 from os import environ
 from pathlib import Path
+import re
+from re import Pattern
 from sys import exit as sys_exit
 from sys import stderr
 from typing import Final
@@ -17,7 +19,9 @@ from ._git import (
     commit_with_message,
     get_repo_root,
     get_staged_diff,
+    has_head_commit,
     has_staged_changes,
+    resolve_amend_base_ref,
 )
 from ._llm import (
     CommitMessageResult,
@@ -31,6 +35,7 @@ class CliArgs(Namespace):
     __slots__ = (
         "description",
         "commit",
+        "amend",
         "edit",
         "provider",
         "model",
@@ -40,6 +45,7 @@ class CliArgs(Namespace):
         "max_length",
         "chunk_tokens",
         "host",
+        "co_authors",
     )
 
     def __init__(
@@ -48,6 +54,7 @@ class CliArgs(Namespace):
     ) -> None:
         self.description: str | None = None
         self.commit: bool = False
+        self.amend: bool = False
         self.edit: bool = False
         self.provider: str | None = None
         self.model: str | None = None
@@ -57,6 +64,83 @@ class CliArgs(Namespace):
         self.max_length: int | None = None
         self.chunk_tokens: int | None = None
         self.host: str | None = None
+        self.co_authors: list[str] | None = None
+
+
+_CO_AUTHOR_LINE_RE: Final[Pattern[str]] = re.compile(
+    r"^\s*([^<>\s\n][^<>\n]*?)\s*<([^<>\s\n]+@[^<>\s\n]+)>\s*$"
+)
+_CO_AUTHOR_ALIASES: Final[dict[str, str]] = {
+    "copilot": "Copilot <copilot@github.com>",
+}
+
+
+def _co_author_alias_keywords_text() -> str:
+    """Return a readable list of accepted co-author alias keywords."""
+
+    keywords: list[str] = sorted(_CO_AUTHOR_ALIASES.keys())
+    return ", ".join(f"'{keyword}'" for keyword in keywords)
+
+
+def _normalize_co_author(
+    raw: str,
+    /,
+) -> str:
+    """Normalize one co-author input into ``Name <email>`` form."""
+
+    value: str = raw.strip()
+    if not value:
+        raise ValueError("Co-author cannot be empty.")
+
+    alias: str | None = _CO_AUTHOR_ALIASES.get(value.lower())
+    if alias is not None:
+        return alias
+
+    match = _CO_AUTHOR_LINE_RE.match(value)
+    if match is None:
+        raise ValueError(
+            "Invalid co-author format: use 'Name <email@example.com>' "
+            f"or an alias keyword ({_co_author_alias_keywords_text()})."
+        )
+
+    name: str = match.group(1).strip()
+    email: str = match.group(2).strip()
+    return f"{name} <{email}>"
+
+
+def _append_co_author_footers(
+    message: str,
+    normalized_co_authors: list[str],
+    /,
+) -> str:
+    """Append Git co-author trailers to a commit message."""
+
+    if not normalized_co_authors:
+        return message
+
+    base: str = message.rstrip()
+    footer_lines: list[str] = [
+        f"Co-authored-by: {author}" for author in normalized_co_authors
+    ]
+    return f"{base}\n\n" + "\n".join(footer_lines)
+
+
+def _normalize_co_authors(
+    co_authors: list[str],
+    /,
+) -> list[str]:
+    """Normalize and deduplicate co-author values in insertion order."""
+
+    seen: set[str] = set()
+    normalized: list[str] = []
+    for raw in co_authors:
+        author = _normalize_co_author(raw)
+        key = author.lower()
+        if key in seen:
+            continue
+        seen.add(key)
+        normalized.append(author)
+    return normalized
 
 
 def _env_chunk_tokens_default() -> int | None:
@@ -99,6 +183,16 @@ def _build_parser() -> ArgumentParser:
         help="Commit immediately with the generated message.",
     )
 
+    parser.add_argument(
+        "--amend",
+        action="store_true",
+        help=(
+            "Generate a message suitable for amending the previous commit. "
+            "When set, the diff is computed from the amended commit's parent to the staged index. "
+            "Use with '--commit' to run the amend, or omit '--commit' to print the message only."
+        ),
+    )
+
     parser.add_argument(
         "--edit",
         action="store_true",
@@ -111,7 +205,8 @@ def _build_parser() -> ArgumentParser:
         help=(
             "LLM provider to use (default: openai). "
             "You may also set GIT_COMMIT_MESSAGE_PROVIDER. "
-            "The CLI flag overrides the environment variable."
+            "The CLI flag overrides the environment variable. "
+            "Supported providers: openai, google, ollama, llamacpp."
         ),
     )
 
@@ -119,7 +214,8 @@ def _build_parser() -> ArgumentParser:
         "--model",
         default=None,
         help=(
-            "Model name to use. If unspecified, uses GIT_COMMIT_MESSAGE_MODEL or a provider-specific default (openai: gpt-5-mini; google: gemini-2.5-flash; ollama: gpt-oss:20b)."
+            "Model name to use. If unspecified, uses GIT_COMMIT_MESSAGE_MODEL or a provider-specific default "
+            "(openai: gpt-5-mini; google: gemini-2.5-flash; ollama: gpt-oss:20b; llamacpp: default)."
         ),
     )
 
@@ -171,8 +267,24 @@ def _build_parser() -> ArgumentParser:
         dest="host",
         default=None,
         help=(
-            "Host URL for API providers like Ollama (default: http://localhost:11434). "
-            "You may also set OLLAMA_HOST for Ollama."
+            "Host URL for API providers like Ollama or llama.cpp "
+            "(default: http://localhost:11434 for Ollama, http://localhost:8080 for llama.cpp). "
+            "You may also set OLLAMA_HOST for Ollama or LLAMACPP_HOST for llama.cpp."
+        ),
+    )
+
+    parser.add_argument(
+        "--co-author",
+        dest="co_authors",
+        action="append",
+        default=None,
+        help=(
+            "Add Co-authored-by trailer(s) to the generated message. "
+            "Repeat for multiple co-authors. "
+            "Use 'Name <email@example.com>' or an alias keyword "
+            f"({_co_author_alias_keywords_text()}). "
+            "When used with --one-line, the subject line remains single-line and these "
+            "trailers are appended on separate lines (i.e., the overall output is multi-line)."
         ),
     )
 
@@ -198,11 +310,19 @@ def _run(
 
     repo_root: Path = get_repo_root()
 
-    if not has_staged_changes(repo_root):
-        print("No staged changes. Run 'git add' and try again.", file=stderr)
-        return 2
+    if args.amend:
+        if not has_head_commit(repo_root):
+            print("Cannot amend: the repository has no commits yet.", file=stderr)
+            return 2
+
+        base_ref = resolve_amend_base_ref(repo_root)
+        diff_text: str = get_staged_diff(repo_root, base_ref=base_ref)
+    else:
+        if not has_staged_changes(repo_root):
+            print("No staged changes. Run 'git add' and try again.", file=stderr)
+            return 2
 
-    diff_text: str = get_staged_diff(repo_root)
+        diff_text = get_staged_diff(repo_root)
 
     hint: str | None = args.description if isinstance(args.description, str) else None
 
@@ -212,6 +332,14 @@ def _run(
     if chunk_tokens is None:
         chunk_tokens = 0
 
+    normalized_co_authors: list[str] | None = None
+    if args.co_authors:
+        try:
+            normalized_co_authors = _normalize_co_authors(args.co_authors)
+        except ValueError as exc:
+            print(str(exc), file=stderr)
+            return 2
+
     result: CommitMessageResult | None = None
     try:
         if args.debug:
@@ -256,6 +384,14 @@ def _run(
         else:
             message = ""
 
+    # Defensive check: one-line normalization can result in an empty message.
+    if not message.strip():
+        print("Failed to generate commit message: generated message is empty.", file=stderr)
+        return 3
+
+    if normalized_co_authors:
+        message = _append_co_author_footers(message, normalized_co_authors)
+
     if not args.commit:
         if args.debug and result is not None:
             # Print debug information
@@ -299,9 +435,9 @@ def _run(
         print(message)
 
     if args.edit:
-        rc: int = commit_with_message(message, True, repo_root)
+        rc: int = commit_with_message(message, True, repo_root, amend=args.amend)
     else:
-        rc = commit_with_message(message, False, repo_root)
+        rc = commit_with_message(message, False, repo_root, amend=args.amend)
 
     return rc
 

git_commit_message-0.8.2/src/git_commit_message/_git.py

@@ -0,0 +1,269 @@
+"""Git-related helper functions.
+
+Provides repository root discovery, extraction of staged changes, and
+creating commits from a message.
+"""
+
+from __future__ import annotations
+
+from pathlib import Path
+from subprocess import CalledProcessError, check_call, check_output, run
+
+
+def _get_empty_tree_hash(
+    cwd: Path,
+    /,
+) -> str:
+    """Return the empty tree hash for this repository.
+
+    Parameters
+    ----------
+    cwd
+        Repository directory in which to run Git.
+
+    Notes
+    -----
+    Do not hard-code the SHA, because repositories may use different
+    hash algorithms (e.g. SHA-1 vs SHA-256). We ask Git to compute the
+    empty tree object ID for the current repo.
+
+    Returns
+    -------
+    str
+        The empty tree object ID for the current repository.
+    """
+
+    try:
+        completed = run(
+            [
+                "git",
+                "hash-object",
+                "-t",
+                "tree",
+                "--stdin",
+            ],
+            cwd=str(cwd),
+            check=True,
+            input=b"",
+            capture_output=True,
+        )
+    except CalledProcessError as exc:
+        stderr_text = (exc.stderr or b"").decode(errors="replace").strip()
+        suffix = f"\nGit stderr: {stderr_text}" if stderr_text else ""
+        raise RuntimeError(
+            f"Failed to compute empty tree hash (git exited with {exc.returncode}).{suffix}"
+        ) from exc
+    except OSError as exc:
+        raise RuntimeError(
+            f"Failed to run git to compute empty tree hash: {exc}"
+        ) from exc
+    oid = completed.stdout.decode().strip()
+    if not oid:
+        raise RuntimeError(
+            "Failed to compute empty tree hash: git returned an empty object ID."
+        )
+    return oid
+
+
+def get_repo_root(
+    cwd: Path | None = None,
+    /,
+) -> Path:
+    """Find the repository root from the current working directory.
+
+    Parameters
+    ----------
+    cwd
+        Starting directory for the search. Defaults to the current working directory.
+
+    Returns
+    -------
+    Path
+        The repository root path.
+    """
+
+    start: Path = cwd or Path.cwd()
+    try:
+        out: bytes = check_output(
+            [
+                "git",
+                "rev-parse",
+                "--show-toplevel",
+            ],
+            cwd=str(start),
+        )
+    except CalledProcessError as exc:  # noqa: TRY003
+        raise RuntimeError("Not a Git repository.") from exc
+
+    root = Path(out.decode().strip())
+    return root
+
+
+def has_staged_changes(
+    cwd: Path,
+    /,
+) -> bool:
+    """Check whether there are staged changes."""
+
+    try:
+        check_call(
+            ["git", "diff", "--cached", "--quiet", "--exit-code"],
+            cwd=str(cwd),
+        )
+        return False
+    except CalledProcessError:
+        return True
+
+
+def has_head_commit(
+    cwd: Path,
+    /,
+) -> bool:
+    """Return True if the repository has at least one commit (HEAD exists).
+
+    Parameters
+    ----------
+    cwd
+        Repository directory in which to run Git.
+
+    Returns
+    -------
+    bool
+        True if ``HEAD`` exists in the repository, False otherwise.
+    """
+
+    completed = run(
+        ["git", "rev-parse", "--verify", "HEAD"],
+        cwd=str(cwd),
+        check=False,
+        capture_output=True,
+    )
+    return completed.returncode == 0
+
+
+def resolve_amend_base_ref(
+    cwd: Path,
+    /,
+) -> str:
+    """Resolve the base ref for an amend diff.
+
+    Parameters
+    ----------
+    cwd
+        Repository directory in which to run Git.
+
+    Notes
+    -----
+    The amended commit keeps the same parent as the current HEAD commit.
+
+    - If HEAD has a parent, base is ``HEAD^``.
+    - If HEAD is a root commit (no parent), base is the empty tree.
+
+    Returns
+    -------
+    str
+        The base reference for the amend diff: either ``HEAD^`` (when the
+        current ``HEAD`` commit has a parent) or the empty tree object ID
+        (when ``HEAD`` is a root commit).
+    """
+
+    completed = run(
+        ["git", "rev-parse", "--verify", "HEAD^"],
+        cwd=str(cwd),
+        check=False,
+        capture_output=True,
+    )
+    if completed.returncode == 0:
+        return "HEAD^"
+    return _get_empty_tree_hash(cwd)
+
+
+def get_staged_diff(
+    cwd: Path,
+    /,
+    *,
+    base_ref: str | None = None,
+) -> str:
+    """Return the staged changes as diff text.
+
+    Parameters
+    ----------
+    cwd
+        Git working directory.
+    base_ref
+        Optional Git reference or tree object ID (e.g., branch name, tag,
+        commit hash, or the empty tree hash) to diff against. When provided,
+        the diff shows changes from ``base_ref`` to the staged index, instead
+        of changes from ``HEAD`` to the staged index.
+
+    Returns
+    -------
+    str
+        Unified diff text for the staged changes.
+    """
+
+    cmd: list[str] = [
+        "git",
+        "diff",
+        "--cached",
+        "--patch",
+        "--minimal",
+        "--no-color",
+    ]
+    if base_ref:
+        cmd.append(base_ref)
+
+    try:
+        out: bytes = check_output(cmd, cwd=str(cwd))
+    except CalledProcessError as exc:
+        message = "Failed to retrieve staged diff from Git."
+        if base_ref:
+            message += (
+                " Ensure that the provided base_ref exists and is a valid Git reference."
+            )
+        raise RuntimeError(message) from exc
+
+    return out.decode()
+
+
+def commit_with_message(
+    message: str,
+    edit: bool,
+    cwd: Path,
+    /,
+    *,
+    amend: bool = False,
+) -> int:
+    """Create a commit with the given message.
+
+    Parameters
+    ----------
+    message
+        Commit message.
+    edit
+        If True, use the `--edit` flag to open an editor for amendments.
+    cwd
+        Git working directory.
+    amend
+        If True, pass ``--amend`` to Git to amend the current ``HEAD`` commit
+        instead of creating a new commit.
+
+    Returns
+    -------
+    int
+        The subprocess exit code.
+    """
+
+    cmd: list[str] = ["git", "commit"]
+    if amend:
+        cmd.append("--amend")
+
+    cmd.extend(["-m", message])
+    if edit:
+        cmd.append("--edit")
+
+    try:
+        completed = run(cmd, cwd=str(cwd), check=False)
+        return int(completed.returncode)
+    except OSError as exc:  # e.g., editor launch failure, etc.
+        raise RuntimeError(f"Failed to run 'git commit': {exc}") from exc

git_commit_message-0.8.2/src/git_commit_message/_llamacpp.py

@@ -0,0 +1,145 @@
+"""llama.cpp provider implementation.
+
+This module contains llama.cpp server-specific API calls and token counting.
+llama.cpp server provides an OpenAI-compatible API, so we use the openai library.
+Provider-agnostic orchestration/prompt logic lives in `_llm.py`.
+"""
+
+from __future__ import annotations
+
+from os import environ
+from typing import ClassVar, Final
+
+from openai import OpenAI
+from openai.types.chat import ChatCompletionMessageParam
+from tiktoken import Encoding, get_encoding
+
+from ._llm import LLMTextResult, LLMUsage
+
+
+_DEFAULT_LLAMACPP_HOST: Final[str] = "http://localhost:8080"
+
+
+def _resolve_llamacpp_host(
+    host: str | None,
+    /,
+) -> str:
+    """Resolve the llama.cpp server host URL from arg, env, or default."""
+
+    return host or environ.get("LLAMACPP_HOST") or _DEFAULT_LLAMACPP_HOST
+
+
+def _get_encoding() -> Encoding:
+    """Get a fallback encoding for token counting."""
+
+    try:
+        return get_encoding("cl100k_base")
+    except Exception:
+        return get_encoding("gpt2")
+
+
+class LlamaCppProvider:
+    """llama.cpp provider implementation for the LLM protocol.
+
+    Uses the OpenAI-compatible API provided by llama.cpp server.
+    """
+
+    __slots__ = (
+        "_host",
+        "_client",
+    )
+
+    name: ClassVar[str] = "llamacpp"
+
+    def __init__(
+        self,
+        /,
+        *,
+        host: str | None = None,
+    ) -> None:
+        self._host = _resolve_llamacpp_host(host)
+        # llama.cpp server uses OpenAI-compatible API
+        # api_key is not required but openai library needs a placeholder
+        self._client = OpenAI(
+            base_url=f"{self._host}/v1",
+            api_key="llamacpp",  # Placeholder, llama.cpp doesn't require auth by default
+        )
+
+    def generate_text(
+        self,
+        /,
+        *,
+        model: str,
+        instructions: str,
+        user_text: str,
+    ) -> LLMTextResult:
+        """Generate text using llama.cpp server (OpenAI-compatible chat/completions API)."""
+
+        messages: list[ChatCompletionMessageParam] = [
+            {"role": "system", "content": instructions},
+            {"role": "user", "content": user_text},
+        ]
+
+        try:
+            resp = self._client.chat.completions.create(
+                model=model,
+                messages=messages,
+            )
+        except Exception as exc:
+            raise RuntimeError(
+                f"Failed to connect to llama.cpp server at {self._host}. "
+                f"Make sure llama.cpp server is running: {exc}"
+            ) from exc
+
+        text: str = ""
+        if resp.choices and len(resp.choices) > 0:
+            choice = resp.choices[0]
+            if choice.message and choice.message.content:
+                text = choice.message.content.strip()
+
+        if not text:
+            raise RuntimeError("An empty response text was generated by the provider.")
+
+        usage: LLMUsage | None = None
+        if resp.usage is not None:
+            usage = LLMUsage(
+                prompt_tokens=resp.usage.prompt_tokens,
+                completion_tokens=resp.usage.completion_tokens,
+                total_tokens=resp.usage.total_tokens,
+            )
+
+        return LLMTextResult(
+            text=text,
+            response_id=resp.id,
+            usage=usage,
+        )
+
+    def count_tokens(
+        self,
+        /,
+        *,
+        model: str,
+        text: str,
+    ) -> int:
+        """Count tokens using llama.cpp's official token counting API."""
+
+        try:
+            # Use llama.cpp's official token counting endpoint via OpenAI client's internal HTTP client
+            response = self._client.post(
+                "/messages/count_tokens",
+                body={
+                    "model": model,
+                    "messages": [
+                        {"role": "user", "content": text}
+                    ]
+                },
+                cast_to=dict,
+            )
+            return response.get("total", 0)
+        except Exception:
+            # Fallback to tiktoken approximation
+            try:
+                encoding = _get_encoding()
+                return len(encoding.encode(text))
+            except Exception:
+                return len(text.split())

{git_commit_message-0.8.0 → git_commit_message-0.8.2}/src/git_commit_message/_llm.py

@@ -19,6 +19,7 @@ _DEFAULT_PROVIDER: Final[str] = "openai"
 _DEFAULT_MODEL_OPENAI: Final[str] = "gpt-5-mini"
 _DEFAULT_MODEL_GOOGLE: Final[str] = "gemini-2.5-flash"
 _DEFAULT_MODEL_OLLAMA: Final[str] = "gpt-oss:20b"
+_DEFAULT_MODEL_LLAMACPP: Final[str] = "default"
 _DEFAULT_LANGUAGE: Final[str] = "en-GB"
 
 
@@ -155,6 +156,9 @@ def _resolve_model(
     elif provider_name == "ollama":
         default_model = _DEFAULT_MODEL_OLLAMA
         provider_model = environ.get("OLLAMA_MODEL")
+    elif provider_name == "llamacpp":
+        default_model = _DEFAULT_MODEL_LLAMACPP
+        provider_model = environ.get("LLAMACPP_MODEL")
     else:
         default_model = _DEFAULT_MODEL_OPENAI
         provider_model = environ.get("OPENAI_MODEL")
@@ -195,8 +199,14 @@ def get_provider(
 
         return OllamaProvider(host=host)
 
+    if name == "llamacpp":
+        # Local import to avoid import cycles: providers may import shared types from this module.
+        from ._llamacpp import LlamaCppProvider
+
+        return LlamaCppProvider(host=host)
+
     raise UnsupportedProviderError(
-        f"Unsupported provider: {name}. Supported providers: openai, google, ollama"
+        f"Unsupported provider: {name}. Supported providers: openai, google, ollama, llamacpp"
     )
 
 

{git_commit_message-0.8.0 → git_commit_message-0.8.2}/src/git_commit_message.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: git-commit-message
-Version: 0.8.0
+Version: 0.8.2
 Summary: Generate Git commit messages from staged changes using LLM
 Maintainer-email: Mina Her <minacle@live.com>
 License: This is free and unencumbered software released into the public domain.
@@ -51,14 +51,14 @@ Requires-Dist: tiktoken>=0.12.0
 
 # git-commit-message
 
-Generate a commit message from your staged changes using OpenAI, Google Gemini, or Ollama.
+Generate a commit message from your staged changes using OpenAI, Google Gemini, Ollama, or llama.cpp.
 
 [![asciicast](https://asciinema.org/a/jk0phFqNnc5vaCiIZEYBwZOyN.svg)](https://asciinema.org/a/jk0phFqNnc5vaCiIZEYBwZOyN)
 
 ## Requirements
 
 - Python 3.13+
-- A Git repo with staged changes (`git add ...`)
+- A Git repo with staged changes (`git add ...`) (or use `--amend` even if nothing is staged)
 
 ## Install
 
@@ -120,6 +120,23 @@ export GIT_COMMIT_MESSAGE_PROVIDER=ollama
 export OLLAMA_MODEL=mistral
 ```
 
+### llama.cpp (local models)
+
+1. Build and run llama.cpp server with your model:
+
+```sh
+llama-server -hf ggml-org/gpt-oss-20b-GGUF --host 0.0.0.0 --port 8080
+```
+
+2. The server runs on `http://localhost:8080` by default.
+
+Optional: set defaults:
+
+```sh
+export GIT_COMMIT_MESSAGE_PROVIDER=llamacpp
+export LLAMACPP_HOST=http://localhost:8080
+```
+
 Note (fish):
 
 ```fish
@@ -141,10 +158,13 @@ git add -A
 git-commit-message "optional extra context about the change"
 ```
 
-Generate a single-line subject only:
+Generate a single-line subject only (when no trailers are appended):
 
 ```sh
 git-commit-message --one-line "optional context"
+
+# with trailers, output is subject plus trailer lines
+git-commit-message --one-line --co-author 'John Doe <john.doe@example.com>'
 ```
 
 Select provider:
@@ -158,6 +178,9 @@ git-commit-message --provider google
 
 # Ollama
 git-commit-message --provider ollama
+
+# llama.cpp
+git-commit-message --provider llamacpp
 ```
 
 Commit immediately (optionally open editor):
@@ -165,6 +188,24 @@ Commit immediately (optionally open editor):
 ```sh
 git-commit-message --commit "refactor parser for speed"
 git-commit-message --commit --edit "refactor parser for speed"
+
+# add co-author trailers
+git-commit-message --commit --co-author 'John Doe <john.doe@example.com>'
+git-commit-message --commit --co-author 'John Doe <john.doe@example.com>' --co-author 'Jane Doe <jane.doe@example.com>'
+git-commit-message --commit --co-author copilot
+```
+
+Amend the previous commit:
+
+```sh
+# print only (useful for pasting into a GUI editor)
+git-commit-message --amend "optional context"
+
+# amend immediately
+git-commit-message --commit --amend "optional context"
+
+# amend immediately, but open editor for final tweaks
+git-commit-message --commit --amend --edit "optional context"
 ```
 
 Limit subject length:
@@ -206,18 +247,26 @@ Configure Ollama host (if running on a different machine):
 git-commit-message --provider ollama --host http://192.168.1.100:11434
 ```
 
+Configure llama.cpp host:
+
+```sh
+git-commit-message --provider llamacpp --host http://192.168.1.100:8080
+```
+
 ## Options
 
-- `--provider {openai,google,ollama}`: provider to use (default: `openai`)
-- `--model MODEL`: model override (provider-specific)
+- `--provider {openai,google,ollama,llamacpp}`: provider to use (default: `openai`)
+- `--model MODEL`: model override (provider-specific; ignored for llama.cpp)
 - `--language TAG`: output language/locale (default: `en-GB`)
-- `--one-line`: output subject only
+- `--one-line`: output subject only when no trailers are appended; with `--co-author`, output is a single-line subject plus `Co-authored-by:` trailer lines
 - `--max-length N`: max subject length (default: 72)
 - `--chunk-tokens N`: token budget per diff chunk (`0` = single summary pass, `-1` disables summarisation)
 - `--debug`: print request/response details
 - `--commit`: run `git commit -m <message>`
+- `--amend`: generate a message suitable for amending the previous commit (diff is from the amended commit's parent to the staged index; if nothing is staged, this effectively becomes the diff introduced by `HEAD`)
 - `--edit`: with `--commit`, open editor for final message
-- `--host URL`: host URL for providers like Ollama (default: `http://localhost:11434`)
+- `--host URL`: host URL for providers like Ollama or llama.cpp (default: `http://localhost:11434` for Ollama, `http://localhost:8080` for llama.cpp)
+- `--co-author VALUE`: append `Co-authored-by:` trailer(s). Repeat to add multiple values. Accepted forms: `Name <email@example.com>` or `copilot` (alias, case-insensitive).
 
 ## Environment variables
 
@@ -233,6 +282,7 @@ Optional:
 - `OPENAI_MODEL`: OpenAI-only model override (used if `--model`/`GIT_COMMIT_MESSAGE_MODEL` are not set)
 - `OLLAMA_MODEL`: Ollama-only model override (used if `--model`/`GIT_COMMIT_MESSAGE_MODEL` are not set)
 - `OLLAMA_HOST`: Ollama server URL (default: `http://localhost:11434`)
+- `LLAMACPP_HOST`: llama.cpp server URL (default: `http://localhost:8080`)
 - `GIT_COMMIT_MESSAGE_LANGUAGE`: default language/locale (default: `en-GB`)
 - `GIT_COMMIT_MESSAGE_CHUNK_TOKENS`: default chunk token budget (default: `0`)
 
@@ -241,6 +291,7 @@ Default models (if not overridden):
 - OpenAI: `gpt-5-mini`
 - Google: `gemini-2.5-flash`
 - Ollama: `gpt-oss:20b`
+- llama.cpp: uses pre-loaded model (model parameter is ignored)
 
 ## AI-generated code notice
 

{git_commit_message-0.8.0 → git_commit_message-0.8.2}/src/git_commit_message.egg-info/SOURCES.txt

@@ -7,6 +7,7 @@ src/git_commit_message/_cli.py
 src/git_commit_message/_gemini.py
 src/git_commit_message/_git.py
 src/git_commit_message/_gpt.py
+src/git_commit_message/_llamacpp.py
 src/git_commit_message/_llm.py
 src/git_commit_message/_ollama.py
 src/git_commit_message.egg-info/PKG-INFO

git_commit_message-0.8.0/src/git_commit_message/_git.py

@@ -1,114 +0,0 @@
-"""Git-related helper functions.
-
-Provides repository root discovery, extraction of staged changes, and
-creating commits from a message.
-"""
-
-from __future__ import annotations
-
-from pathlib import Path
-from subprocess import CalledProcessError, check_call, check_output, run
-
-
-def get_repo_root(
-    cwd: Path | None = None,
-    /,
-) -> Path:
-    """Find the repository root from the current working directory.
-
-    Parameters
-    ----------
-    cwd
-        Starting directory for the search. Defaults to the current working directory.
-
-    Returns
-    -------
-    Path
-        The repository root path.
-    """
-
-    start: Path = cwd or Path.cwd()
-    try:
-        out: bytes = check_output(
-            [
-                "git",
-                "rev-parse",
-                "--show-toplevel",
-            ],
-            cwd=str(start),
-        )
-    except CalledProcessError as exc:  # noqa: TRY003
-        raise RuntimeError("Not a Git repository.") from exc
-
-    root = Path(out.decode().strip())
-    return root
-
-
-def has_staged_changes(
-    cwd: Path,
-    /,
-) -> bool:
-    """Check whether there are staged changes."""
-
-    try:
-        check_call(
-            ["git", "diff", "--cached", "--quiet", "--exit-code"],
-            cwd=str(cwd),
-        )
-        return False
-    except CalledProcessError:
-        return True
-
-
-def get_staged_diff(
-    cwd: Path,
-    /,
-) -> str:
-    """Return the staged changes as diff text."""
-
-    out: bytes = check_output(
-        [
-            "git",
-            "diff",
-            "--cached",
-            "--patch",
-            "--minimal",
-            "--no-color",
-        ],
-        cwd=str(cwd),
-    )
-    return out.decode()
-
-
-def commit_with_message(
-    message: str,
-    edit: bool,
-    cwd: Path,
-    /,
-) -> int:
-    """Create a commit with the given message.
-
-    Parameters
-    ----------
-    message
-        Commit message.
-    edit
-        If True, use the `--edit` flag to open an editor for amendments.
-    cwd
-        Git working directory.
-
-    Returns
-    -------
-    int
-        The subprocess exit code.
-    """
-
-    cmd: list[str] = ["git", "commit", "-m", message]
-    if edit:
-        cmd.append("--edit")
-
-    try:
-        completed = run(cmd, cwd=str(cwd), check=False)
-        return int(completed.returncode)
-    except OSError as exc:  # e.g., editor launch failure, etc.
-        raise RuntimeError(f"Failed to run 'git commit': {exc}") from exc