commitai 1.0.4__tar.gz → 2.0.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {commitai-1.0.4 → commitai-2.0.0}/.github/workflows/main.yml +1 -1
- {commitai-1.0.4 → commitai-2.0.0}/.pre-commit-config.yaml +8 -10
- {commitai-1.0.4 → commitai-2.0.0}/PKG-INFO +35 -20
- {commitai-1.0.4 → commitai-2.0.0}/README.md +28 -10
- {commitai-1.0.4 → commitai-2.0.0}/commitai/__init__.py +1 -1
- commitai-2.0.0/commitai/agent.py +222 -0
- commitai-2.0.0/commitai/chains.py +135 -0
- {commitai-1.0.4 → commitai-2.0.0}/commitai/cli.py +146 -105
- {commitai-1.0.4 → commitai-2.0.0}/commitai/template.py +21 -0
- {commitai-1.0.4 → commitai-2.0.0}/pyproject.toml +12 -13
- commitai-2.0.0/tests/test_agent.py +27 -0
- commitai-2.0.0/tests/test_agent_chains.py +80 -0
- {commitai-1.0.4 → commitai-2.0.0}/tests/test_cli.py +110 -194
- {commitai-1.0.4 → commitai-2.0.0}/.gitignore +0 -0
- {commitai-1.0.4 → commitai-2.0.0}/LICENSE +0 -0
- {commitai-1.0.4 → commitai-2.0.0}/assets/commitaai.gif +0 -0
- {commitai-1.0.4 → commitai-2.0.0}/bitmap.png +0 -0
- {commitai-1.0.4 → commitai-2.0.0}/commitai/git.py +0 -0
- {commitai-1.0.4 → commitai-2.0.0}/tests/test_git.py +0 -0
- {commitai-1.0.4 → commitai-2.0.0}/tests/test_template.py +0 -0
|
@@ -25,16 +25,14 @@ repos:
|
|
|
25
25
|
hooks:
|
|
26
26
|
- id: mypy
|
|
27
27
|
# Ensure mypy runs with the necessary dependencies installed
|
|
28
|
-
additional_dependencies:
|
|
29
|
-
|
|
30
|
-
|
|
31
|
-
|
|
32
|
-
|
|
33
|
-
|
|
34
|
-
|
|
35
|
-
|
|
36
|
-
"types-setuptools"
|
|
37
|
-
]
|
|
28
|
+
additional_dependencies:
|
|
29
|
+
- "click>=8.0,<9.0"
|
|
30
|
+
- "langchain>=0.1.0,<=0.3.25"
|
|
31
|
+
- "langchain-core>=0.1.0,<=0.3.58"
|
|
32
|
+
- "langchain-community>=0.0.20,<=0.3.23"
|
|
33
|
+
- "langchain-google-genai~=2.1.4"
|
|
34
|
+
- "pydantic>=2.0,<3.0"
|
|
35
|
+
- "types-setuptools"
|
|
38
36
|
args: [--config-file=pyproject.toml] # Point mypy to the config
|
|
39
37
|
# You might need to adjust entry if your structure changes
|
|
40
38
|
# entry: mypy commitai commitai/tests
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: commitai
|
|
3
|
-
Version:
|
|
3
|
+
Version: 2.0.0
|
|
4
4
|
Summary: Commitai helps you generate git commit messages using AI
|
|
5
5
|
Project-URL: Bug Tracker, https://github.com/lguibr/commitai/issues
|
|
6
6
|
Project-URL: Documentation, https://github.com/lguibr/commitai/blob/main/README.md
|
|
@@ -34,23 +34,20 @@ Classifier: Intended Audience :: Developers
|
|
|
34
34
|
Classifier: License :: OSI Approved :: MIT License
|
|
35
35
|
Classifier: Operating System :: OS Independent
|
|
36
36
|
Classifier: Programming Language :: Python :: 3
|
|
37
|
-
Classifier: Programming Language :: Python :: 3.9
|
|
38
37
|
Classifier: Programming Language :: Python :: 3.10
|
|
39
38
|
Classifier: Programming Language :: Python :: 3.11
|
|
40
39
|
Classifier: Programming Language :: Python :: 3.12
|
|
41
40
|
Classifier: Topic :: Software Development :: Version Control :: Git
|
|
42
41
|
Classifier: Topic :: Utilities
|
|
43
|
-
Requires-Python: >=3.
|
|
42
|
+
Requires-Python: >=3.10
|
|
44
43
|
Requires-Dist: click<9.0,>=8.0
|
|
45
|
-
Requires-Dist: langchain-
|
|
46
|
-
Requires-Dist: langchain-
|
|
47
|
-
Requires-Dist: langchain-
|
|
48
|
-
Requires-Dist: langchain
|
|
49
|
-
Requires-Dist: langchain-openai<0.3.0,>=0.1.0
|
|
50
|
-
Requires-Dist: langchain<0.3.0,>=0.1.0
|
|
44
|
+
Requires-Dist: langchain-community<=0.3.23,>=0.0.20
|
|
45
|
+
Requires-Dist: langchain-core<=0.3.58,>=0.1.0
|
|
46
|
+
Requires-Dist: langchain-google-genai~=2.1.4
|
|
47
|
+
Requires-Dist: langchain<=0.3.25,>=0.1.0
|
|
51
48
|
Requires-Dist: pydantic<3.0,>=2.0
|
|
52
49
|
Provides-Extra: test
|
|
53
|
-
Requires-Dist: langchain-google-genai~=
|
|
50
|
+
Requires-Dist: langchain-google-genai~=2.1.4; extra == 'test'
|
|
54
51
|
Requires-Dist: mypy>=1.9.0; extra == 'test'
|
|
55
52
|
Requires-Dist: pytest-cov>=3.0; extra == 'test'
|
|
56
53
|
Requires-Dist: pytest>=7.0; extra == 'test'
|
|
@@ -67,9 +64,13 @@ Description-Content-Type: text/markdown
|
|
|
67
64
|
[](https://github.com/lguibr/CommitAi/blob/main/LICENSE)
|
|
68
65
|
[](https://github.com/astral-sh/ruff)
|
|
69
66
|
|
|
70
|
-
|
|
71
|
-
|
|
72
|
-
|
|
67
|
+
|
|
68
|
+
<img
|
|
69
|
+
src="https://raw.githubusercontent.com/lguibr/commitai/main/bitmap.png"
|
|
70
|
+
alt="screenshot"
|
|
71
|
+
width="400"
|
|
72
|
+
/>
|
|
73
|
+
|
|
73
74
|
|
|
74
75
|
**Tired of writing Git commit messages? Let AI do the heavy lifting!**
|
|
75
76
|
|
|
@@ -97,12 +98,12 @@ Simply stage your files and run `commitai`. It analyzes the diff, optionally tak
|
|
|
97
98
|
|
|
98
99
|
## Features
|
|
99
100
|
|
|
100
|
-
* 🧠 **Intelligent Commit Generation**: Analyzes staged code differences (`git diff --staged`) using state-of-the-art AI models (GPT, Claude
|
|
101
|
+
* 🧠 **Intelligent Commit Generation**: Analyzes staged code differences (`git diff --staged`) using state-of-the-art AI models (Gemini, GPT, Claude) to create meaningful commit messages.
|
|
101
102
|
* 📄 **Conventional Commits**: Automatically formats messages according to the Conventional Commits specification (e.g., `feat(auth): add JWT authentication`). This improves readability and enables automated changelog generation.
|
|
102
103
|
* 📝 **Optional Explanations**: Provide a high-level description of your changes as input to guide the AI, or let it infer the context solely from the code diff.
|
|
103
104
|
* ✅ **Pre-commit Hook Integration**: Automatically runs your existing native Git pre-commit hook (`.git/hooks/pre-commit`) before generating the message, ensuring code quality and style checks pass.
|
|
104
105
|
* 🔧 **Customizable Prompts via Templates**: Add custom instructions or context to the AI prompt using global environment variables or repository-specific template files.
|
|
105
|
-
* 🤖 **Multiple AI Provider Support**: Choose your preferred AI model from OpenAI, Anthropic, or
|
|
106
|
+
* 🤖 **Multiple AI Provider Support**: Choose your preferred AI model from OpenAI, Anthropic, Google or local AI models with Ollama.
|
|
106
107
|
* ⚙️ **Flexible Workflow**:
|
|
107
108
|
* Stages all changes automatically (`-a` flag).
|
|
108
109
|
* Reviews message in your default Git editor (default behavior).
|
|
@@ -152,7 +153,15 @@ CommitAi requires API keys for the AI provider you intend to use. Set these as e
|
|
|
152
153
|
export GOOGLE_API_KEY="your_google_api_key_here"
|
|
153
154
|
```
|
|
154
155
|
|
|
155
|
-
You only need to set the key for the provider corresponding to the model you select (or the default, Gemini).
|
|
156
|
+
You only need to set the key for the provider corresponding to the model you select (or the default, Gemini 3 Flash with Google).
|
|
157
|
+
|
|
158
|
+
### Ollama
|
|
159
|
+
|
|
160
|
+
CommitAi can also work with Ollama models:
|
|
161
|
+
```bash
|
|
162
|
+
export OLLAMA_HOST="your_ollama_base_url"
|
|
163
|
+
```
|
|
164
|
+
|
|
156
165
|
|
|
157
166
|
### Commit Templates (Optional)
|
|
158
167
|
|
|
@@ -211,14 +220,20 @@ The `commitai` command (which is an alias for `commitai generate`) accepts the f
|
|
|
211
220
|
* Example: `commitai -c "Fix typo in documentation"` (for minor changes)
|
|
212
221
|
* Can be combined with `-a`: `commitai -a -c "Quick fix and commit all"`
|
|
213
222
|
|
|
223
|
+
* `--review` / `--no-review`:
|
|
224
|
+
* Toggle a preliminary AI review of the staged diff before generating the commit message. Default is `--review` (enabled).
|
|
225
|
+
* When enabled, CommitAi prints a brief review and asks if you want to proceed.
|
|
226
|
+
* Example: `commitai --no-review` to skip the review step.
|
|
227
|
+
|
|
214
228
|
* `-m <model_name>`, `--model <model_name>`:
|
|
215
229
|
* Specifies which AI model to use.
|
|
216
|
-
*
|
|
230
|
+
* Specifies which AI model to use.
|
|
231
|
+
* Defaults to `gemini-3-flash-preview`.
|
|
217
232
|
* Ensure the corresponding API key environment variable is set.
|
|
218
233
|
* Examples:
|
|
234
|
+
* `commitai -m gemini-3-pro-preview "Use Google's Gemini 3 Pro"`
|
|
219
235
|
* `commitai -m gpt-4 "Use OpenAI's GPT-4"`
|
|
220
|
-
* `commitai -m claude-3-opus
|
|
221
|
-
* `commitai -m gemini-2.5-flash-preview-04-17 "Use Google's Gemini 1.5 Flash"`
|
|
236
|
+
* `commitai -m claude-3-opus "Use Anthropic's Claude 3 Opus"`
|
|
222
237
|
|
|
223
238
|
### Creating Repository Templates
|
|
224
239
|
|
|
@@ -290,7 +305,7 @@ Contributions are highly welcome! Please follow these steps:
|
|
|
290
305
|
9. Run checks locally before committing:
|
|
291
306
|
* Format code: `ruff format .`
|
|
292
307
|
* Lint code: `ruff check .`
|
|
293
|
-
* Run type checks: `mypy commitai
|
|
308
|
+
* Run type checks: `mypy commitai tests`
|
|
294
309
|
* Run tests: `pytest`
|
|
295
310
|
10. Commit your changes (you can use `commitai`!).
|
|
296
311
|
11. Push your branch to your fork: `git push origin my-feature-branch`
|
|
@@ -7,9 +7,13 @@
|
|
|
7
7
|
[](https://github.com/lguibr/CommitAi/blob/main/LICENSE)
|
|
8
8
|
[](https://github.com/astral-sh/ruff)
|
|
9
9
|
|
|
10
|
-
|
|
11
|
-
|
|
12
|
-
|
|
10
|
+
|
|
11
|
+
<img
|
|
12
|
+
src="https://raw.githubusercontent.com/lguibr/commitai/main/bitmap.png"
|
|
13
|
+
alt="screenshot"
|
|
14
|
+
width="400"
|
|
15
|
+
/>
|
|
16
|
+
|
|
13
17
|
|
|
14
18
|
**Tired of writing Git commit messages? Let AI do the heavy lifting!**
|
|
15
19
|
|
|
@@ -37,12 +41,12 @@ Simply stage your files and run `commitai`. It analyzes the diff, optionally tak
|
|
|
37
41
|
|
|
38
42
|
## Features
|
|
39
43
|
|
|
40
|
-
* 🧠 **Intelligent Commit Generation**: Analyzes staged code differences (`git diff --staged`) using state-of-the-art AI models (GPT, Claude
|
|
44
|
+
* 🧠 **Intelligent Commit Generation**: Analyzes staged code differences (`git diff --staged`) using state-of-the-art AI models (Gemini, GPT, Claude) to create meaningful commit messages.
|
|
41
45
|
* 📄 **Conventional Commits**: Automatically formats messages according to the Conventional Commits specification (e.g., `feat(auth): add JWT authentication`). This improves readability and enables automated changelog generation.
|
|
42
46
|
* 📝 **Optional Explanations**: Provide a high-level description of your changes as input to guide the AI, or let it infer the context solely from the code diff.
|
|
43
47
|
* ✅ **Pre-commit Hook Integration**: Automatically runs your existing native Git pre-commit hook (`.git/hooks/pre-commit`) before generating the message, ensuring code quality and style checks pass.
|
|
44
48
|
* 🔧 **Customizable Prompts via Templates**: Add custom instructions or context to the AI prompt using global environment variables or repository-specific template files.
|
|
45
|
-
* 🤖 **Multiple AI Provider Support**: Choose your preferred AI model from OpenAI, Anthropic, or
|
|
49
|
+
* 🤖 **Multiple AI Provider Support**: Choose your preferred AI model from OpenAI, Anthropic, Google or local AI models with Ollama.
|
|
46
50
|
* ⚙️ **Flexible Workflow**:
|
|
47
51
|
* Stages all changes automatically (`-a` flag).
|
|
48
52
|
* Reviews message in your default Git editor (default behavior).
|
|
@@ -92,7 +96,15 @@ CommitAi requires API keys for the AI provider you intend to use. Set these as e
|
|
|
92
96
|
export GOOGLE_API_KEY="your_google_api_key_here"
|
|
93
97
|
```
|
|
94
98
|
|
|
95
|
-
You only need to set the key for the provider corresponding to the model you select (or the default, Gemini).
|
|
99
|
+
You only need to set the key for the provider corresponding to the model you select (or the default, Gemini 3 Flash with Google).
|
|
100
|
+
|
|
101
|
+
### Ollama
|
|
102
|
+
|
|
103
|
+
CommitAi can also work with Ollama models:
|
|
104
|
+
```bash
|
|
105
|
+
export OLLAMA_HOST="your_ollama_base_url"
|
|
106
|
+
```
|
|
107
|
+
|
|
96
108
|
|
|
97
109
|
### Commit Templates (Optional)
|
|
98
110
|
|
|
@@ -151,14 +163,20 @@ The `commitai` command (which is an alias for `commitai generate`) accepts the f
|
|
|
151
163
|
* Example: `commitai -c "Fix typo in documentation"` (for minor changes)
|
|
152
164
|
* Can be combined with `-a`: `commitai -a -c "Quick fix and commit all"`
|
|
153
165
|
|
|
166
|
+
* `--review` / `--no-review`:
|
|
167
|
+
* Toggle a preliminary AI review of the staged diff before generating the commit message. Default is `--review` (enabled).
|
|
168
|
+
* When enabled, CommitAi prints a brief review and asks if you want to proceed.
|
|
169
|
+
* Example: `commitai --no-review` to skip the review step.
|
|
170
|
+
|
|
154
171
|
* `-m <model_name>`, `--model <model_name>`:
|
|
155
172
|
* Specifies which AI model to use.
|
|
156
|
-
*
|
|
173
|
+
* Specifies which AI model to use.
|
|
174
|
+
* Defaults to `gemini-3-flash-preview`.
|
|
157
175
|
* Ensure the corresponding API key environment variable is set.
|
|
158
176
|
* Examples:
|
|
177
|
+
* `commitai -m gemini-3-pro-preview "Use Google's Gemini 3 Pro"`
|
|
159
178
|
* `commitai -m gpt-4 "Use OpenAI's GPT-4"`
|
|
160
|
-
* `commitai -m claude-3-opus
|
|
161
|
-
* `commitai -m gemini-2.5-flash-preview-04-17 "Use Google's Gemini 1.5 Flash"`
|
|
179
|
+
* `commitai -m claude-3-opus "Use Anthropic's Claude 3 Opus"`
|
|
162
180
|
|
|
163
181
|
### Creating Repository Templates
|
|
164
182
|
|
|
@@ -230,7 +248,7 @@ Contributions are highly welcome! Please follow these steps:
|
|
|
230
248
|
9. Run checks locally before committing:
|
|
231
249
|
* Format code: `ruff format .`
|
|
232
250
|
* Lint code: `ruff check .`
|
|
233
|
-
* Run type checks: `mypy commitai
|
|
251
|
+
* Run type checks: `mypy commitai tests`
|
|
234
252
|
* Run tests: `pytest`
|
|
235
253
|
10. Commit your changes (you can use `commitai`!).
|
|
236
254
|
11. Push your branch to your fork: `git push origin my-feature-branch`
|
|
@@ -3,7 +3,7 @@
|
|
|
3
3
|
|
|
4
4
|
# This __version__ string is read by hatchling during the build process
|
|
5
5
|
# Make sure to update it for new releases.
|
|
6
|
-
__version__ = "0.
|
|
6
|
+
__version__ = "2.0.0"
|
|
7
7
|
|
|
8
8
|
# The importlib.metadata approach is generally for reading the version
|
|
9
9
|
# of an *already installed* package at runtime. We don't need it here
|
|
@@ -0,0 +1,222 @@
|
|
|
1
|
+
import glob
|
|
2
|
+
import os
|
|
3
|
+
import subprocess
|
|
4
|
+
from typing import Any, Dict, Type
|
|
5
|
+
|
|
6
|
+
from langchain.agents import AgentExecutor, create_tool_calling_agent
|
|
7
|
+
from langchain_core.language_models import BaseChatModel
|
|
8
|
+
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
|
|
9
|
+
from langchain_core.runnables import Runnable
|
|
10
|
+
from langchain_core.tools import BaseTool
|
|
11
|
+
from pydantic import BaseModel, Field
|
|
12
|
+
|
|
13
|
+
# --- TOOLS ---
|
|
14
|
+
|
|
15
|
+
|
|
16
|
+
class ShellInput(BaseModel):
|
|
17
|
+
command: str = Field(
|
|
18
|
+
description=(
|
|
19
|
+
"The git command to execute (e.g., 'git status', 'git log'). "
|
|
20
|
+
"Must start with 'git'."
|
|
21
|
+
)
|
|
22
|
+
)
|
|
23
|
+
|
|
24
|
+
|
|
25
|
+
class ReadOnlyShellTool(BaseTool):
|
|
26
|
+
name: str = "git_shell"
|
|
27
|
+
description: str = (
|
|
28
|
+
"Run read-only git commands to inspect the repository state. "
|
|
29
|
+
"Only 'git' commands are allowed. Write operations are blocked."
|
|
30
|
+
)
|
|
31
|
+
args_schema: Type[BaseModel] = ShellInput
|
|
32
|
+
|
|
33
|
+
def _run(self, command: str) -> str:
|
|
34
|
+
command = command.strip()
|
|
35
|
+
if not command.startswith("git"):
|
|
36
|
+
return "Error: Only 'git' commands are allowed."
|
|
37
|
+
|
|
38
|
+
# Simple blocklist for write operations
|
|
39
|
+
forbidden = [
|
|
40
|
+
"push",
|
|
41
|
+
"pull",
|
|
42
|
+
"commit",
|
|
43
|
+
"merge",
|
|
44
|
+
"rebase",
|
|
45
|
+
"cherry-pick",
|
|
46
|
+
"stash",
|
|
47
|
+
"clean",
|
|
48
|
+
"reset",
|
|
49
|
+
"checkout",
|
|
50
|
+
"switch",
|
|
51
|
+
"branch",
|
|
52
|
+
]
|
|
53
|
+
if any(w in command.split() for w in forbidden):
|
|
54
|
+
return f"Error: Command '{command}' contains forbidden write operations."
|
|
55
|
+
|
|
56
|
+
try:
|
|
57
|
+
# shell=True is dangerous in general, but we heavily restricted input above
|
|
58
|
+
result = subprocess.run(
|
|
59
|
+
command, shell=True, capture_output=True, text=True, cwd=os.getcwd()
|
|
60
|
+
)
|
|
61
|
+
if result.returncode != 0:
|
|
62
|
+
return f"Error ({result.returncode}): {result.stderr}"
|
|
63
|
+
return result.stdout
|
|
64
|
+
except Exception as e:
|
|
65
|
+
return f"Execution Error: {str(e)}"
|
|
66
|
+
|
|
67
|
+
|
|
68
|
+
class FileSearchInput(BaseModel):
|
|
69
|
+
pattern: str = Field(
|
|
70
|
+
description="The glob pattern to search for files (e.g., 'src/**/*.py')."
|
|
71
|
+
)
|
|
72
|
+
|
|
73
|
+
|
|
74
|
+
class FileSearchTool(BaseTool):
|
|
75
|
+
name: str = "file_search"
|
|
76
|
+
description: str = (
|
|
77
|
+
"Search for file paths in the project using glob patterns. "
|
|
78
|
+
"Useful to find files to inspect."
|
|
79
|
+
)
|
|
80
|
+
args_schema: Type[BaseModel] = FileSearchInput
|
|
81
|
+
|
|
82
|
+
def _run(self, pattern: str) -> str:
|
|
83
|
+
try:
|
|
84
|
+
# Security: prevent breaking out of repo?
|
|
85
|
+
# For simplicity, just run glob.
|
|
86
|
+
if ".." in pattern:
|
|
87
|
+
return "Error: '..' not allowed in patterns."
|
|
88
|
+
|
|
89
|
+
files = glob.glob(pattern, recursive=True)
|
|
90
|
+
if not files:
|
|
91
|
+
return "No files found."
|
|
92
|
+
return "\n".join(files[:20]) # Limit output
|
|
93
|
+
except Exception as e:
|
|
94
|
+
return f"Error: {str(e)}"
|
|
95
|
+
|
|
96
|
+
|
|
97
|
+
class FileReadInput(BaseModel):
|
|
98
|
+
file_path: str = Field(description="The path of the file to read.")
|
|
99
|
+
|
|
100
|
+
|
|
101
|
+
class FileReadTool(BaseTool):
|
|
102
|
+
name: str = "file_read"
|
|
103
|
+
description: str = "Read the contents of a specific file."
|
|
104
|
+
args_schema: Type[BaseModel] = FileReadInput
|
|
105
|
+
|
|
106
|
+
def _run(self, file_path: str) -> str:
|
|
107
|
+
if ".." in file_path:
|
|
108
|
+
return "Error: Traversing up directories is not allowed."
|
|
109
|
+
if not os.path.exists(file_path):
|
|
110
|
+
return "Error: File does not exist."
|
|
111
|
+
try:
|
|
112
|
+
with open(file_path, "r") as f:
|
|
113
|
+
content = f.read()
|
|
114
|
+
return content[:2000] # Truncate large files
|
|
115
|
+
except Exception as e:
|
|
116
|
+
return f"Error reading file: {str(e)}"
|
|
117
|
+
|
|
118
|
+
|
|
119
|
+
# --- MIDDLEWARE (Simulated for Agent) ---
|
|
120
|
+
|
|
121
|
+
|
|
122
|
+
class SummarizationMiddleware:
|
|
123
|
+
"""Uses LLM to summarize diff before agent sees it."""
|
|
124
|
+
|
|
125
|
+
def __init__(self, llm: BaseChatModel):
|
|
126
|
+
self.llm = llm
|
|
127
|
+
|
|
128
|
+
def process(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
|
|
129
|
+
diff = inputs.get("diff", "")
|
|
130
|
+
if not diff:
|
|
131
|
+
return inputs
|
|
132
|
+
|
|
133
|
+
# Simple summarization chain (inline invocation)
|
|
134
|
+
# Truncate for summary
|
|
135
|
+
msg = f"Summarize these changes in 2 sentences:\n\n{diff[:5000]}"
|
|
136
|
+
resp = self.llm.invoke(msg)
|
|
137
|
+
inputs["summary"] = resp.content
|
|
138
|
+
return inputs
|
|
139
|
+
|
|
140
|
+
|
|
141
|
+
class TodoMiddleware:
|
|
142
|
+
"""Scans diff for TODOs and adds to inputs."""
|
|
143
|
+
|
|
144
|
+
def process(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
|
|
145
|
+
diff = inputs.get("diff", "")
|
|
146
|
+
todos = []
|
|
147
|
+
for line in diff.splitlines():
|
|
148
|
+
if line.startswith("+") and any(
|
|
149
|
+
x in line.lower() for x in ["todo", "fixme"]
|
|
150
|
+
):
|
|
151
|
+
todos.append(line[1:].strip())
|
|
152
|
+
inputs["todos"] = todos
|
|
153
|
+
inputs["todo_str"] = "\n".join(f"- {t}" for t in todos) if todos else "None"
|
|
154
|
+
return inputs
|
|
155
|
+
|
|
156
|
+
|
|
157
|
+
# --- AGENT ---
|
|
158
|
+
|
|
159
|
+
|
|
160
|
+
def create_commit_agent(llm: BaseChatModel) -> Runnable:
|
|
161
|
+
# 1. Init Tools
|
|
162
|
+
tools = [ReadOnlyShellTool(), FileSearchTool(), FileReadTool()]
|
|
163
|
+
|
|
164
|
+
# 2. Middlewares
|
|
165
|
+
summ_mw = SummarizationMiddleware(llm)
|
|
166
|
+
todo_mw = TodoMiddleware()
|
|
167
|
+
|
|
168
|
+
# 3. Prompt
|
|
169
|
+
system_prompt = """You are an expert software engineer acting as a Commit Assistant.
|
|
170
|
+
Your goal is to generate a conventional commit message.
|
|
171
|
+
|
|
172
|
+
Context:
|
|
173
|
+
- User Explanation: {explanation}
|
|
174
|
+
- Detected TODOs: {todo_str}
|
|
175
|
+
- Auto-Summary: {summary}
|
|
176
|
+
- Staged Diff: {diff}
|
|
177
|
+
|
|
178
|
+
You have access to tools to explore the codebase if the diff + explanation is ambiguous.
|
|
179
|
+
- Use `git_shell` to check status or logs.
|
|
180
|
+
- Use `file_search` and `file_read` to understand context of modified files.
|
|
181
|
+
|
|
182
|
+
Protocol:
|
|
183
|
+
1. Analyze the input.
|
|
184
|
+
2. If detecting POTENTIAL SENSITIVE DATA (API keys, secrets) in the diff, you MUST stop
|
|
185
|
+
and ask the user (simulated by returning a warning message).
|
|
186
|
+
3. If clarification is needed, explore files.
|
|
187
|
+
4. Final Answer MUST be ONLY the commit message.
|
|
188
|
+
"""
|
|
189
|
+
prompt = ChatPromptTemplate.from_messages(
|
|
190
|
+
[
|
|
191
|
+
("system", system_prompt),
|
|
192
|
+
MessagesPlaceholder("chat_history", optional=True),
|
|
193
|
+
("human", "Generate the commit message."),
|
|
194
|
+
MessagesPlaceholder("agent_scratchpad"),
|
|
195
|
+
]
|
|
196
|
+
)
|
|
197
|
+
|
|
198
|
+
# 4. Construct Agent
|
|
199
|
+
agent = create_tool_calling_agent(llm, tools, prompt)
|
|
200
|
+
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=False)
|
|
201
|
+
|
|
202
|
+
# 5. Pipeline with Middleware
|
|
203
|
+
def run_pipeline(inputs: Dict[str, Any]) -> str:
|
|
204
|
+
# Run Middleware
|
|
205
|
+
state = inputs.copy()
|
|
206
|
+
state = todo_mw.process(state)
|
|
207
|
+
state = summ_mw.process(state)
|
|
208
|
+
|
|
209
|
+
# Inject formatted fields if missing
|
|
210
|
+
state.setdefault("explanation", "None")
|
|
211
|
+
state.setdefault("summary", "None")
|
|
212
|
+
state.setdefault("todo_str", "None")
|
|
213
|
+
state.setdefault("chat_history", [])
|
|
214
|
+
|
|
215
|
+
# Run Agent
|
|
216
|
+
result = agent_executor.invoke(state)
|
|
217
|
+
return str(result["output"])
|
|
218
|
+
|
|
219
|
+
# Wrap in RunnableLambda to expose 'invoke'
|
|
220
|
+
from langchain_core.runnables import RunnableLambda
|
|
221
|
+
|
|
222
|
+
return RunnableLambda(run_pipeline)
|
|
@@ -0,0 +1,135 @@
|
|
|
1
|
+
from typing import Any, Dict, List, Optional, TypedDict
|
|
2
|
+
|
|
3
|
+
from langchain_core.language_models import BaseChatModel
|
|
4
|
+
from langchain_core.output_parsers import StrOutputParser
|
|
5
|
+
from langchain_core.prompts import ChatPromptTemplate
|
|
6
|
+
from langchain_core.runnables import Runnable, RunnableLambda, RunnablePassthrough
|
|
7
|
+
|
|
8
|
+
|
|
9
|
+
class CommitState(TypedDict):
|
|
10
|
+
diff: str
|
|
11
|
+
explanation: Optional[str]
|
|
12
|
+
summary: Optional[str]
|
|
13
|
+
todos: Optional[List[str]]
|
|
14
|
+
|
|
15
|
+
|
|
16
|
+
class SummarizationMiddleware:
|
|
17
|
+
"""Middleware to summarize the diff before generating the commit message."""
|
|
18
|
+
|
|
19
|
+
def __init__(self, llm: BaseChatModel):
|
|
20
|
+
self.llm = llm
|
|
21
|
+
self.prompt = ChatPromptTemplate.from_template(
|
|
22
|
+
"Summarize the following code changes concisely in 1-2 sentences:\n\n{diff}"
|
|
23
|
+
)
|
|
24
|
+
self.chain = self.prompt | self.llm | StrOutputParser()
|
|
25
|
+
|
|
26
|
+
def __call__(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
|
|
27
|
+
"""Run the summarizer and add 'summary' to the state."""
|
|
28
|
+
diff = inputs.get("diff", "")
|
|
29
|
+
if not diff:
|
|
30
|
+
return {**inputs, "summary": ""}
|
|
31
|
+
|
|
32
|
+
# We invoke the chain synchronously here
|
|
33
|
+
summary = self.chain.invoke({"diff": diff})
|
|
34
|
+
return {**inputs, "summary": summary}
|
|
35
|
+
|
|
36
|
+
|
|
37
|
+
class TodoMiddleware:
|
|
38
|
+
"""Middleware to scan the diff for TODO/FIXME/HACK comments."""
|
|
39
|
+
|
|
40
|
+
def __call__(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
|
|
41
|
+
diff = inputs.get("diff", "")
|
|
42
|
+
todos = []
|
|
43
|
+
for line in diff.splitlines():
|
|
44
|
+
if line.startswith("+"):
|
|
45
|
+
lower_line = line.lower()
|
|
46
|
+
if (
|
|
47
|
+
"todo" in lower_line
|
|
48
|
+
or "fixme" in lower_line
|
|
49
|
+
or "hack" in lower_line
|
|
50
|
+
):
|
|
51
|
+
# Strip the + and whitespace
|
|
52
|
+
clean_line = line[1:].strip()
|
|
53
|
+
todos.append(clean_line)
|
|
54
|
+
|
|
55
|
+
return {**inputs, "todos": todos}
|
|
56
|
+
|
|
57
|
+
|
|
58
|
+
def create_commit_chain(llm: BaseChatModel) -> Runnable:
|
|
59
|
+
"""Creates the full commit generation pipeline with middleware."""
|
|
60
|
+
|
|
61
|
+
# 1. Initialize Middlewares
|
|
62
|
+
summarizer = SummarizationMiddleware(llm)
|
|
63
|
+
todo_scanner = TodoMiddleware()
|
|
64
|
+
|
|
65
|
+
# 2. Define the Prompt
|
|
66
|
+
# We include placeholders for summary and todos if they exist
|
|
67
|
+
system_template = (
|
|
68
|
+
"You are an expert software engineer and git commit message generator.\n"
|
|
69
|
+
"Your task is to generate a clean, concise commit message following the "
|
|
70
|
+
"Conventional Commits specification.\n\n"
|
|
71
|
+
"Values from middleware:\n"
|
|
72
|
+
"{summary_section}\n"
|
|
73
|
+
"{todo_section}\n\n"
|
|
74
|
+
"Input context:\n"
|
|
75
|
+
"{explanation_section}\n\n"
|
|
76
|
+
"Existing Code Changes (Diff):\n"
|
|
77
|
+
"{diff}\n\n"
|
|
78
|
+
"Instructions:\n"
|
|
79
|
+
"1. Use the format: <type>(<scope>): <subject>\n"
|
|
80
|
+
"2. Keep the subject line under 50 characters if possible.\n"
|
|
81
|
+
"3. If there are multiple changes, provide a bulleted body.\n"
|
|
82
|
+
"4. If TODOs were detected, mention them in the footer or body as "
|
|
83
|
+
"appropriate.\n"
|
|
84
|
+
"5. If an explanation is provided, prioritize it.\n"
|
|
85
|
+
)
|
|
86
|
+
prompt = ChatPromptTemplate.from_template(system_template)
|
|
87
|
+
|
|
88
|
+
# 3. Helper to format the prompt inputs from state
|
|
89
|
+
def format_inputs(state: CommitState) -> Dict[str, Any]:
|
|
90
|
+
summary = state.get("summary")
|
|
91
|
+
todos = state.get("todos")
|
|
92
|
+
explanation = state.get("explanation")
|
|
93
|
+
|
|
94
|
+
summary_section = f"Summary of changes:\n{summary}\n" if summary else ""
|
|
95
|
+
|
|
96
|
+
todo_section = ""
|
|
97
|
+
if todos:
|
|
98
|
+
todo_section = (
|
|
99
|
+
"Detected TODOs in this diff:\n"
|
|
100
|
+
+ "\n".join(f"- {t}" for t in todos)
|
|
101
|
+
+ "\n"
|
|
102
|
+
)
|
|
103
|
+
|
|
104
|
+
explanation_section = ""
|
|
105
|
+
if explanation:
|
|
106
|
+
explanation_section = f"User Explanation:\n{explanation}\n"
|
|
107
|
+
|
|
108
|
+
return {
|
|
109
|
+
"diff": state["diff"],
|
|
110
|
+
"summary_section": summary_section,
|
|
111
|
+
"todo_section": todo_section,
|
|
112
|
+
"explanation_section": explanation_section,
|
|
113
|
+
}
|
|
114
|
+
|
|
115
|
+
# 4. Construct the Pipeline
|
|
116
|
+
# Parallel step to run middlewares
|
|
117
|
+
# (conceptually, though here we chain them or use RunnablePassthrough)
|
|
118
|
+
# Since middlewares modify state, we can chain them:
|
|
119
|
+
|
|
120
|
+
middleware_chain: Runnable = (
|
|
121
|
+
RunnablePassthrough()
|
|
122
|
+
| RunnableLambda(todo_scanner)
|
|
123
|
+
| RunnableLambda(summarizer)
|
|
124
|
+
)
|
|
125
|
+
|
|
126
|
+
# Final generation chain
|
|
127
|
+
generation_chain = (
|
|
128
|
+
middleware_chain
|
|
129
|
+
| RunnableLambda(format_inputs)
|
|
130
|
+
| prompt
|
|
131
|
+
| llm
|
|
132
|
+
| StrOutputParser()
|
|
133
|
+
)
|
|
134
|
+
|
|
135
|
+
return generation_chain
|