feat(commit-prompter): Add new configuration option in config.json

- Introduce a new `model` field in the `config.json` file - Parse the `model` configuration in the `gpt-commit-prompter` script - Adapt the `config` section in the README.md file to reflect the changes - Move the `config` section from the README.md to the config.json file - Perform code cleanup and refactorings
5n00py · Dec 13, 2023 · c100550 · c100550
1 parent 4307ada
commit c100550
Show file tree

Hide file tree

Showing 3 changed files with 42 additions and 49 deletions.
diff --git a/README.md b/README.md
@@ -36,7 +36,7 @@ possible options, run `gc-smart --help`.
 - Staged changes in the current repo to be commited
 - `tmp_commit_msg.txt` has to be configured as commit template (see below)
 
-## Installation and Setup
+## Installation
 
 1. Clone this repository to your local machine using git:
 
@@ -119,6 +119,36 @@ possible options, run `gc-smart --help`.
     ```
     (Don't forget to replace `/path/to/gc-smart` with the actual path to the script.
 
+## Configuration
+
+The `gpt-commit-prompter` used in `gc-smart` allows configuration through the
+[config.json](config.json) file. This flexibility ensures that the tool can be
+easily adjusted to specific needs for generating commit messages.
+
+### Commit Message Style
+
+By default, the script uses the `imperative` style for commit messages. You can
+choose a different style by using the `-s` or `--style` option when running the
+script. The available styles, as defined in the `config.json` file, are:
+
+- `imperative`: Generates a message in the imperative mood with a conventional
+  title and bullet points.
+- `simple`: Produces a concise, one-line commit message.
+- `detailed`: Creates a verbose commit message, elaborating on the changes.
+- `conventional`: Follows the Conventional Commits specification for the commit
+  message format.
+
+You can view and customize the styles by editing their corresponding entries in
+the `config.json` file.
+
+### GPT Model Configuration
+
+In addition to style customization, the `config.json` file allows you to
+configure the GPT model used by the script. This feature enables you to select
+the most appropriate model version for your needs, ensuring optimal performance
+and relevance of the generated commit messages. To change the model, simply
+update the `model` section in the `config.json` file.
+
 ## Note on Commit Template Handling from gc-smart Version 0.4.0
 
 Starting from version 0.4.0 of `gc-smart`, there is no longer a need to
@@ -193,26 +223,6 @@ message will appear as a template in your default Git editor. You can then
 finalize the message or make any necessary modifications before committing the
 changes to your repository.
 
-## Commit Message Style
-
-The `gpt-commit-prompter` script uses the `imperative` style as its default for
-the commit messages it generates. However, you have the flexibility to
-configure the style through the `-s` or `--style` option.
-
-Available styles are:
-
-- `imperative`: Generates a message in the imperative mood and uses
-  conventional title with bullets.
-- `simple`: Generates a concise, straightforward commit message in one line.
-- `detailed`: Generates a more verbose commit message, detailing the changes.
-- `conventional`: Generates a commit message following the conventional commits
-  specification.
-
-These styles can also be used in combination with the `gc-smart` script.
-
-You can see how the styles are configured and further customize them by
-checking the `config.json` file.
-
 ## License
 
 This project is licensed under the MIT License - see the LICENSE.md file for

diff --git a/config.json b/config.json
@@ -1,4 +1,7 @@
 {
+    "model": {
+        "name": "gpt-3.5-turbo"
+    },
     "style": {
         "imperative": {
             "system_prompt": "You are a helpful assistant that generates meaningful commit messages based on a diff file. The commit message should include a title in the conventional commits style with possible types feat, fix, chore, docs, style, refactor, test or build and a scope and description. The details should be summarized using bullets '-' and imperatives. Each change should be on a new line in an imperative style. Please ensure that no line exceeds 80 characters in length."

diff --git a/python/gpt-commit-prompter b/python/gpt-commit-prompter
@@ -91,33 +91,6 @@ def get_changes(args):
     else:
         return args.changes
 
-# def generate_commit_message(changes, instruction, style="imperative"):
-#     config = load_config()
-#
-#     # Ensure the provided style is in the configuration
-#     if style not in config["style"]:
-#         raise ValueError(f"Style '{style}' not found in configuration.")
-#
-#     system_prompt = config["style"][style]["system_prompt"]
-#
-#     # Constructing the prompt for the AI
-#     prompt = f"{system_prompt}\n\nChanges:\n{changes}\n\n"
-#     if instruction:
-#         prompt += f"Instruction: {instruction}\n"
-#
-#     # Call the OpenAI API with the constructed prompt
-#     response = openai.Completion.create(
-#         engine="text-davinci-003",  # or another model of your choice
-#         prompt=prompt,
-#         max_tokens=150,  # Adjust as needed
-#         n=1,  # Number of completions to generate
-#         stop=None  # String where the API will stop generating further content
-#     )
-#
-#     # Extracting the commit message from the response
-#     commit_message = response.choices[0].text.strip()
-#     return commit_message
-
 def generate_commit_message(changes, instruction, style="imperative"):
     config = load_config()
 
@@ -135,13 +108,20 @@ def generate_commit_message(changes, instruction, style="imperative"):
     # Instantiate the OpenAI client
     client = OpenAI()
 
+    # Get the model name from the configuration
+    model_config = config.get("model")
+    if not model_config or "name" not in model_config:
+        raise ValueError("Model configuration not found in config.json")
+
+    model_name = model_config["name"]
+
     # Call the OpenAI API with the constructed prompt
     response = client.chat.completions.create(
         messages=[
             {"role": "system", "content": system_prompt},
             {"role": "user", "content": prompt}
         ],
-        model="gpt-3.5-turbo", 
+        model=model_name, 
     )
 
     # Extracting the commit message from the response