<table>
<tr>
<td align="center" style="border: none; padding: 0;">
<img src="https://raw.githubusercontent.com/scribe-works/prompt-scribe/main/docs/assets/logo.png" alt="The Scribe Works Logo" width="200"/>
</td>
<td style="border: none; padding: 0; vertical-align: middle;">
<h1>Prompt Scribe</h1>
<p>A powerful, template-based prompt composer for crafting and managing complex instructions for LLMs.</p>
</td>
</tr>
</table>
---
## The Problem
Managing large, multi-part prompts for Large Language Models can be messy. You often combine a persona, rules, context from different files, and a specific query. Doing this manually by copy-pasting is tedious and error-prone.
## The Solution
**Prompt Scribe** automates this process. It uses a simple YAML configuration and the powerful Jinja2 templating engine to compose your final prompts from various source files, ready to be used in any LLM chat or API.
## Key Features
- **Template-Based:** Use Jinja2 templates for maximum flexibility.
- **Modular:** Split your prompts into reusable parts (personas, includes, rules).
- **Safe Initialization:** Creates a dedicated `.prompt_scribe/` directory to avoid cluttering your project root.
- **Watch Mode:** Automatically recompose prompts when source files change.
- **Beautiful CLI:** A clean, helpful command-line interface.
## Installation
```bash
pip install prompt-scribe
```
## Quick Start
1. **Initialize a new project:**
Navigate to your project directory and run:
```bash
prompt-scribe init
```
This will create a self-contained `.prompt_scribe/` directory with a default structure:
```
.
└── .prompt_scribe/
├── composed_prompts/
├── includes/
│ └── development-rules.md
├── personas/
│ └── code-reviewer.md
├── templates/
│ └── master.jinja2
└── prompts.yml
```
2. **Compose your prompts:**
Run the compose command from your project root:
```bash
prompt-scribe compose
```
This will read `.prompt_scribe/prompts.yml`, process the `example-code-reviewer` agent, and generate the final prompt in `.prompt_scribe/composed_prompts/`.
3. **Use the output:**
Open the generated file, copy the content, and paste it into your LLM chat interface.
## Usage
### Composing Prompts
- **Compose all agents**:
```bash
prompt-scribe compose
```
- **Compose specific agents**:
```bash
prompt-scribe compose agent-one agent-two
```
- **Watch for changes** and automatically recompose:
```bash
prompt-scribe compose --watch
```
## Configuration (`.prompt_scribe/prompts.yml`)
The `prompts.yml` file is the heart of your project. It resolves all paths relative to its own location.
```yaml
# Global settings for all agents
settings:
personas_dir: "personas"
includes_dir: "includes"
templates_dir: "templates"
output_dir: "composed_prompts"
# Map of agents to be composed
agents:
example-code-reviewer:
# Jinja2 template to use for rendering
template: master.jinja2
# The final output file name
output_file: example_code_reviewer.md
# The persona section, available as `persona` in the template
persona:
file: personas/code-reviewer.md
# A list of content sections, available as `sections` in the template
sections:
- title: "📜 Key Development Rules"
prologue: "These are the mandatory rules and principles for this project."
file: includes/development-rules.md
- title: "📄 Code Snippet for Review"
# You can also include content directly
content: |
```python
def hello_world():
print("Hello, Scribe!")
```
```
## License
This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.
Raw data
{
"_id": null,
"home_page": null,
"name": "prompt-scribe",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.14",
"maintainer_email": null,
"keywords": "llm, ai, prompt-engineering, developer-tools, jinja2, cli",
"author": "Plyanter Viktor",
"author_email": null,
"download_url": "https://files.pythonhosted.org/packages/f2/90/d53909b54911fc259974cd4bb2d375ea4468c1e7170653f4885f8484adb2/prompt_scribe-0.1.0.tar.gz",
"platform": null,
"description": "<table>\n <tr>\n <td align=\"center\" style=\"border: none; padding: 0;\">\n <img src=\"https://raw.githubusercontent.com/scribe-works/prompt-scribe/main/docs/assets/logo.png\" alt=\"The Scribe Works Logo\" width=\"200\"/>\n </td>\n <td style=\"border: none; padding: 0; vertical-align: middle;\">\n <h1>Prompt Scribe</h1>\n <p>A powerful, template-based prompt composer for crafting and managing complex instructions for LLMs.</p>\n </td>\n </tr>\n</table>\n\n---\n\n## The Problem\n\nManaging large, multi-part prompts for Large Language Models can be messy. You often combine a persona, rules, context from different files, and a specific query. Doing this manually by copy-pasting is tedious and error-prone.\n\n## The Solution\n\n**Prompt Scribe** automates this process. It uses a simple YAML configuration and the powerful Jinja2 templating engine to compose your final prompts from various source files, ready to be used in any LLM chat or API.\n\n## Key Features\n\n- **Template-Based:** Use Jinja2 templates for maximum flexibility.\n- **Modular:** Split your prompts into reusable parts (personas, includes, rules).\n- **Safe Initialization:** Creates a dedicated `.prompt_scribe/` directory to avoid cluttering your project root.\n- **Watch Mode:** Automatically recompose prompts when source files change.\n- **Beautiful CLI:** A clean, helpful command-line interface.\n\n## Installation\n\n```bash\npip install prompt-scribe\n```\n\n## Quick Start\n\n1. **Initialize a new project:**\n Navigate to your project directory and run:\n ```bash\n prompt-scribe init\n ```\n This will create a self-contained `.prompt_scribe/` directory with a default structure:\n ```\n .\n \u2514\u2500\u2500 .prompt_scribe/\n \u251c\u2500\u2500 composed_prompts/\n \u251c\u2500\u2500 includes/\n \u2502 \u2514\u2500\u2500 development-rules.md\n \u251c\u2500\u2500 personas/\n \u2502 \u2514\u2500\u2500 code-reviewer.md\n \u251c\u2500\u2500 templates/\n \u2502 \u2514\u2500\u2500 master.jinja2\n \u2514\u2500\u2500 prompts.yml\n ```\n\n2. **Compose your prompts:**\n Run the compose command from your project root:\n ```bash\n prompt-scribe compose\n ```\n This will read `.prompt_scribe/prompts.yml`, process the `example-code-reviewer` agent, and generate the final prompt in `.prompt_scribe/composed_prompts/`.\n\n3. **Use the output:**\n Open the generated file, copy the content, and paste it into your LLM chat interface.\n\n## Usage\n\n### Composing Prompts\n\n- **Compose all agents**:\n ```bash\n prompt-scribe compose\n ```\n- **Compose specific agents**:\n ```bash\n prompt-scribe compose agent-one agent-two\n ```\n- **Watch for changes** and automatically recompose:\n ```bash\n prompt-scribe compose --watch\n ```\n\n## Configuration (`.prompt_scribe/prompts.yml`)\n\nThe `prompts.yml` file is the heart of your project. It resolves all paths relative to its own location.\n\n```yaml\n# Global settings for all agents\nsettings:\n personas_dir: \"personas\"\n includes_dir: \"includes\"\n templates_dir: \"templates\"\n output_dir: \"composed_prompts\"\n\n# Map of agents to be composed\nagents:\n example-code-reviewer:\n # Jinja2 template to use for rendering\n template: master.jinja2\n # The final output file name\n output_file: example_code_reviewer.md\n # The persona section, available as `persona` in the template\n persona:\n file: personas/code-reviewer.md\n # A list of content sections, available as `sections` in the template\n sections:\n - title: \"\ud83d\udcdc Key Development Rules\"\n prologue: \"These are the mandatory rules and principles for this project.\"\n file: includes/development-rules.md\n - title: \"\ud83d\udcc4 Code Snippet for Review\"\n # You can also include content directly\n content: |\n ```python\n def hello_world():\n print(\"Hello, Scribe!\")\n ```\n```\n\n## License\n\nThis project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.",
"bugtrack_url": null,
"license": null,
"summary": "A powerful, template-based prompt composer for crafting and managing complex instructions for LLMs.",
"version": "0.1.0",
"project_urls": null,
"split_keywords": [
"llm",
" ai",
" prompt-engineering",
" developer-tools",
" jinja2",
" cli"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "603da1458be887dae7954ef0b20c68fc0c3ad6e716645e77585c86b4183d1fdf",
"md5": "fd5b206b8c30f8a0dbcabd7eb935efaf",
"sha256": "813bba9cc850b8869a236badce099503aed3e480dea82e18768cf28e52164b3b"
},
"downloads": -1,
"filename": "prompt_scribe-0.1.0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "fd5b206b8c30f8a0dbcabd7eb935efaf",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.14",
"size": 12701,
"upload_time": "2025-10-29T04:44:20",
"upload_time_iso_8601": "2025-10-29T04:44:20.699643Z",
"url": "https://files.pythonhosted.org/packages/60/3d/a1458be887dae7954ef0b20c68fc0c3ad6e716645e77585c86b4183d1fdf/prompt_scribe-0.1.0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "f290d53909b54911fc259974cd4bb2d375ea4468c1e7170653f4885f8484adb2",
"md5": "f18153e92965eeb85511b3e2429b2a02",
"sha256": "1e8b4e0191e8ba638a3f547fc5e84079047d4f8f1c7c71ad00c3f4ef4d2c2839"
},
"downloads": -1,
"filename": "prompt_scribe-0.1.0.tar.gz",
"has_sig": false,
"md5_digest": "f18153e92965eeb85511b3e2429b2a02",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.14",
"size": 10684,
"upload_time": "2025-10-29T04:44:22",
"upload_time_iso_8601": "2025-10-29T04:44:22.151269Z",
"url": "https://files.pythonhosted.org/packages/f2/90/d53909b54911fc259974cd4bb2d375ea4468c1e7170653f4885f8484adb2/prompt_scribe-0.1.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-10-29 04:44:22",
"github": false,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"lcname": "prompt-scribe"
}
JSON