Perfect AI development, like perfect paella: quality ingredients, structure, and expertise.
⭐ If you find PAELLADOC useful, please consider starring the repo! ⭐
Version 0.3.7: Hotfix release restoring core project CRUD tools inadvertently omitted in v0.3.6 build. Check the CHANGELOG for details!
"In the AI era, context isn't supplementary to code—it's the primary creation."
PAELLADOC is an AI-First Development framework that implements the 5 Philosophical Principles of AI-First Development, transforming how we create software in the age of AI.
PAELLADOC implements Anthropic's Model Context Protocol (MCP) (see Anthropic's news). This protocol provides a structured way for Large Language Models (LLMs) to interact with external tools and context, enabling more sophisticated capabilities.
By implementing MCP, PAELLADOC allows LLMs to leverage its specific AI-First development tools and workflows directly through this standard. This approach facilitates functionalities similar to Tool Use or Function Calling seen in other platforms, but specifically utilizes the Anthropic MCP standard for interaction.
Traditional development treats documentation as an afterthought. AI-First Development inverts this paradigm:
- Context becomes the primary artifact
- Code becomes its manifestation
- Knowledge evolves alongside systems
- Decisions preserve their philosophical context
- Human-AI collaboration is seamless
# Traditional Way
write_code() -> document()
# PAELLADOC Way
create_context() -> manifest_as_code()- Every artifact has a UUID for perfect traceability
- Context is versioned alongside code
- Knowledge graphs capture relationships
- Intent is preserved at every step
graph TD
A[Business Intent] --> B[Context Creation]
B --> C[Architecture Manifestation]
C --> D[Code Generation]
D --> E[Living Documentation]
- Architecture flows from intent, not implementation
- Every decision captures its philosophical context
- Systems adapt to evolving purpose
# Knowledge evolves with your system
paella continue my-project- Project memory tracks evolution of understanding
- Documentation updates automatically with changes
- Context remains fresh and relevant
- Knowledge graphs show relationships
# Not just code generation, but true collaboration
with paelladoc.context() as ctx:
ctx.understand_intent()
ctx.propose_solutions()
ctx.implement_with_human()- Natural language conversations
- Intent preservation
- Contextual awareness
- Seamless collaboration
decision:
id: uuid-123
intent: "Why we chose this path"
context: "What we knew at the time"
alternatives: "What we considered"
implications: "Future impact"- Every decision preserves its context
- Future developers understand the "why"
- Changes respect historical context
- Intent remains clear
PAELLADOC is a Python application and should be installed in its own dedicated Python virtual environment. This keeps its dependencies separate and avoids conflicts. You'll need one PAELLADOC environment, regardless of how many different projects (Python, JS, Ruby, etc.) you plan to document.
(Requires Python 3.12 or later)
To install PAELLADOC for Claude Desktop automatically via Smithery:
npx -y @smithery/cli install @jlcases/paelladoc --client claudeFirst, choose a permanent location for this environment. Your home directory is often a good choice.
# Navigate to where you want to store the environment (e.g., your home directory)
# cd ~ # Uncomment and run if you want it in your home directory
# Create the virtual environment (using python3.12 or your installed 3.12+ version)
# We'll name the folder '.paelladoc_venv' (starting with a dot makes it hidden)
python3.12 -m venv .paelladoc_venv
# Activate the environment
# (The command depends on your shell. Use ONE of the following)
# For Bash/Zsh:
source .paelladoc_venv/bin/activate
# For Fish:
# source .paelladoc_venv/bin/activate.fish
# For Powershell (Windows):
# .\.paelladoc_venv\Scripts\activate.ps1 (You should see (.paelladoc_venv) at the beginning of your terminal prompt now)
# Make sure your (.paelladoc_venv) prompt is visible before running pip
pip install paelladocPAELLADOC needs to know where to store its memory database (memory.db). There are two main ways to configure this:
Option 1: Environment Variable (Less Reliable for LLM Integration)
You can set the PAELLADOC_DB_PATH environment variable. This works well if you run PAELLADOC directly from your terminal.
# Example: Set the variable in your current terminal session
export PAELLADOC_DB_PATH="$HOME/.paelladoc/memory.db"
# Optional: Add the export line to your shell's startup file
# (.bashrc, .zshrc, etc.) for it to persist across sessions.Important: When PAELLADOC is run by an LLM tool (like Cursor via MCP), it might not inherit environment variables set this way. Therefore, this method is less reliable for LLM integration.
Option 2: MCP Configuration (Recommended for LLM Integration)
The most reliable way to ensure your LLM tool uses the correct database path is to configure it directly within the tool's MCP JSON file (.cursor/mcp.json for Cursor). This injects the variable directly into the server process launched by the LLM.
See the examples in the next section.
Now, tell your LLM tool (like Cursor) how to find and run PAELLADOC.
Key Information Needed:
- Full Path to Python Executable: Absolute path to
pythoninside your.paelladoc_venv.
Edit your .cursor/mcp.json file. Add a server configuration for PAELLADOC. Here's a typical example:
{
"mcpServers": {
"Paelladoc": {
"command": "/absolute/path/to/.paelladoc_venv/bin/python",
"args": [
"-m",
"paelladoc.ports.input.mcp_server_adapter",
"--stdio"
],
"cwd": "/path/to/your/project/directory", // Optional: Set working directory
"env": {
// Recommended for local dev: Use a DB in your project folder
"PAELLADOC_DB_PATH": "/path/to/your/project/directory/paelladoc_memory.db",
// Optional: Add src to PYTHONPATH if needed for local development imports
"PYTHONPATH": "/path/to/your/project/directory/src:/path/to/your/project/directory"
},
"disabled": false
}
},
"mcp.timeout": 120000
}Important Notes:
- The
commandpath must be the absolute path to the Python executable inside your.paelladoc_venv(created in Step 1). Replace/absolute/path/to/with the actual path on your system (e.g.,/Users/your_username/). - Database Path:
- By default (if
PAELLADOC_DB_PATHis not set inenv), PAELLADOC uses~/.paelladoc/memory.db. - For local development where you might want the database alongside your project code, setting
PAELLADOC_DB_PATHin theenvsection (as shown in the example) is the recommended and most reliable approach. Replace/path/to/your/project/directory/with the actual path to your project.
- By default (if
- Working Directory (
cwd): Setting this to your project directory can be helpful but is often optional. - PYTHONPATH: Setting this in
envmight be necessary if you are doing local development on PAELLADOC itself and need the server to find your source code.
Once connected, your LLM will have access to all PAELLADOC commands:
PAELLA: Start new documentation projectsCONTINUE: Continue existing documentationVERIFY: Verify documentation coverageGENERATE: Generate documentation or code
The LLM will handle all the complexity - you just need to express your intent in natural language!
- PyPI Version (Stable): The versions published on PyPI (
pip install paelladoc) are stable releases recommended for general use. - GitHub Repository (Development): The
mainbranch (and other branches) on the GitHub repository contains the latest development code. This version may include new features or changes that are not yet fully tested and should be considered unstable. Use this version if you want to try out cutting-edge features or contribute to development.
Note on Current Development: Active development is currently focused internally on delivering an MVP with significant new capabilities. While the PyPI version remains stable, expect major advancements in future releases as we work towards this goal in a more private setting for now.
-
Ensure PAELLADOC is installed (
pip install paelladoc) and configured in your LLM's tool/MCP settings (see examples above). -
Start interacting with PAELLADOC through your LLM by issuing a command. The primary command to initiate a new project or list existing ones is
PAELLA.- In Cursor or a similar chat interface, simply type:
PAELLA - Alternatively, you can instruct the LLM more explicitly:
Use PAELLADOC to start documenting a new project.Tell PAELLADOC I want to create documentation.
- In Cursor or a similar chat interface, simply type:
-
Follow the LLM's lead: PAELLADOC (via the LLM) will then guide you through the process interactively, asking for project details, template choices, etc.
This version provides the following core commands, exposed via MCP for interaction with your LLM:
-
ping:- Description: Basic health check to confirm the server is running and responsive.
- Arguments: None (or optional
random_string). - Returns:
{ "status": "ok", "message": "pong" }.
-
paella_init:- Description: Initializes a new PAELLADOC project, creating the necessary structure and initial memory file.
- <
83EB
strong>Arguments:
base_path(str),documentation_language(str, e.g., "es-ES"),interaction_language(str, e.g., "en-US"),new_project_name(str). - Returns: Dictionary confirming project creation status, name, and path.
-
paella_list:- Description: Lists the names of all existing PAELLADOC projects found in the memory database.
- Arguments: None.
- Returns: Dictionary containing a list of project names (
projects).
-
paella_select:- Description: Selects an existing PAELLADOC project to work on (loads its memory).
- Arguments:
project_name(str). - Returns: Dictionary confirming project selection and its base path.
-
core_continue:- Description: Continues work on a previously selected project, loading its memory and suggesting next steps (basic implementation).
- Arguments:
project_name(str). - Returns: Dictionary with project status and suggested next step.
-
core_help:- Description: Provides help information about available commands (basic stub implementation).
- Arguments: None (future: specific command).
- Returns: Placeholder success message.
-
core_list_projects:- Description: (Likely redundant with
paella_list) Lists the names of existing PAELLADOC projects. - Arguments:
db_path(str, optional, for testing). - Returns: Dictionary containing a list of project names (
projects).
- Description: (Likely redundant with
-
core_verification:- Description: Checks documentation quality and completeness (basic stub implementation).
- Arguments: None.
- Returns: Placeholder success message.
Based on the Unified Roadmap, future versions aim to include:
- Full interactive documentation generation flows (
GENERATE-DOC). - Code analysis and context generation (
GENERATE_CONTEXT). - Automatic code generation from documentation (
code_generation). - Management of coding styles and Git workflows (
styles.coding_styles,styles.git_workflows). - Project memory commands for decisions, issues, achievements (
DECISION,ISSUE,ACHIEVEMENT). - And much more, aligning with the MECE taxonomy and A2A capabilities.
Our AI-First taxonomy ensures complete context preservation: