knowcode 0.1.3

Structural cognition engine for code repositories

KnowCode

KnowCode is a structural cognition engine designed specifically for Agentic IDEs (like Cursor, Gemini, and Copilot). It bridges the gap between physical repository code, deterministic structural snapshots, and semantic architectural knowledge, allowing AI agents and human developers to seamlessly inspect, track, and maintain codebase architectures.

---

Table of Contents

• Overview
• Key Features
• Installation
• Usage: The Agentic Workflow
• Ecosystem Layout
• CLI Reference & Manual Execution
• License

---

Overview

KnowCode completely flips the traditional documentation model. Instead of forcing developers to write specs before they code, KnowCode uses an AI agent to track your intent during a development session. When you sync your physical code changes, the AI agent is automatically invoked to synthesize your intent against the deterministic reality of the codebase, ensuring your architectural knowledge naturally accumulates as a byproduct of development.

KnowCode partitions codebase understanding into three clean domains:

1. Physical Reality: The actual source files inside the repository.
2. Structural Truth: Deterministic representation of files, components, and code symbols parsed via abstract syntax trees (ASTs).
3. Semantic Knowledge: Permanent architectural rules, decisions, and constraints synthesized by your AI agent.

---

Key Features

• Native Agent Integration: Built-in slash commands (/knowcode, /know-sync) that plug directly into Agentic IDEs without configuration.
• Multi-Language AST Parsing: Built on top of tree-sitter, with out-of-the-box support for Python, JavaScript, and TypeScript.
• Deterministic State & Revision Tracking: Tracks codebase changes across sequential revisions (S-001, S-002, etc.).
• Diff & Report Generation: Automatically computes additions, deletions, line-boundary modifications, and maps changes to their top-level affected components.

---

Installation

Prerequisites

• Python 3.13+
• uv Package Manager

Installation

You can install Knowcode directly via pip or npm.

Using pip:

pip install knowcode

Using npm:

npm install -g knowcode

Development Setup

If you want to contribute or modify the code, clone the repository and install it in editable mode:

git clone https://github.com/knowiki/knowcode.git
cd knowcode
uv sync
uv pip install -e .

---

Usage: The Agentic Workflow

KnowCode is designed to be driven by you and your AI agent in tandem.

1. Initialize the Repository

Run the initialization command in your target repository terminal:

know .

This scaffolds the .agent/ and .knowcode/ directories, computes your first AST snapshot, and drops knowcode.md at your root.

Note: To prevent polluting your repository, know . automatically adds .knowcode/, .agent/, and knowcode.md to your .gitignore. If you want to share your synthesized semantic knowledge with your team, you can selectively un-ignore .knowcode/knowledge/.

2. Activate the Agent

In your AI chat (Cursor, Copilot, Gemini), simply type:

/knowcode

The agent will instantly read knowcode.md, lock into the KnowCode governance rules, and begin silently tracking your intent in .agent/memory/active_context.md while you code.

3. Sync & Synthesize

After you finish a coding session or fix a bug, instruct the agent to synchronize by typing:

/know-sync

The agent will orchestrate the entire lifecycle:

1. Run know sync to compute structural AST diffs.
2. Read the generated structural report (R-XXX.md).
3. Reconcile the structural reality against your tracked intent.
4. Distribute the extracted knowledge into the 5 semantic buckets (architecture, decisions, etc.).
5. Run know sync-semantic to commit the knowledge and bump the revision.

---

Ecosystem Layout

When KnowCode is initialized in a repository via know ., it generates a governed agent architecture:

/
├── knowcode.md                # The Agent Ignition Switch (Read by the AI)
├── .agent/                    # The Semantic Governance & Memory Layer
│   ├── memory/
│   │   ├── active_context.md  # The AI's short-term intent tracking buffer
│   │   └── previous_context.md
│   ├── skills/
│   │   ├── track_intent.md    # Background skill for tracking architectural intent
│   │   └── synthesize_knowledge.md
│   └── workflows/
│       ├── sync_reconciliation.md # The /know-sync workflow
│       └── ingest_legacy.md       # The /know-ingest workflow
│
└── .knowcode/                 # The Deterministic Knowledge Artifact
    ├── state.yaml             # Authoritative state registry (managed by the Engine)
    ├── structure/             # Deterministic AST snapshots (e.g. S-001.json)
    ├── reports/               # Change and diff analysis reports (e.g. R-001.md)
    └── knowledge/             # Permanent semantic knowledge (AI synthesized)
        ├── architecture/
        ├── components/
        ├── constraints/
        ├── conventions/
        ├── decisions/
        └── raw/               # Inbox for legacy or unstructured documentation

---

CLI Reference & Manual Execution

While the Agentic Workflow relies on slash commands to orchestrate operations, developers can bypass the AI entirely and execute any step manually using the know CLI. This is useful for CI/CD pipelines, programmatic integration, or direct inspection.

• know .: Initializes Knowcode in the current directory.
• know status: Displays current structural and semantic revisions.
• know sync: Computes AST diffs, generates a report, and rolls over the memory buffer.
• know sync-semantic: Bumps the semantic revision after the AI (or a human) finishes synthesis.
• know ingest-semantic .: Wipes the raw documentation inbox and bumps the revision.

---

License

This project is licensed under the MIT License. See the LICENSE file for details.