💠 Silice AI Native Codebase Mapper

Silice Codebase Mapper is a local, AI-powered documentation and dependency mapping suite. It transforms a messy codebase into a Structured Knowledge Graph specifically optimized for ingestion by Large Language Models (LLMs) via Ollama.

By breaking down code into atomic JSON "neurons" and a central "synapse" (index.json), it allows AI agents to understand your project's architecture, intent, and impact without exceeding context windows.

---

🚀 Features

• Atomic Ingestion: Generates a unique JSON map for every source file.
• AI Enrichment: Uses Ollama + Instructor to extract logic summaries and conceptual relations.
• Static & Dynamic Analysis: Combines Python's ast for reliable structure with LLMs for "intent" analysis.
• Impact Analysis: Query-tool to see how changing one function ripples through the graph.
• AI Bridge: A RAG (Retrieval-Augmented Generation) chat interface to talk to your code locally.

---

🛠️ Installation

1. Requirement: Ensure you have Ollama installed and running.
2. Model: Pull a compatible model (Llama 3 is recommended):

ollama pull llama3

3. Dependencies:

pip install ollama instructor pydantic

---

📂 The Core "Trinity"

1. The Mapper (silice_file_mapper.py)

The "Worker." It crawls your directories, performs static analysis, and asks the AI to document the logic of every file.

• Output: silice_output/*.json and index.json.
• Usage:

python silice_file_mapper.py ./src ./lib

2. The Query Tool (silice_query.py)

The "Analyst." Use this to traverse the graph and check dependencies.

• Usage:

# See what depends on a specific class or function
python silice_query.py --impact AuthService

# Get a quick AI-generated summary of a file
python silice_query.py --info models/user.py

3. The Bridge (silice_bridge.py)

The "Interface." An interactive chat that uses the generated JSON maps as a local brain to answer complex architectural questions.

• Usage:

python silice_bridge.py

---

📊 Data Schema

The system uses Pydantic to enforce strict adherence to the Silice Protocol. Every file is mapped into a FileNode:

Field	Description
file_path	Absolute location for reference.
functions	List of names, signatures, and AI-summarized logic.
dependencies	Graph-edges (imports, calls, inheritance).
summary	A high-level overview of the file's "Reason for Being."

---

🛠️ Workflow

1. Scan: Run the Mapper to build your JSON library.
2. Index: The index.json creates a master map of all file relations.
3. Consult: Use the Bridge or Query tool to navigate the project without ever reading the raw source code.

---

⚠️ Notes

• Context Control: This suite is designed to avoid "Context Bloat." The Bridge only feeds the AI the specific JSON maps relevant to your current question.
• Performance: For faster mapping, you can edit the script to use model="phi3" or model="mistral" in the Ollama client.

---

Made with ❤️ and ☕ by the Plantacerium

⭐ Star us on GitHub ⭐