Status: Active Development / Prototype
Mapping the hidden
connections inside code.
Synapse is an experimental semantic "second brain" for software repositories. It scans workspaces, parses Abstract Syntax Trees via Tree-sitter, extracts symbols, tracks cross-file imports, and maps complex projects into an interactive neural graph.
// 01 · The Central Question
“What if a software codebase could be explored like a neural map instead of a flat folder tree?”
As software projects grow, relationships between modules, functions, classes, and types become invisible behind directory hierarchies. Developers lose hours tracing imports or understanding the blast radius of changes. Synapse was started to turn codebases into transparent, semantic knowledge graphs.
// 02 · Honest Development Breakdown
Current vs. Exploring vs. Future Direction
Current / Prototyped
Working Capabilities
- Recursive workspace directory scanning via CLI
- Tree-sitter parser manager and symbol extraction
- Initial AST parsers for Python, JS, TS, Rust, Dart
- Import tracking & SQLite symbol database models
- Workspace statistics CLI command (
synapse.cli info)
Actively Exploring
In Active Research
- Small Language Model (SLM) embeddings with small Qwen models
- Semantic concept clustering beyond exact symbol names
- Interactive Three.js 3D node constellation interface
- Bidirectional dependency graph traversal
- Incremental indexing cache on git commits
Future Ambition
Long-Term Vision
- Natural language code queries ("where is payment auth routed?")
- Semantic code search that understands intent
- Integration with Zeta for voice-driven codebase navigation
- Spatial code architecture visualization in IDEs
// 03 · Command-Line Architecture
CLI Tooling & Modules
Synapse is developed as a modular Python package with clean separation between the scanner, parsers, and graph database storage.
$ python -m synapse.cli scan "D:\projects\berry-engine"
✓ Discovered 148 source files [Python, TypeScript, Rust]
✓ Tree-sitter extracted 842 symbols, 319 imports into SQLite
✓ Generated graph representation: 842 nodes, 1,204 edges
$ python -m synapse.cli info
Workspace: berry-engine | Symbol density: 5.68/file | Cache: valid
$ python -m synapse.cli open
» Preparing Three.js constellation renderer on localhost:4173...
// 04 · Project Timeline
Development Milestones
2026
Tree-sitter parser manager integration and SQLite symbol relational models.
2026
Investigating small Qwen model quantization for semantic cluster tagging on 8GB workstation.
2025
Initial architecture conception for code representation beyond folder trees.