Constellation

What is Constellation?

Constellation turns your codebase into a queryable knowledge graph, giving AI assistants the structural understanding they need to reason about real software — not just plain text. It integrates with AI assistants via a single code_intel tool that exposes a rich JavaScript API for code intelligence queries.

How to use Constellation?

Invoke the code_intel tool with JavaScript API calls such as api.searchSymbols, api.impactAnalysis, and api.getDependents. The tool composes queries like code, allowing the assistant to ask about symbol definitions, usage, impact, and dependencies.

Key features of Constellation

• Turns codebase into a semantic graph, not a grep loop
• Single code_intel tool with JavaScript API
• Multi-language: TypeScript, JavaScript, Python (more coming)
• Multi-project support with per‑project git root resolution
• Fast: parallel API calls, inline source snippets
• Safe: hardened sandbox with isolated child process and memory limits
• Privacy by design: source code never leaves your machine

Use cases of Constellation

• Find where a symbol is defined and used across the codebase
• Assess the blast radius of a code change before editing
• Detect circular dependencies or dead code in modules
• Trace data flow through the call graph

FAQ from Constellation

How does Constellation differ from plain text search?

Text search tells you where a string appears everywhere; Constellation gives you the exact location of a symbol, its meaning, its callers, and the impact of changing it — from a semantic graph, not a grep loop.

What languages does Constellation support?

TypeScript, JavaScript, and Python are supported, with more languages planned.

Where does my source code live?

Your source code never leaves your machine. Constellation parses it to a compressed AST locally; only structural metadata is sent to the service. Access keys scope each project to its own namespace.

What are the runtime requirements?

Constellation runs queries in an isolated child process with hard memory limits. It auto‑resolves per‑project configuration by git root with LRU caching.

How do I configure access?

Each project is scoped to its own namespace via access keys. No further authentication details are provided in the README.