Docs/Advanced: command line/Context Engine

Context Engine

Dynamo indexes your game project locally. The assistant reads that index at chat time so answers point at real types, real files, and real call chains. Today the index comes from the Dynamo Unity package; in a folder that isn't a game project, the project-wide tools are off.

Heads up on data flow: the index is built on your machine without network calls, but every chat turn still sends your prompts and the indexed snippets the model retrieves to whichever LLM provider you've configured. Local-first applies to the index, not to inference.

How it works

A hover card on a type in a reply, showing its shape and where it is defined
In the app: hover a type to see it
The What Dynamo has read card: files indexed and knowledge base created, with Read again
In the app: what Dynamo has read
  • ●Local, deterministic index — the engine package catalogs classes, methods, scenes, prefabs, singletons, and input bindings. The indexer itself doesn't call an LLM and doesn't make network requests.
  • ●Stays current on its own — Dynamo reads the project again when it opens and after a turn that changed files; the app also does it when you resume a chat or start a new one. When the index already exists, only the changed files are read again.
  • ●Ready-made subsystem summaries — rendered from the index as markdown. The assistant reads them; so can you.
  • ●Structural answers in one call — "how does X work", "what depends on Y", "trace the Z pipeline" come back complete: call chains, file paths, summary context inline.
  • ●Handles vague questions — no need to name the exact class. The engine picks the most likely match, or lists the close candidates.
  • ●Type details in one lookup — asking about a class returns its file path, fields, properties, method signatures, and inheritance together. Per-method call lists are added on request. Very large types (over 50 methods) show the 20 most important.
  • ●Works across providers — the engine is designed to give a consistent experience on Claude, GPT, Gemini, OpenRouter, and Ollama. Actual results vary by model capability and tool-calling fidelity; smaller and local models may converge less reliably on complex prompts.

Configure

The context engine is on by default. Tune it in dynamo.yaml:

yaml
ai:
  context:
    enabled: true
    auto_index: true
    cache:
      type: none         # or 'vcs' to share via version control

On the command line, use /context to change settings live.

ConfigurationProject Instructions