Cross-Node Context

One of Kommit's most powerful features is that the AI is never working in isolation. When you chat on a Feature node, the AI already knows your tech stack, who your users are, and what your competitors are doing. It uses all of that to give better answers.

Why this matters

Specs are interconnected. A feature's complexity depends on the tech stack. Acceptance criteria depend on the target user. The AI cannot write useful, specific requirements if it only sees the feature in isolation.

Cross-node context solves this by injecting relevant information from across your canvas into every conversation turn.

How context is prioritised

Each conversation turn receives context from across your project. Connected nodes receive the highest priority and are included with the most detail, while other nodes contribute summaries. This ensures the AI always has the most relevant information for the conversation at hand.

The injected context is size-limited per turn. For most projects this is more than enough to include everything in full. On very large canvases with dozens of nodes, the priority ordering ensures the most relevant information always makes it in.

Practical examples

Chatting on a Feature node

If your Feature node is connected to Tech Stack and Audience:

• The AI knows your full stack (React, Node.js, Postgres, etc.) and will tailor acceptance criteria accordingly
• The AI knows your primary persona and will frame user stories in their context
• The AI has one-line summaries of your other features and can flag dependencies

Chatting on a Tech Stack node

• The AI knows your product goals from the Project node
• The AI knows the features you're planning and can flag where your stack choices may cause friction
• The AI knows your competitors and can note where their known tech choices differ

Chatting on a Competitors node

• The AI knows your target audience and can assess competitor gaps relative to your users' pain points
• The AI knows your own feature set and can help you articulate differentiation clearly

Performance

Context delivery is optimised to keep response times fast even as your canvas grows. You do not need to manage context manually — Kommit handles it automatically.

Repository context

If you've connected a GitHub repository, its ingested code chunks are also available as context — retrieved via semantic search at query time. The most relevant code excerpts are appended to the context payload when the AI determines they are useful.

This means the AI can reference your actual codebase when helping you spec a new feature — noting existing patterns, database schemas, or API conventions that the feature should follow.

See GitHub integration for setup details.