Context

You’ve probably heard when working with AI that “Context is King”. Here’s everything you need to know about context - what context does Fine have? How do you add more?

What is Context in AI Coding?

Context in AI coding refers to the additional information an AI system has about a task, allowing it to make more accurate and relevant coding suggestions. By understanding the code’s purpose, dependencies, and surrounding logic, AI models can generate more sophisticated and functional solutions. Contextual awareness enables smoother integration of AI tools within development workflows, enhancing productivity and reducing the need for manual corrections.

What is a context window?

A context window in AI refers to the amount of information an AI model can process at once when generating responses or making predictions. This window typically includes a fixed number of words, tokens, or lines of code, which provide the model with essential context, such as prior interactions or code snippets, to produce coherent and relevant outputs. The size of the context window limits how much the model can "remember" from prior input, impacting the depth and continuity of its responses.

As AI providers such as OpenAI and Anthropic improve the context windows of their models, Fine will be able to process more context in each task you give it. Currently, it’s better to break large tasks into smaller steps which are digestible for the LLMs.

How does Fine get enough context?

When you create an account and integrate your platforms, Fine creates a Knowledge Graph based on your codebase, from where it can quickly use Retrieval-Augmented Generation (RAG) to search the codebase and draw on the relevant information for the AI to generate a response. This means that Fine can edit multiple files in one task, across your architecture.

Guiding Fine with specific context

Imagine being taken to an enormous library filled with books, given a list of tasks and told “all the information you need is here”. It’s much more helpful to be told which books you need to use to complete the task - you’re less likely to miss something or duplicate something that already exists.

The same goes for Fine. Although it has access to your entire codebase and can perform a search, the more you guide it, the better the results will be. This is why we’ve created a number of ways to add context to your tasks and questions.

How to Add Context to Fine

Files, folders, documents and components in GitHub

Use @ to reference different parts of your codebase.

Linear Issues

To reference a Linear issue, use @ or copy and paste the link into the conversation (public or private).

GitHub PRs

To reference a GitHub PR, copy and paste the link into the conversation.

Sentry Issues

To reference a Sentry issue, copy and paste the public link into the conversation.

View the Context Used

When asking a question or delegating a task to Fine, you can view the context it has drawn upon in the Knowledge pane. Click the book icon at the top right of the page to view the pane.