Master the terminology behind codebase-grounded AI coding assistance.
0 / 5 completed
1 / 5
At standup, a dev wants an AI coding assistant that can answer questions grounded in an entire large codebase's actual code, not just the open file. Which tool fits?
Cody is Sourcegraph's AI coding assistant built to answer questions and generate code grounded in context retrieved from across an entire codebase, leveraging Sourcegraph's code search and indexing. This differs from assistants limited to only the currently open file. Codebase-wide grounding is its core differentiator.
2 / 5
During a design review, the team wants Cody's answers to cite the actual files and functions it drew context from. Which capability supports this?
Cody surfaces citations pointing to the specific files and code it used as context for an answer, letting developers verify the grounding rather than trusting an unreferenced claim. This traceability builds confidence in the assistant's output. It parallels how retrieval-augmented systems generally expose their sources.
3 / 5
In a code review, a dev asks Cody to find all usages of a deprecated internal function across a large monorepo. Which underlying capability enables this?
Cody relies on Sourcegraph's codebase-wide search and indexing infrastructure to locate all relevant usages across potentially millions of lines, something a single-file or manual search would struggle to do reliably at scale. This search foundation is what enables accurate, codebase-grounded answers. It is central to Sourcegraph's broader code intelligence platform.
4 / 5
An incident report shows Cody suggested a pattern that was actually deprecated in the current codebase. What should the team check?
If retrieved context is stale relative to recent code changes, an assistant grounded in that index can surface outdated or deprecated patterns as if they were current. Ensuring the index stays up to date with the codebase reduces this risk. This is a general challenge for any retrieval-grounded coding assistant, not unique to Cody.
5 / 5
During a PR review, a teammate asks how Cody differs from a general-purpose chat assistant with no codebase integration. What is the distinction?
A general-purpose assistant without codebase integration answers from its trained knowledge alone, while Cody actively retrieves and grounds its answers in the real, current state of the specific codebase being worked on. This grounding is especially valuable for large, private, or fast-changing codebases. It reduces the risk of confidently wrong answers based on outdated or generic assumptions.