Implement agent to aid with algorithm development

In the algorithm writing screen, the agent will be used to interact with the code.
The assistant chat should allow the user to ask the agent for help.
The agent can reply with new configuration and updated code.

The new code should be shown as a diff that the user can accept or reject.
Given the simplicity of the configuration changes, those changes (eg: date) are accepted automatically.

The assistant can also be used to help with error handling. Upon an error it should be possible to send the error to the assistant for interpretation and it should respond with a solution to the problem and, if warranted, a code change.

The last aid provided by the assistant should be AI code completion, to be added to the code editor itself instead of the assistant chat.

No files or tests are named. Start by locating the algorithm writing screen and its code editor, then trace how configuration changes and errors are handled. Done would require agreed behavior for assistant chat, code/configuration diffs, error assistance, and editor completion, with corresponding tests.

Written by the indexing model from the issue text.