Project
Check quoted commas and missing fields without overwriting the original input.
AI coding workspace
From task to working code.
Work inside the code you already have. Kiteforge brings the project conversation and its files into one workspace, with model selection and changes you can inspect.
● ● ●Kiteforge●
A concrete starting point · Project changes ×
const rows = parseCSV(input);
const valid = rows.filter(validateRow);
return { rows: valid, errors };✳Import CSV data and report invalid rows.
✓ ChangesReview before applying ↗
Illustrative workspace
Review before applying
Compare actual behavior with the expected result. If a check fails, investigate it before applying more changes.
Connection
Base URL, credentials and a reachable server.
A concrete starting point
const rows = parseCSV(input);
const valid = rows.filter(validateRow);
return { rows: valid, errors };Understand what a maintenance change will touch
A maintenance request often sounds smaller than the code behind it. Adding clearer errors to a CSV import can involve the parser, row validation and the screen that reports rejected records. In Kiteforge, the Windows coding workspace, begin by asking the assistant to inspect those connections in your project folder before proposing edits.
That inspection gives the request useful boundaries. Ask where invalid rows are handled, which callers consume the result and what existing tests expect. For an importer that already accepts valid rows, specify that this behavior must stay intact while rejected rows gain an explanation. The proposal should identify the files involved and explain why each needs attention.
Choose the connected model with that work in mind. If you are considering an uncensored AI model, Kiteforge supports connections through supported OpenAI-compatible or Anthropic endpoints. The model and server determine context limits, tool support and response restrictions. Check the connection before relying on it to inspect or edit project files.
Once edits are proposed, compare the diff with the original request. Look for unrelated cleanup, changed return values or dependency updates that widen the task. Run the relevant project tests, then exercise the import with valid and rejected records. Those checks help establish whether the error message improved without disrupting successful imports.
Planning a change in an existing repository
What should I ask for before any files are edited?
Request a read-only inspection of the affected code, its callers and relevant tests. Ask for a proposed file list and an explanation of the current behavior, then define the change you want.
How do I keep an importer fix from becoming a rewrite?
State what must remain stable: accepted rows, output fields and existing callers, for example. Ask for rejected-row explanations separately, and review any broader restructuring before including it in the task.
Can I assume the assistant has read the whole repository?
No. Ask which files informed the proposal and which dependencies remain unexamined. Context limits and available tools depend on the connected model and server; narrow the task if important code is missing.
How should I prepare an existing project for the first task?
Keep a recoverable copy or commit your current work, then open the intended project folder. Start with an inspection request for one behavior and confirm the referenced files before asking for edits. Keep unrelated local changes outside the proposed patch.
What evidence should I check before accepting the patch?
Review the actual diff and test output. For the importer, exercise a valid file, rejected records and mixed input, checking that the displayed errors agree with the rows the application accepts.
Kiteforge · Updated 2026-10-25
