You paste one strange error from production.
The message names a function or constant, but not the file that caused it.
Plan
FixMap uses the exact words from the error and code names it can confirm. Then it shows matching files and tests.
When a coding problem leaves you asking, “Where do I even start?”
FixMap checks the problem and the project together. These are 13 real moments when that starting list can help—and what FixMap can honestly show.
01
You open a project you have never seen.
The bug says TOKEN_TTL_MINUTES is ignored and reset links expire immediately. You do not know which file handles resets, which test proves it, or what else should be reviewed.
Real output from the sample project
- File to check first
src/auth/reset-password.ts- Test to run
npm run test- Other code to review
src/http/routes.ts- Area to review
- authentication
sample-api project.More moments where FixMap fits.
Each row describes a real FixMap workflow. No customer story or guaranteed result is invented.
Your AI tool suggests a file that looks wrong.
You expected one file, but it is missing from FixMap's first results.
Explain
Ask FixMap about that file. It can explain why the file appears lower, why FixMap left it out, or why it could not read it.
The project has 40 packages and too many test commands.
The main test command may be slow or may not cover the part you are changing.
Test routing
FixMap finds the package that owns the file, then points to that package's test command and nearby tests.
Your AI tool keeps opening built files instead of source code.
A generated file in a build folder looks relevant because it repeats every useful code word.
Explain
FixMap can warn that the file was generated and point you back toward the source code people actually maintain.
One small edit may affect code somewhere else.
The file looks isolated, but another route, test, or package may depend on it.
Impact Graph
It shows code this file uses, code that uses it, matching tests, and files often changed with it.
The ticket only says “make it better.”
There is not enough detail to choose a file honestly.
Uncertainty
FixMap can return no file suggestions and ask for a failing behavior, error message, command, code name, or path instead of making up an answer.
You add one useful detail to a vague task.
You want to know whether the better wording actually changed the result.
Compare
Compare shows which files appeared, disappeared, moved up or down, or became more or less certain.
The AI tool finished editing. Did it follow the plan?
The final change may touch files that were never discussed or skip the leading file entirely.
Verify
Verify compares the saved list with the files the AI tool actually changed. It points out extra files, important files it skipped, and tests or risk areas to review.
The fix takes several rounds.
The files and risks can change while the AI tool keeps editing.
Watch
Watch checks again each time files change. It reads the files and their history but does not run the project.
Your AI tool needs source code, not just file names.
Sending the whole project is noisy, but a list of paths is not enough to work from.
Context
Context packs the most useful pieces of code into a fixed size limit and tells you what it left out.
You want the same map on every pull request.
Reviewers should not have to run a local command just to see the planned files, checks, and risks.
GitHub Action
The GitHub Action can add or update one FixMap report on a pull request instead of posting duplicate comments.
You want to know whether FixMap works on your own project.
A public test cannot tell you how file finding behaves in your project.
Benchmark
Benchmark checks older changes from your project and compares FixMap with a basic code search. It reads files and history without running the project.
See a situation you recognize?
Try FixMap on the sample project first. If the output makes sense, add it to your own workflow.