style(ui): prettier formatting pass over the dashboard

[ryan-crabbe-berri]

Relevant issues

Linear ticket

Refs LIT-3556

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have added meaningful tests
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Screenshots / Proof of Fix

Prettier is configured in the UI (.prettierrc, format/format:check scripts) but the codebase had drifted out of compliance. This is a mechanical prettier --write . pass that brings every file into line; it changes formatting only, no logic.

Before, format:check failed on 608 files; after the pass it is clean:

$ npm run format:check   # before
Checking formatting...
[warn] ... 608 files ...
[warn] Code style issues found in 608 files. Run Prettier to fix.

$ npm run format         # prettier --write .
$ npm run format:check   # after
Checking formatting...
All matched files use Prettier code style!

npm ci && npm run build still passes after the pass, confirming nothing broke.

Enforcing the check in CI is intentionally left to a follow-up PR so this one stays a reviewable, logic-free diff.

Type

🧹 Refactoring

Changes

A single formatting-only pass (prettier --write .) over ui/litellm-dashboard, touching 608 files. Reflows, quote normalization, and trailing commas only; no behavior changes. Tailwind class sorting and import sorting were intentionally left out to avoid scope creep.

[greptile-apps]

Too many files changed for review. (610 files found, 100 file limit)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[mateo-berri]

AST equivalence check: confirmed format-only

I compared the parsed syntax tree of every changed file between the PR head (1b06a428) and its base (5ee526d7) to verify this branch changes formatting only and touches no behavior. All 608 changed files are semantically identical to their base versions.

Method

Each .ts / .tsx / .jsx file is parsed with the dashboard's own TypeScript compiler (v5.9.3, the same parser the build uses) at both commits, then serialized to a position-free token stream and compared. The serialization deliberately folds away exactly the transformations Prettier is allowed to make, so anything Prettier does registers as "equal" while any real code change registers as "different":

• whitespace, indentation, line wrapping, semicolons and trailing commas are not part of the AST at all
• string quote style (' vs ") and numeric formatting (0xAB vs 0xab) are compared by decoded value, not raw text
• redundant parentheses are unwrapped (a precedence-changing paren is not, and stays a difference)
• object property keys are compared by key value, so quoteProps removal ("Akto": -> Akto:) is equal
• JSX children are compared by their rendered text, which covers Prettier's {" "} reflow and the DOM's collapsing of whitespace runs
• HTML entities in JSX text and attribute values are decoded, so description="..."X"" equals description='..."X"'
• comments are trivia and are ignored

The two JSON files are compared by parsed value, and the one Markdown file by its word sequence.

Results

Group	Files	Verdict
.ts / .tsx / .jsx	605	identical normalized AST
knip.json, tsconfig.json	2	identical parsed JSON value
(dashboard)/README.md	1	identical word sequence (blank-line / trailing-whitespace only)
Total	608	no semantic change

The only diffs that went beyond pure whitespace were all behavior-preserving Prettier defaults: quoteProps: "as-needed" dropping redundant object-key quotes in four files, and one JSX attribute switching quote style to drop " entities in RouterConfigBuilder.tsx. Both decode to identical values and are accounted for above.

Why the "identical" verdict is trustworthy

To make sure the comparator is not simply blind, I ran it against a 24-case harness. The ten pure-reformat variants (quote flips, paren removal, reindent + trailing commas, quoteProps, numeric reformat, {" "} reflow, JSX double-space collapse, entity vs literal quote, added comment) all compared equal, and all fourteen genuine mutations were flagged as different: identifier rename, numeric value change, string value change, === vs !==, + vs -, let vs const, removed prefix !, a precedence-changing paren, a dropped call argument, changed JSX visible text, changed JSX attribute value, swapped ternary branches, removed await, and a negated if condition.

Reproduce

Both scripts (the comparator and the 24-case validation harness) plus a short README are here: https://gist.github.com/mateo-berri/a6397ba65f76a258ab4aa160874b5053

cd ui/litellm-dashboard
FILES=$(git -C ../.. diff --name-only 5ee526d7..1b06a428 | grep -E '\.(tsx|ts|jsx|js)$')
TS_PATH="$PWD/node_modules/typescript" BASE_REF=5ee526d7 HEAD_REF=1b06a428 \
  node ast_compare.js $FILES        # -> 605 identical AST, 0 mismatched
TS_PATH="$PWD/node_modules/typescript" node ast_selftest.js   # -> 24 passed, 0 failed

Net: this branch is a pure Prettier pass and is safe to merge as format-only

[mateo-berri]

AST compare'ed and it LGTM. Just do the git blame ignore thing so this PR doesn't show up everywhere

[mateo-berri]

LGTM; agreed that we lose the hash when we squash. Fast-follow with git blame ignore