Output

Source files: src/output/progress.ts, src/output/terminal.ts

The output module provides real-time terminal feedback while tests are running. LLM invocations can take anywhere from seconds to minutes per test file, so without live feedback the tool would appear frozen. The spinner system animates Unicode Braille characters at 80ms intervals, showing which test is currently running and progress through the suite. When a test completes, the spinner is replaced with a result line showing pass/fail counts.

The colour system uses raw ANSI escape codes rather than a library like chalk. It detects whether stdout is a TTY — in a terminal, it shows animated spinners and coloured output. In piped or CI output (where escape codes and line overwriting would produce garbage), it falls back to static, unadorned lines.

Spinner system

createProgress(total) returns a Progress object with three methods:

start(testName, index)

Starts an animated spinner (TTY) or prints a static line (non-TTY):

  ⠹ [1/5] Running auth-middleware.spec.md...

Uses Unicode Braille frames (⠋⠙⠹⠸⠼⠴⠦⠧⠇⠏) cycling at 80ms intervals.

retry(testName, attempt, maxRetries)

Replaces the spinner with a retry indicator:

  ⟳ [1/3] Retrying auth-middleware.spec.md...

repeatRun(testName, repetition, total)

Shows the current repeat iteration when --repeat is greater than 1:

  ⟳ [1/5] auth-middleware.spec.md (run 2/3)...

result(testName, results, index, verbose?)

Stops the spinner and prints the final result with an icon:

  ✔ [1/5] auth-middleware.spec.md  ·  2 passed
  ✗ [2/5] api-routes.spec.md  ·  1 failed
  ⚠ [3/5] broken-test.spec.md  ·  1 errored

Results from a single test file may contain multiple scenarios (e.g. “2 passed, 1 failed”).

Verbose mode

Verbose output appears in two places:

1. Progress results — The result() method prints per-scenario lines below each test result, showing the individual test ID and status icon.
2. Terminal summary — printSummary() prints a grouped breakdown of all results. For failures, it includes the expectation, observed behaviour, location, and suggested resolution.

This is useful during development when you want to see exactly what the LLM reported without opening the full Markdown report.

Colour utilities

The colors object is exported for use across the output layer:

Function	ANSI code	Usage
green	\x1b[32m	Passed counts
red	\x1b[31m	Failed counts
yellow	\x1b[33m	Errored counts, warnings
cyan	\x1b[36m	Spinner frames
dim	\x1b[2m	Index tags, separators
bold	\x1b[1m	Emphasis
greenBold	\x1b[1;32m	Final “PASSED” verdict
redBold	\x1b[1;31m	Final “FAILED” verdict
yellowBold	\x1b[1;33m	Emphasis warnings

Colour is disabled when NO_COLOR is set or FORCE_COLOR=0, following the no-color standard.

Terminal summary

printSummary() in terminal.ts prints the final output after all tests complete:

Semantic tests completed

Report:    semtest-results/latest.md
CI Output: semtest-results/ci-results.json
JUnit:     semtest-results/junit-results.xml
Debug:     semtest-results/debug/

Passed: 10
Failed: 2
Errored: 1

Validation Issues:
  - [duplicate-id] Duplicate test ID "auth-check" found in: auth.spec.md, session.spec.md

FAILED

Sections

Line	Colour	Condition
Report path	—	Always
CI Output path	—	Always
JUnit path	—	Only if --junit enabled
Debug path	—	Only if --debug enabled
Passed count	Green	Always
Failed count	Red	Always
Errored count	Yellow	Only if > 0
Skipped count	Dim	Only if > 0
Invalid count	Dim	Only if > 0
Validation issues	Yellow header	Only if validation ran and has issues
Final verdict	Green bold / Red bold	”PASSED” or “FAILED”