lukaszraczylo/llm-testbench
1
0
Fork 0
mirror of https://github.com/lukaszraczylo/llm-testbench.git synced 2026-10-16 01:07:47 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update

Browse Source
This commit is contained in:
lukaszraczylo
2026-08-30 12:37:38 +01:00
parent 5eef0482de
commit cde7c7ce14
29 changed files with 4548 additions and 434 deletions
Download Patch File Download Diff File Expand all files Collapse all files
internal/report/markdown.go
+23 -6
View File
@@ -10,8 +10,8 @@ import (
"github.com/lukaszraczylo/llm-testbench/internal/testkit"
)
// renderMarkdown writes the same three tables as renderTable, formatted as
// GitHub-flavored markdown tables.
// renderMarkdown writes the same four sections as renderTable, formatted
// as GitHub-flavored markdown tables.
func renderMarkdown(w io.Writer, tests []testkit.Test, models []string, results []runner.Result) error {
idx := indexResults(results)
ew := &errWriter{w: w}
@@ -24,9 +24,9 @@ func renderMarkdown(w io.Writer, tests []testkit.Test, models []string, results
for _, t := range sortedTests(tests) {
row := []string{t.ID, t.Category, t.Subcategory}
for _, model := range models {
r, ok := idx[t.ID][model]
row = append(row, cellText(r, ok))
anyTruncated = anyTruncated || r.Truncated()
c := idx[t.ID][model]
row = append(row, cellText(c))
anyTruncated = anyTruncated || c.result.Truncated()
}
writeMarkdownRow(ew, row)
}
@@ -48,11 +48,16 @@ func renderMarkdown(w io.Writer, tests []testkit.Test, models []string, results
writeMarkdownRow(ew, row)
}
ew.println()
ew.println("## Discrimination")
ew.println()
writeDiscriminationRows(markdownPrinter{ew: ew}, testStats(tests, models, idx), len(models) > 1)
ew.println()
ew.println("## Model summary")
ew.println()
writeMarkdownHeader(ew, []string{"Model", "Mean", "Passed", "Partial", "Failed", "Errors", "Mean latency", "Total tokens", "Tok/s"})
for _, ms := range summarize(models, results) {
for _, ms := range summarize(models, idx) {
writeMarkdownRow(ew, []string{
ms.model,
meanCellText(ms.overallMean),
@@ -69,6 +74,18 @@ func renderMarkdown(w io.Writer, tests []testkit.Test, models []string, results
return ew.err
}
// markdownPrinter adapts writeDiscriminationRows to pipe-delimited rows.
// GFM needs a blank line between the note paragraph and the table, so
// header() emits one first.
type markdownPrinter struct{ ew *errWriter }
func (mp markdownPrinter) header(cells ...string) {
mp.ew.println()
writeMarkdownHeader(mp.ew, cells)
}
func (mp markdownPrinter) row(cells ...string) { writeMarkdownRow(mp.ew, cells) }
func (mp markdownPrinter) note(text string) { mp.ew.println(text) }
func writeMarkdownHeader(ew *errWriter, cols []string) {
ew.println("| " + strings.Join(cols, " | ") + " |")
sep := make([]string, len(cols))