fix: harden upstream path + binary-body validation (gadfly round 1)

- upstreamPath rejects '..' in model ids AND in the rest path — the rest
  can embed SERVER-SUPPLIED components (ACE-Step result file URLs), so
  dot-dot/scheme smuggling toward other proxy endpoints is refused
- singleImageResult requires positive image evidence (sniffed magic OR
  declared image/*): an empty-Content-Type error page can no longer pass
  as 'the image' via sniffImageMIME's PNG-default labelling
- upscale/background responses get a dedicated 256MB cap (the 64MB cap
  is JSON-sized; a 4x PNG legitimately exceeds it)
- mesh JSON-detection widened (512-byte whitespace-tolerant peek + reject
  declared application/json)
- Transcribe now reuses buildMultipart; transcriptionFilename takes
  (filename, mime) so diarize shares it without a fake request struct;
  truncateForError stops shadowing builtin cap; OnlyMask doc de-ambiguated

Co-Authored-By: Claude Fable 5 <[email protected]>

provider/llamaswap/mesh.go

@@ -102,7 +102,7 @@ func (m *meshModel) Generate(ctx context.Context, req meshgen.Request, opts ...m
 	if err != nil {
 		return nil, fmt.Errorf("llama-swap: encode mesh request: %w", err)
 	}
-	raw, _, err := m.p.doRaw(ctx, http.MethodPost, path, m.id, "application/json", bytes.NewReader(encoded), maxMeshResponseBytes)
+	raw, respType, err := m.p.doRaw(ctx, http.MethodPost, path, m.id, "application/json", bytes.NewReader(encoded), maxMeshResponseBytes)
 	if err != nil {
 		return nil, err
 	}
@@ -110,8 +110,16 @@ func (m *meshModel) Generate(ctx context.Context, req meshgen.Request, opts ...m
 		return nil, &llm.APIError{Provider: m.p.name, Model: m.id, Message: "mesh response contained no data"}
 	}
 	// A JSON body on the success path is the server reporting a soft error
-	// (or an API drift) — never hand it back as "the mesh".
-	if first := strings.TrimLeft(string(raw[:min(len(raw), 64)]), " \t\r\n"); strings.HasPrefix(first, "{") || strings.HasPrefix(first, "[") {
+	// (or an API drift) — never hand it back as "the mesh". Two signals:
+	// the declared Content-Type, and a whitespace-tolerant peek at the
+	// leading bytes (512 covers any indented error envelope; a binary STL
+	// header theoretically CAN start with '{', but a real one also won't
+	// be all-whitespace-then-brace).
+	if strings.Contains(strings.ToLower(respType), "application/json") {
+		return nil, &llm.APIError{Provider: m.p.name, Model: m.id,
+			Message: "mesh response is JSON, not mesh bytes: " + truncateForError(raw)}
+	}
+	if first := strings.TrimLeft(string(raw[:min(len(raw), 512)]), " \t\r\n"); strings.HasPrefix(first, "{") || strings.HasPrefix(first, "[") {
 		return nil, &llm.APIError{Provider: m.p.name, Model: m.id,
 			Message: "mesh response is JSON, not mesh bytes: " + truncateForError(raw)}
 	}
@@ -120,9 +128,9 @@ func (m *meshModel) Generate(ctx context.Context, req meshgen.Request, opts ...m
 
 // truncateForError bounds a payload quoted into an error message.
 func truncateForError(b []byte) string {
-	const cap = 500
-	if len(b) > cap {
-		return string(b[:cap]) + "..."
+	const maxErrLen = 500
+	if len(b) > maxErrLen {
+		return string(b[:maxErrLen]) + "..."
 	}
 	return string(b)
 }