fix: gadfly review — video-sized response cap, stale comment

maxVideoResponseBytes (512MB) replaces the shared 64MB JSON cap on the
/v1/videos/sync read path (doRaw now takes the cap per call) — 3/6
models flagged that a legitimate long/high-bitrate clip would be
discarded after minutes of GPU work. Plus a stale stable-diffusion
comment in initImageFilename and a test-handler early return.

Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01AXQxVhXBw8PwFAtsVrXSmj

provider/llamaswap/audio.go

@@ -63,7 +63,7 @@ func (m *speechModel) Speak(ctx context.Context, req audio.SpeechRequest, opts .
 	if err != nil {
 		return nil, fmt.Errorf("llama-swap: encode speech request: %w", err)
 	}
-	audioBytes, contentType, err := m.p.doRaw(ctx, http.MethodPost, "/v1/audio/speech", m.id, "application/json", bytes.NewReader(body))
+	audioBytes, contentType, err := m.p.doRaw(ctx, http.MethodPost, "/v1/audio/speech", m.id, "application/json", bytes.NewReader(body), maxResponseBytes)
 	if err != nil {
 		return nil, err
 	}
@@ -140,7 +140,7 @@ func (m *transcriptionModel) Transcribe(ctx context.Context, req audio.Transcrip
 		return nil, fmt.Errorf("llama-swap: build transcription form: %w", err)
 	}
 
-	raw, _, err := m.p.doRaw(ctx, http.MethodPost, "/v1/audio/transcriptions", m.id, w.FormDataContentType(), &buf)
+	raw, _, err := m.p.doRaw(ctx, http.MethodPost, "/v1/audio/transcriptions", m.id, w.FormDataContentType(), &buf, maxResponseBytes)
 	if err != nil {
 		return nil, err
 	}
@@ -201,7 +201,7 @@ func (p *Provider) ListVoices(ctx context.Context, model string) ([]string, erro
 	if model == "" {
 		return nil, fmt.Errorf("llama-swap: ListVoices requires a model id")
 	}
-	raw, _, err := p.doRaw(ctx, http.MethodGet, "/v1/audio/voices?model="+url.QueryEscape(model), model, "", nil)
+	raw, _, err := p.doRaw(ctx, http.MethodGet, "/v1/audio/voices?model="+url.QueryEscape(model), model, "", nil, maxResponseBytes)
 	if err != nil {
 		return nil, err
 	}
@@ -256,10 +256,10 @@ func parseVoices(raw []byte) ([]string, error) {
 
 // doRaw performs a request to a llama-swap endpoint and returns the raw
 // response body and its Content-Type — the sibling of doJSON for endpoints
-// whose success payload is not JSON (audio bytes) or whose shape varies.
-// contentType sets the request Content-Type when body is non-nil. A response
-// larger than maxResponseBytes is an error, never a silent truncation.
-func (p *Provider) doRaw(ctx context.Context, method, path, model, contentType string, body io.Reader) ([]byte, string, error) {
+// whose success payload is not JSON (audio/video bytes) or whose shape
+// varies. contentType sets the request Content-Type when body is non-nil.
+// A response larger than maxBytes is an error, never a silent truncation.
+func (p *Provider) doRaw(ctx context.Context, method, path, model, contentType string, body io.Reader, maxBytes int64) ([]byte, string, error) {
 	if err := p.requireBaseURL(); err != nil {
 		return nil, "", err
 	}
@@ -275,12 +275,12 @@ func (p *Provider) doRaw(ctx context.Context, method, path, model, contentType s
 	if resp.StatusCode/100 != 2 {
 		return nil, "", p.apiError(resp, model)
 	}
-	data, err := io.ReadAll(io.LimitReader(resp.Body, maxResponseBytes+1))
+	data, err := io.ReadAll(io.LimitReader(resp.Body, maxBytes+1))
 	if err != nil {
 		return nil, "", fmt.Errorf("llama-swap: read response: %w", err)
 	}
-	if len(data) > maxResponseBytes {
-		return nil, "", fmt.Errorf("llama-swap: response exceeds %d bytes", maxResponseBytes)
+	if int64(len(data)) > maxBytes {
+		return nil, "", fmt.Errorf("llama-swap: response exceeds %d bytes", maxBytes)
 	}
 	return data, resp.Header.Get("Content-Type"), nil
 }