Compare commits

...

3 Commits

Author SHA1 Message Date
Daniel Hiltgen
554172759c win: warn if ggml-base detected in PATH (#13289)
If the user has somehow installed another GGML based app which places a
ggml-base lib somewhere in their PATH, we can experience runtime problems
due to incompatibilities.  This change adds a warning message if we detect
a ggml-base outside of our install location to aid in troubleshooting.
2025-12-01 15:36:47 -08:00
Bruce MacDonald
5b6a8e6001 api/client: handle non-json streaming errors (#13007)
While processing the response stream during a chat or generation if an error is occurred it is parsed and returned to the user. The issue with the existing code is that this assumed the response would be valid JSON, which is not a safe assumption and caused cryptic error messages to be displayed due to parsing failures:
`invalid character 'i' looking for beginning of value`

This change updates the stream function to return the raw error string if it cant be parsed as JSON. This should help with debugging issues by making sure the actual error reaches the user.
2025-12-01 15:10:16 -08:00
Daniel Hiltgen
467bbc0dd5 jetpack: require exact match or skip cuda_jetpack* (#13288)
The cuda_jetpack libs will enumerate discrete GPUs on SBSA systems
which leads to runtime failures of missing kernels.  This fix
requires an exact match to enable jetpacks instead of relying on
enumeration to filter out supported libraries.
2025-12-01 12:48:16 -08:00
3 changed files with 93 additions and 11 deletions

View File

@@ -226,7 +226,14 @@ func (c *Client) stream(ctx context.Context, method, path string, data any, fn f
bts := scanner.Bytes()
if err := json.Unmarshal(bts, &errorResponse); err != nil {
return fmt.Errorf("unmarshal: %w", err)
if response.StatusCode >= http.StatusBadRequest {
return StatusError{
StatusCode: response.StatusCode,
Status: response.Status,
ErrorMessage: string(bts),
}
}
return errors.New(string(bts))
}
if response.StatusCode == http.StatusUnauthorized {

View File

@@ -55,6 +55,7 @@ func TestClientFromEnvironment(t *testing.T) {
type testError struct {
message string
statusCode int
raw bool // if true, write message as-is instead of JSON encoding
}
func (e testError) Error() string {
@@ -111,6 +112,20 @@ func TestClientStream(t *testing.T) {
},
},
},
{
name: "plain text error response",
responses: []any{
"internal server error",
},
wantErr: "internal server error",
},
{
name: "HTML error page",
responses: []any{
"<html><body>404 Not Found</body></html>",
},
wantErr: "404 Not Found",
},
}
for _, tc := range testCases {
@@ -135,6 +150,12 @@ func TestClientStream(t *testing.T) {
return
}
if str, ok := resp.(string); ok {
fmt.Fprintln(w, str)
flusher.Flush()
continue
}
if err := json.NewEncoder(w).Encode(resp); err != nil {
t.Fatalf("failed to encode response: %v", err)
}
@@ -173,9 +194,10 @@ func TestClientStream(t *testing.T) {
func TestClientDo(t *testing.T) {
testCases := []struct {
name string
response any
wantErr string
name string
response any
wantErr string
wantStatusCode int
}{
{
name: "immediate error response",
@@ -183,7 +205,8 @@ func TestClientDo(t *testing.T) {
message: "test error message",
statusCode: http.StatusBadRequest,
},
wantErr: "test error message",
wantErr: "test error message",
wantStatusCode: http.StatusBadRequest,
},
{
name: "server error response",
@@ -191,7 +214,8 @@ func TestClientDo(t *testing.T) {
message: "internal error",
statusCode: http.StatusInternalServerError,
},
wantErr: "internal error",
wantErr: "internal error",
wantStatusCode: http.StatusInternalServerError,
},
{
name: "successful response",
@@ -203,6 +227,26 @@ func TestClientDo(t *testing.T) {
Success: true,
},
},
{
name: "plain text error response",
response: testError{
message: "internal server error",
statusCode: http.StatusInternalServerError,
raw: true,
},
wantErr: "internal server error",
wantStatusCode: http.StatusInternalServerError,
},
{
name: "HTML error page",
response: testError{
message: "<html><body>404 Not Found</body></html>",
statusCode: http.StatusNotFound,
raw: true,
},
wantErr: "<html><body>404 Not Found</body></html>",
wantStatusCode: http.StatusNotFound,
},
}
for _, tc := range testCases {
@@ -210,11 +254,16 @@ func TestClientDo(t *testing.T) {
ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if errResp, ok := tc.response.(testError); ok {
w.WriteHeader(errResp.statusCode)
err := json.NewEncoder(w).Encode(map[string]string{
"error": errResp.message,
})
if err != nil {
t.Fatal("failed to encode error response:", err)
if !errResp.raw {
err := json.NewEncoder(w).Encode(map[string]string{
"error": errResp.message,
})
if err != nil {
t.Fatal("failed to encode error response:", err)
}
} else {
// Write raw message (simulates non-JSON error responses)
fmt.Fprint(w, errResp.message)
}
return
}
@@ -241,6 +290,15 @@ func TestClientDo(t *testing.T) {
if err.Error() != tc.wantErr {
t.Errorf("error message mismatch: got %q, want %q", err.Error(), tc.wantErr)
}
if tc.wantStatusCode != 0 {
if statusErr, ok := err.(StatusError); ok {
if statusErr.StatusCode != tc.wantStatusCode {
t.Errorf("status code mismatch: got %d, want %d", statusErr.StatusCode, tc.wantStatusCode)
}
} else {
t.Errorf("expected StatusError, got %T", err)
}
}
return
}

View File

@@ -65,6 +65,7 @@ func GPUDevices(ctx context.Context, runners []ml.FilteredRunnerDiscovery) []ml.
}
slog.Info("discovering available GPUs...")
detectIncompatibleLibraries()
// Warn if any user-overrides are set which could lead to incorrect GPU discovery
overrideWarnings()
@@ -98,6 +99,9 @@ func GPUDevices(ctx context.Context, runners []ml.FilteredRunnerDiscovery) []ml.
continue
} else if jetpack != "" && filepath.Base(dir) != "cuda_"+jetpack {
continue
} else if jetpack == "" && strings.Contains(filepath.Base(dir), "cuda_jetpack") {
slog.Debug("jetpack not detected (set JETSON_JETPACK or OLLAMA_LLM_LIBRARY to override), skipping", "libDir", dir)
continue
} else if !envconfig.EnableVulkan() && strings.Contains(filepath.Base(dir), "vulkan") {
slog.Info("experimental Vulkan support disabled. To enable, set OLLAMA_VULKAN=1")
continue
@@ -484,3 +488,16 @@ func overrideWarnings() {
slog.Warn("if GPUs are not correctly discovered, unset and try again")
}
}
func detectIncompatibleLibraries() {
if runtime.GOOS != "windows" {
return
}
basePath, err := exec.LookPath("ggml-base.dll")
if err != nil || basePath == "" {
return
}
if !strings.HasPrefix(basePath, ml.LibOllamaPath) {
slog.Warn("potentially incompatible library detected in PATH", "location", basePath)
}
}