mirror of
https://github.com/ollama/ollama.git
synced 2026-03-09 07:16:38 -05:00
don't require pulling stubs for cloud models (#14574)
* don't require pulling stubs for cloud models This is a first in a series of PRs that will better integrate Ollama's cloud into the API and CLI. Previously we used to have a layer of indirection where you'd first have to pull a "stub" model that contains a reference to a cloud model. With this change, you don't have to pull first, you can just use a cloud model in various routes like `/api/chat` and `/api/show`. This change respects <https://github.com/ollama/ollama/pull/14221>, so if cloud is disabled, these models won't be accessible. There's also a new, simpler pass-through proxy that doesn't convert the requests ahead of hitting the cloud models, which they themselves already support various formats (e.g., `v1/chat/completions` or Open Responses, etc.). This will help prevent issues caused by double converting (e.g., `v1/chat/completions` converted to `api/chat` on the client, then calling cloud and converting back to a `v1/chat/completions` response instead of the cloud model handling the original `v1/chat/completions` request first). There's now a notion of "source tags", which can be mixed with existing tags. So instead of having different formats like`gpt-oss:20b-cloud` vs. `kimi-k2.5:cloud` (`-cloud` suffix vs. `:cloud`), you can now specify cloud by simply appending `:cloud`. This PR doesn't change model resolution yet, but sets us up to allow for things like omitting the non-source tag, which would make something like `ollama run gpt-oss:cloud` work the same way that `ollama run gpt-oss` already works today. More detailed changes: - Added a shared model selector parser in `types/modelselector`: - supports `:cloud` and `:local` - accepts source tags in any position - supports legacy `:<tag>-cloud` - rejects conflicting source tags - Integrated selector handling across server inference/show routes: - `GenerateHandler`, `ChatHandler`, `EmbedHandler`, `EmbeddingsHandler`, `ShowHandler` - Added explicit-cloud passthrough proxy for ollama.com: - same-endpoint forwarding for `/api/*`, `/v1/*`, and `/v1/messages` - normalizes `model` (and `name` for `/api/show`) before forwarding - forwards request headers except hop-by-hop/proxy-managed headers - uses bounded response-header timeout - handles auth failures in a friendly way - Preserved cloud-disable behavior (`OLLAMA_NO_CLOUD`) - Updated create flow to support `FROM ...:cloud` model sources (though this flow uses the legacy proxy still, supporting Modelfile overrides is more complicated with the direct proxy approach) - Updated CLI/TUI/config cloud detection to use shared selector logic - Updated CLI preflight behavior so explicit cloud requests do not auto-pull local stubs What's next? - Cloud discovery/listing and cache-backed `ollama ls` / `/api/tags` - Modelfile overlay support for virtual cloud models on OpenAI/Anthropic request families - Recommender/default-selection behavior for ambiguous model families - Fully remove the legacy flow Fixes: https://github.com/ollama/ollama/issues/13801 * consolidate pull logic into confirmAndPull helper pullIfNeeded and ShowOrPull shared identical confirm-and-pull logic. Extract confirmAndPull to eliminate the duplication. * skip local existence checks for cloud models ModelExists and the TUI's modelExists both check the local model list, which causes cloud models to appear missing. Return true early for explicit cloud models so the TUI displays them beside the integration name and skips re-prompting the model picker on relaunch. * support optionally pulling stubs for newly-style names We now normalize names like `<family>:<size>:cloud` into legacy-style names like `<family>:<size>-cloud` for pulling and deleting (this also supports stripping `:local`). Support for pulling cloud models is temporary, once we integrate properly into `/api/tags` we won't need this anymore. * Fix server alias syncing * Update cmd/cmd.go Co-authored-by: Parth Sareen <parth.sareen@ollama.com> * address comments * improve some naming --------- Co-authored-by: ParthSareen <parth.sareen@ollama.com>
This commit is contained in:
Devon Rifkin
GitHub
115
internal/modelref/modelref.go
Normal file
115
internal/modelref/modelref.go
Normal file
@@ -0,0 +1,115 @@
|
||||
package modelref
|
||||
|
||||
import (
|
||||
"errors"
|
||||
"fmt"
|
||||
"strings"
|
||||
)
|
||||
|
||||
type ModelSource uint8
|
||||
|
||||
const (
|
||||
ModelSourceUnspecified ModelSource = iota
|
||||
ModelSourceLocal
|
||||
ModelSourceCloud
|
||||
)
|
||||
|
||||
var (
|
||||
ErrConflictingSourceSuffix = errors.New("use either :local or :cloud, not both")
|
||||
ErrModelRequired = errors.New("model is required")
|
||||
)
|
||||
|
||||
type ParsedRef struct {
|
||||
Original string
|
||||
Base string
|
||||
Source ModelSource
|
||||
}
|
||||
|
||||
func ParseRef(raw string) (ParsedRef, error) {
|
||||
var zero ParsedRef
|
||||
|
||||
raw = strings.TrimSpace(raw)
|
||||
if raw == "" {
|
||||
return zero, ErrModelRequired
|
||||
}
|
||||
|
||||
base, source, explicit := parseSourceSuffix(raw)
|
||||
if explicit {
|
||||
if _, _, nested := parseSourceSuffix(base); nested {
|
||||
return zero, fmt.Errorf("%w: %q", ErrConflictingSourceSuffix, raw)
|
||||
}
|
||||
}
|
||||
|
||||
return ParsedRef{
|
||||
Original: raw,
|
||||
Base: base,
|
||||
Source: source,
|
||||
}, nil
|
||||
}
|
||||
|
||||
func HasExplicitCloudSource(raw string) bool {
|
||||
parsedRef, err := ParseRef(raw)
|
||||
return err == nil && parsedRef.Source == ModelSourceCloud
|
||||
}
|
||||
|
||||
func HasExplicitLocalSource(raw string) bool {
|
||||
parsedRef, err := ParseRef(raw)
|
||||
return err == nil && parsedRef.Source == ModelSourceLocal
|
||||
}
|
||||
|
||||
func StripCloudSourceTag(raw string) (string, bool) {
|
||||
parsedRef, err := ParseRef(raw)
|
||||
if err != nil || parsedRef.Source != ModelSourceCloud {
|
||||
return strings.TrimSpace(raw), false
|
||||
}
|
||||
|
||||
return parsedRef.Base, true
|
||||
}
|
||||
|
||||
func NormalizePullName(raw string) (string, bool, error) {
|
||||
parsedRef, err := ParseRef(raw)
|
||||
if err != nil {
|
||||
return "", false, err
|
||||
}
|
||||
|
||||
if parsedRef.Source != ModelSourceCloud {
|
||||
return parsedRef.Base, false, nil
|
||||
}
|
||||
|
||||
return toLegacyCloudPullName(parsedRef.Base), true, nil
|
||||
}
|
||||
|
||||
func toLegacyCloudPullName(base string) string {
|
||||
if hasExplicitTag(base) {
|
||||
return base + "-cloud"
|
||||
}
|
||||
|
||||
return base + ":cloud"
|
||||
}
|
||||
|
||||
func hasExplicitTag(name string) bool {
|
||||
lastSlash := strings.LastIndex(name, "/")
|
||||
lastColon := strings.LastIndex(name, ":")
|
||||
return lastColon > lastSlash
|
||||
}
|
||||
|
||||
func parseSourceSuffix(raw string) (string, ModelSource, bool) {
|
||||
idx := strings.LastIndex(raw, ":")
|
||||
if idx >= 0 {
|
||||
suffixRaw := strings.TrimSpace(raw[idx+1:])
|
||||
suffix := strings.ToLower(suffixRaw)
|
||||
|
||||
switch suffix {
|
||||
case "cloud":
|
||||
return raw[:idx], ModelSourceCloud, true
|
||||
case "local":
|
||||
return raw[:idx], ModelSourceLocal, true
|
||||
}
|
||||
|
||||
if !strings.Contains(suffixRaw, "/") && strings.HasSuffix(suffix, "-cloud") {
|
||||
return raw[:idx+1] + suffixRaw[:len(suffixRaw)-len("-cloud")], ModelSourceCloud, true
|
||||
}
|
||||
}
|
||||
|
||||
return raw, ModelSourceUnspecified, false
|
||||
}
|
||||
268
internal/modelref/modelref_test.go
Normal file
268
internal/modelref/modelref_test.go
Normal file
@@ -0,0 +1,268 @@
|
||||
package modelref
|
||||
|
||||
import (
|
||||
"errors"
|
||||
"testing"
|
||||
)
|
||||
|
||||
func TestParseRef(t *testing.T) {
|
||||
tests := []struct {
|
||||
name string
|
||||
input string
|
||||
wantBase string
|
||||
wantSource ModelSource
|
||||
wantErr error
|
||||
wantCloud bool
|
||||
wantLocal bool
|
||||
wantStripped string
|
||||
wantStripOK bool
|
||||
}{
|
||||
{
|
||||
name: "cloud suffix",
|
||||
input: "gpt-oss:20b:cloud",
|
||||
wantBase: "gpt-oss:20b",
|
||||
wantSource: ModelSourceCloud,
|
||||
wantCloud: true,
|
||||
wantStripped: "gpt-oss:20b",
|
||||
wantStripOK: true,
|
||||
},
|
||||
{
|
||||
name: "legacy cloud suffix",
|
||||
input: "gpt-oss:20b-cloud",
|
||||
wantBase: "gpt-oss:20b",
|
||||
wantSource: ModelSourceCloud,
|
||||
wantCloud: true,
|
||||
wantStripped: "gpt-oss:20b",
|
||||
wantStripOK: true,
|
||||
},
|
||||
{
|
||||
name: "local suffix",
|
||||
input: "qwen3:8b:local",
|
||||
wantBase: "qwen3:8b",
|
||||
wantSource: ModelSourceLocal,
|
||||
wantLocal: true,
|
||||
wantStripped: "qwen3:8b:local",
|
||||
},
|
||||
{
|
||||
name: "no source suffix",
|
||||
input: "llama3.2",
|
||||
wantBase: "llama3.2",
|
||||
wantSource: ModelSourceUnspecified,
|
||||
wantStripped: "llama3.2",
|
||||
},
|
||||
{
|
||||
name: "bare cloud name is not explicit cloud",
|
||||
input: "my-cloud-model",
|
||||
wantBase: "my-cloud-model",
|
||||
wantSource: ModelSourceUnspecified,
|
||||
wantStripped: "my-cloud-model",
|
||||
},
|
||||
{
|
||||
name: "slash in suffix blocks legacy cloud parsing",
|
||||
input: "foo:bar-cloud/baz",
|
||||
wantBase: "foo:bar-cloud/baz",
|
||||
wantSource: ModelSourceUnspecified,
|
||||
wantStripped: "foo:bar-cloud/baz",
|
||||
},
|
||||
{
|
||||
name: "conflicting source suffixes",
|
||||
input: "foo:cloud:local",
|
||||
wantErr: ErrConflictingSourceSuffix,
|
||||
wantSource: ModelSourceUnspecified,
|
||||
},
|
||||
{
|
||||
name: "empty input",
|
||||
input: " ",
|
||||
wantErr: ErrModelRequired,
|
||||
wantSource: ModelSourceUnspecified,
|
||||
},
|
||||
}
|
||||
|
||||
for _, tt := range tests {
|
||||
t.Run(tt.name, func(t *testing.T) {
|
||||
got, err := ParseRef(tt.input)
|
||||
if tt.wantErr != nil {
|
||||
if !errors.Is(err, tt.wantErr) {
|
||||
t.Fatalf("ParseRef(%q) error = %v, want %v", tt.input, err, tt.wantErr)
|
||||
}
|
||||
return
|
||||
}
|
||||
if err != nil {
|
||||
t.Fatalf("ParseRef(%q) returned error: %v", tt.input, err)
|
||||
}
|
||||
|
||||
if got.Base != tt.wantBase {
|
||||
t.Fatalf("base = %q, want %q", got.Base, tt.wantBase)
|
||||
}
|
||||
|
||||
if got.Source != tt.wantSource {
|
||||
t.Fatalf("source = %v, want %v", got.Source, tt.wantSource)
|
||||
}
|
||||
|
||||
if HasExplicitCloudSource(tt.input) != tt.wantCloud {
|
||||
t.Fatalf("HasExplicitCloudSource(%q) = %v, want %v", tt.input, HasExplicitCloudSource(tt.input), tt.wantCloud)
|
||||
}
|
||||
|
||||
if HasExplicitLocalSource(tt.input) != tt.wantLocal {
|
||||
t.Fatalf("HasExplicitLocalSource(%q) = %v, want %v", tt.input, HasExplicitLocalSource(tt.input), tt.wantLocal)
|
||||
}
|
||||
|
||||
stripped, ok := StripCloudSourceTag(tt.input)
|
||||
if ok != tt.wantStripOK {
|
||||
t.Fatalf("StripCloudSourceTag(%q) ok = %v, want %v", tt.input, ok, tt.wantStripOK)
|
||||
}
|
||||
if stripped != tt.wantStripped {
|
||||
t.Fatalf("StripCloudSourceTag(%q) base = %q, want %q", tt.input, stripped, tt.wantStripped)
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
func TestNormalizePullName(t *testing.T) {
|
||||
tests := []struct {
|
||||
name string
|
||||
input string
|
||||
wantName string
|
||||
wantCloud bool
|
||||
wantErr error
|
||||
}{
|
||||
{
|
||||
name: "explicit local strips source",
|
||||
input: "gpt-oss:20b:local",
|
||||
wantName: "gpt-oss:20b",
|
||||
},
|
||||
{
|
||||
name: "explicit cloud with size maps to legacy dash cloud tag",
|
||||
input: "gpt-oss:20b:cloud",
|
||||
wantName: "gpt-oss:20b-cloud",
|
||||
wantCloud: true,
|
||||
},
|
||||
{
|
||||
name: "legacy cloud with size remains stable",
|
||||
input: "gpt-oss:20b-cloud",
|
||||
wantName: "gpt-oss:20b-cloud",
|
||||
wantCloud: true,
|
||||
},
|
||||
{
|
||||
name: "explicit cloud without tag maps to cloud tag",
|
||||
input: "qwen3:cloud",
|
||||
wantName: "qwen3:cloud",
|
||||
wantCloud: true,
|
||||
},
|
||||
{
|
||||
name: "host port without tag keeps host port and appends cloud tag",
|
||||
input: "localhost:11434/library/foo:cloud",
|
||||
wantName: "localhost:11434/library/foo:cloud",
|
||||
wantCloud: true,
|
||||
},
|
||||
{
|
||||
name: "conflicting source suffixes fail",
|
||||
input: "foo:cloud:local",
|
||||
wantErr: ErrConflictingSourceSuffix,
|
||||
},
|
||||
}
|
||||
|
||||
for _, tt := range tests {
|
||||
t.Run(tt.name, func(t *testing.T) {
|
||||
gotName, gotCloud, err := NormalizePullName(tt.input)
|
||||
if tt.wantErr != nil {
|
||||
if !errors.Is(err, tt.wantErr) {
|
||||
t.Fatalf("NormalizePullName(%q) error = %v, want %v", tt.input, err, tt.wantErr)
|
||||
}
|
||||
return
|
||||
}
|
||||
if err != nil {
|
||||
t.Fatalf("NormalizePullName(%q) returned error: %v", tt.input, err)
|
||||
}
|
||||
|
||||
if gotName != tt.wantName {
|
||||
t.Fatalf("normalized name = %q, want %q", gotName, tt.wantName)
|
||||
}
|
||||
if gotCloud != tt.wantCloud {
|
||||
t.Fatalf("cloud = %v, want %v", gotCloud, tt.wantCloud)
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
func TestParseSourceSuffix(t *testing.T) {
|
||||
tests := []struct {
|
||||
name string
|
||||
input string
|
||||
wantBase string
|
||||
wantSource ModelSource
|
||||
wantExplicit bool
|
||||
}{
|
||||
{
|
||||
name: "explicit cloud suffix",
|
||||
input: "gpt-oss:20b:cloud",
|
||||
wantBase: "gpt-oss:20b",
|
||||
wantSource: ModelSourceCloud,
|
||||
wantExplicit: true,
|
||||
},
|
||||
{
|
||||
name: "explicit local suffix",
|
||||
input: "qwen3:8b:local",
|
||||
wantBase: "qwen3:8b",
|
||||
wantSource: ModelSourceLocal,
|
||||
wantExplicit: true,
|
||||
},
|
||||
{
|
||||
name: "legacy cloud suffix on tag",
|
||||
input: "gpt-oss:20b-cloud",
|
||||
wantBase: "gpt-oss:20b",
|
||||
wantSource: ModelSourceCloud,
|
||||
wantExplicit: true,
|
||||
},
|
||||
{
|
||||
name: "legacy cloud suffix does not match model segment",
|
||||
input: "my-cloud-model",
|
||||
wantBase: "my-cloud-model",
|
||||
wantSource: ModelSourceUnspecified,
|
||||
wantExplicit: false,
|
||||
},
|
||||
{
|
||||
name: "legacy cloud suffix blocked when suffix includes slash",
|
||||
input: "foo:bar-cloud/baz",
|
||||
wantBase: "foo:bar-cloud/baz",
|
||||
wantSource: ModelSourceUnspecified,
|
||||
wantExplicit: false,
|
||||
},
|
||||
{
|
||||
name: "unknown suffix is not explicit source",
|
||||
input: "gpt-oss:clod",
|
||||
wantBase: "gpt-oss:clod",
|
||||
wantSource: ModelSourceUnspecified,
|
||||
wantExplicit: false,
|
||||
},
|
||||
{
|
||||
name: "uppercase suffix is accepted",
|
||||
input: "gpt-oss:20b:CLOUD",
|
||||
wantBase: "gpt-oss:20b",
|
||||
wantSource: ModelSourceCloud,
|
||||
wantExplicit: true,
|
||||
},
|
||||
{
|
||||
name: "no suffix",
|
||||
input: "llama3.2",
|
||||
wantBase: "llama3.2",
|
||||
wantSource: ModelSourceUnspecified,
|
||||
wantExplicit: false,
|
||||
},
|
||||
}
|
||||
|
||||
for _, tt := range tests {
|
||||
t.Run(tt.name, func(t *testing.T) {
|
||||
gotBase, gotSource, gotExplicit := parseSourceSuffix(tt.input)
|
||||
if gotBase != tt.wantBase {
|
||||
t.Fatalf("base = %q, want %q", gotBase, tt.wantBase)
|
||||
}
|
||||
if gotSource != tt.wantSource {
|
||||
t.Fatalf("source = %v, want %v", gotSource, tt.wantSource)
|
||||
}
|
||||
if gotExplicit != tt.wantExplicit {
|
||||
t.Fatalf("explicit = %v, want %v", gotExplicit, tt.wantExplicit)
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user