fix(picker): catalog is authority for @custom labels; keep plain-lane colon models whole
The #7240 fix peeled the leading @custom: segment on first-colon split, but a plain-lane custom id whose model carries a tag/variant colon (@custom:ollamacloud/qwen3.5:397b) was still misread as slug=ollamacloud/qwen3.5, rendering only "397b". Two changes: - _dynamicModelLabels (the backend-populated catalog) is now consulted BEFORE the string peel: the backend knows the true provider/model split and ships an exact m.label per routing id, so catalogued ids — including plain-lane colon models and collision cases — render verbatim. - The legacy string fallback now treats a slash-bearing first segment as the model itself (a provider slug is a config key or host:port authority and never contains '/'), mirroring the '/' means routable rule api/config.py applies when building ids. Tests: add plain-lane @custom:ollamacloud/qwen3.5:397b controls, a post-catalog dynamic-label control, and a collision case proving labels stay distinct by full routing id. Driver now accepts an optional pre-hydrated catalog map.
This commit is contained in:
1 parent
e71c3e1f97
commit
da4e4ef316
2 files changed
+73
-8
No files matched your search
+15
-2
@@ -7386,6 +7386,15 @@ function _stripDottedModelPrefix(bare){
|
|||||||
function getModelLabel(modelId){
|
function getModelLabel(modelId){
|
||||||
if(!modelId) return 'Unknown';
|
if(!modelId) return 'Unknown';
|
||||||
const rawId=String(modelId||'');
|
const rawId=String(modelId||'');
|
||||||
|
// The catalog is the authority on model identity: the backend knows the real
|
||||||
|
// provider/model split and ships an exact `m.label` per routing id, so a
|
||||||
|
// catalogued id — including a plain-lane `@custom:` id whose model contains
|
||||||
|
// colons (`@custom:ollamacloud/qwen3.5:397b`) — renders verbatim. The
|
||||||
|
// string parsing below is only a legacy fallback for ids the catalog has
|
||||||
|
// never seen (pre-hydration, stale sessions, removed providers), where a
|
||||||
|
// first-colon split alone cannot tell a `@custom:<slug>:<model>` from a
|
||||||
|
// plain-lane `@custom:<model-with-colon>` (#7240).
|
||||||
|
if(_dynamicModelLabels[modelId]) return _dynamicModelLabels[modelId];
|
||||||
// Preserve custom gateway model IDs exactly as configured. A custom id is
|
// Preserve custom gateway model IDs exactly as configured. A custom id is
|
||||||
// `@custom:<model>` in the plain custom lane or `@custom:<slug>:<model>` for
|
// `@custom:<model>` in the plain custom lane or `@custom:<slug>:<model>` for
|
||||||
// a named custom provider; the provider slug may itself be an endpoint
|
// a named custom provider; the provider slug may itself be an endpoint
|
||||||
@@ -7400,6 +7409,12 @@ function getModelLabel(modelId){
|
|||||||
const rest=rawId.slice('@custom:'.length);
|
const rest=rawId.slice('@custom:'.length);
|
||||||
const sep=rest.indexOf(':');
|
const sep=rest.indexOf(':');
|
||||||
if(sep<0) return rest||rawId;
|
if(sep<0) return rest||rawId;
|
||||||
|
// A provider slug is a config key or a host:port authority — it never
|
||||||
|
// contains a `/`. A slash-bearing first segment is therefore the model
|
||||||
|
// itself in the plain custom lane (`@custom:ollamacloud/qwen3.5:397b`
|
||||||
|
// must render the whole remainder, not just `397b`), mirroring the
|
||||||
|
// `/`-means-routable rule api/config.py applies when building ids.
|
||||||
|
if(rest.slice(0,sep).includes('/')) return rest||rawId;
|
||||||
let model=rest.slice(sep+1);
|
let model=rest.slice(sep+1);
|
||||||
// Endpoint-style slug (`custom:10.8.71.41:8080:model`): the `:port` belongs
|
// Endpoint-style slug (`custom:10.8.71.41:8080:model`): the `:port` belongs
|
||||||
// to the provider segment, mirroring the host:port slug check in
|
// to the provider segment, mirroring the host:port slug check in
|
||||||
@@ -7416,8 +7431,6 @@ function getModelLabel(modelId){
|
|||||||
}
|
}
|
||||||
return model||rawId;
|
return model||rawId;
|
||||||
}
|
}
|
||||||
// Check dynamic labels first, then fall back to splitting the ID
|
|
||||||
if(_dynamicModelLabels[modelId]) return _dynamicModelLabels[modelId];
|
|
||||||
// Static fallback for common models
|
// Static fallback for common models
|
||||||
const STATIC_LABELS={'openai/gpt-5.4-mini':'GPT-5.4 Mini','openai/gpt-4o':'GPT-4o','openai/o3':'o3','openai/o4-mini':'o4-mini','anthropic/claude-sonnet-4.6':'Sonnet 4.6','anthropic/claude-sonnet-4-5':'Sonnet 4.5','anthropic/claude-haiku-3-5':'Haiku 3.5','google/gemini-3.1-pro-preview':'Gemini 3.1 Pro','google/gemini-3-flash-preview':'Gemini 3 Flash','google/gemini-3.1-flash-lite-preview':'Gemini 3.1 Flash Lite','google/gemini-2.5-pro':'Gemini 2.5 Pro','google/gemini-2.5-flash':'Gemini 2.5 Flash','deepseek/deepseek-v4-flash':'DeepSeek V4 Flash','deepseek/deepseek-v4-pro':'DeepSeek V4 Pro','deepseek/deepseek-chat-v3-0324':'DeepSeek V3 (legacy)','meta-llama/llama-4-scout':'Llama 4 Scout'};
|
const STATIC_LABELS={'openai/gpt-5.4-mini':'GPT-5.4 Mini','openai/gpt-4o':'GPT-4o','openai/o3':'o3','openai/o4-mini':'o4-mini','anthropic/claude-sonnet-4.6':'Sonnet 4.6','anthropic/claude-sonnet-4-5':'Sonnet 4.5','anthropic/claude-haiku-3-5':'Haiku 3.5','google/gemini-3.1-pro-preview':'Gemini 3.1 Pro','google/gemini-3-flash-preview':'Gemini 3 Flash','google/gemini-3.1-flash-lite-preview':'Gemini 3.1 Flash Lite','google/gemini-2.5-pro':'Gemini 2.5 Pro','google/gemini-2.5-flash':'Gemini 2.5 Flash','deepseek/deepseek-v4-flash':'DeepSeek V4 Flash','deepseek/deepseek-v4-pro':'DeepSeek V4 Pro','deepseek/deepseek-chat-v3-0324':'DeepSeek V3 (legacy)','meta-llama/llama-4-scout':'Llama 4 Scout'};
|
||||||
if(STATIC_LABELS[modelId]) return STATIC_LABELS[modelId];
|
if(STATIC_LABELS[modelId]) return STATIC_LABELS[modelId];
|
||||||
|
|||||||
@@ -6,9 +6,11 @@ following the LAST ``:``, so a tag/variant suffix like ``:free``/``:31b``/
|
|||||||
``:397b`` inside the model's own name became the whole label
|
``:397b`` inside the model's own name became the whole label
|
||||||
(``@custom:omni:kg/stepfun/step-3.7-flash:free`` rendered as ``free``).
|
(``@custom:omni:kg/stepfun/step-3.7-flash:free`` rendered as ``free``).
|
||||||
|
|
||||||
The fix peels only the leading provider segment (``@custom:<slug>:`` — the
|
The fix makes the catalog the authority — a catalogued routing id renders its
|
||||||
``<slug>`` is itself optional in the plain custom lane, and may be an endpoint
|
exact `m.label` (the backend knows the real provider/model split). The string
|
||||||
authority ``host:port``), leaving the full configured model id as the label.
|
peel is only a legacy fallback for ids the catalog has never seen, where a
|
||||||
|
provider slug (a config key or host:port authority) is distinguished from a
|
||||||
|
plain-lane model by its lack of a `/`.
|
||||||
|
|
||||||
Runs the live getModelLabel() via Node so drift between the test and the real
|
Runs the live getModelLabel() via Node so drift between the test and the real
|
||||||
code is caught immediately.
|
code is caught immediately.
|
||||||
@@ -36,7 +38,9 @@ if (start < 0) throw new Error('getModelLabel not found');
|
|||||||
const after = ui.indexOf('\nfunction _gatewayProviderName(', start);
|
const after = ui.indexOf('\nfunction _gatewayProviderName(', start);
|
||||||
if (after < 0) throw new Error('getModelLabel end boundary not found');
|
if (after < 0) throw new Error('getModelLabel end boundary not found');
|
||||||
const fnSrc = ui.slice(start, after);
|
const fnSrc = ui.slice(start, after);
|
||||||
const _dynamicModelLabels = {};
|
// Optional pre-hydrated catalog: {routingId: label}. Absent -> empty, which
|
||||||
|
// exercises the string-only legacy fallback.
|
||||||
|
const _dynamicModelLabels = process.argv[3] ? JSON.parse(process.argv[3]) : {};
|
||||||
function _fmtOllamaLabel(s){ return s; }
|
function _fmtOllamaLabel(s){ return s; }
|
||||||
// getModelLabel() calls the dotted Bedrock/Vertex prefix normalizer, which lives
|
// getModelLabel() calls the dotted Bedrock/Vertex prefix normalizer, which lives
|
||||||
// just above it with two Sets it closes over. Pull all three in, or the eval'd
|
// just above it with two Sets it closes over. Pull all three in, or the eval'd
|
||||||
@@ -52,9 +56,12 @@ process.stdout.write(JSON.stringify(out));
|
|||||||
"""
|
"""
|
||||||
|
|
||||||
|
|
||||||
def _labels(model_ids):
|
def _labels(model_ids, catalog=None):
|
||||||
|
argv = [NODE, "-e", _DRIVER, str(UI_JS_PATH), json.dumps(model_ids)]
|
||||||
|
if catalog is not None:
|
||||||
|
argv.append(json.dumps(catalog))
|
||||||
proc = subprocess.run(
|
proc = subprocess.run(
|
||||||
[NODE, "-e", _DRIVER, str(UI_JS_PATH), json.dumps(model_ids)],
|
argv,
|
||||||
capture_output=True, text=True, timeout=30,
|
capture_output=True, text=True, timeout=30,
|
||||||
)
|
)
|
||||||
assert proc.returncode == 0, f"node driver failed: {proc.stderr}"
|
assert proc.returncode == 0, f"node driver failed: {proc.stderr}"
|
||||||
@@ -100,3 +107,48 @@ def test_custom_model_label_plain_lane_and_host_port_slugs():
|
|||||||
assert out["@custom:10.8.71.41:8080:Qwen3-235B"] == "Qwen3-235B"
|
assert out["@custom:10.8.71.41:8080:Qwen3-235B"] == "Qwen3-235B"
|
||||||
assert out["@custom:localhost:11434:llama3"] == "llama3"
|
assert out["@custom:localhost:11434:llama3"] == "llama3"
|
||||||
assert out["@custom:omni:11434:Qwen3"] == "11434:Qwen3"
|
assert out["@custom:omni:11434:Qwen3"] == "11434:Qwen3"
|
||||||
|
|
||||||
|
|
||||||
|
def test_plain_custom_lane_colon_bearing_model_slash_segment():
|
||||||
|
"""#7240: a plain-lane ``@custom:<model-with-colon>`` whose first segment
|
||||||
|
is slash-bearing (`ollamacloud/qwen3.5:397b`) is the model itself — a
|
||||||
|
provider slug never contains a `/`, so it must not be peeled as one."""
|
||||||
|
out = _labels([
|
||||||
|
"@custom:ollamacloud/qwen3.5:397b",
|
||||||
|
"@custom:ollamacloud/gemma4:31b",
|
||||||
|
"@custom:some-vendor/step-3.7-flash:free",
|
||||||
|
])
|
||||||
|
assert out["@custom:ollamacloud/qwen3.5:397b"] == "ollamacloud/qwen3.5:397b"
|
||||||
|
assert out["@custom:ollamacloud/gemma4:31b"] == "ollamacloud/gemma4:31b"
|
||||||
|
assert out["@custom:some-vendor/step-3.7-flash:free"] == "some-vendor/step-3.7-flash:free"
|
||||||
|
|
||||||
|
|
||||||
|
def test_catalog_label_wins_over_legacy_peel():
|
||||||
|
"""Post-hydration: the catalog ships the exact backend label per routing
|
||||||
|
id, so a catalogued id renders verbatim even where the string peel alone
|
||||||
|
would be ambiguous or lossy."""
|
||||||
|
catalog = {
|
||||||
|
# Plain lane, colon-bearing model: the backend knows this is ONE model.
|
||||||
|
"@custom:ollamacloud/qwen3.5:397b": "ollamacloud/qwen3.5:397b",
|
||||||
|
# Named lane, same native model: distinct routing id, distinct entry.
|
||||||
|
"@custom:omni:ollamacloud/qwen3.5:397b": "ollamacloud/qwen3.5:397b",
|
||||||
|
# A custom display label is preserved exactly.
|
||||||
|
"@custom:ai_gateway:Qwen3.6-35B-A3B": "My Qwen Gateway",
|
||||||
|
}
|
||||||
|
out = _labels(list(catalog), catalog=catalog)
|
||||||
|
assert out["@custom:ollamacloud/qwen3.5:397b"] == "ollamacloud/qwen3.5:397b"
|
||||||
|
assert out["@custom:omni:ollamacloud/qwen3.5:397b"] == "ollamacloud/qwen3.5:397b"
|
||||||
|
assert out["@custom:ai_gateway:Qwen3.6-35B-A3B"] == "My Qwen Gateway"
|
||||||
|
|
||||||
|
|
||||||
|
def test_collision_labels_stay_distinct_by_full_routing_id():
|
||||||
|
"""Two providers exposing the same native model stay distinguishable: the
|
||||||
|
catalog keys labels by the FULL routing id, never by the bare model."""
|
||||||
|
catalog = {
|
||||||
|
"@custom:omni:ollamacloud/qwen3.5:397b": "ollamacloud/qwen3.5:397b (omni)",
|
||||||
|
"@custom:ai_gateway:ollamacloud/qwen3.5:397b": "ollamacloud/qwen3.5:397b (gateway)",
|
||||||
|
}
|
||||||
|
out = _labels(list(catalog), catalog=catalog)
|
||||||
|
assert out["@custom:omni:ollamacloud/qwen3.5:397b"] == "ollamacloud/qwen3.5:397b (omni)"
|
||||||
|
assert out["@custom:ai_gateway:ollamacloud/qwen3.5:397b"] == "ollamacloud/qwen3.5:397b (gateway)"
|
||||||
|
assert out["@custom:omni:ollamacloud/qwen3.5:397b"] != out["@custom:ai_gateway:ollamacloud/qwen3.5:397b"]
|
||||||
Reference in new issue
Block a user