mirror of
https://github.com/diegosouzapw/OmniRoute.git
synced 2026-09-15 19:32:20 +03:00
65e81158ab
Landed with the design call resolved per the owner's pick — **option 1**: the synced store is now endpoint-agnostic (persistDiscoveredModels and managedModelImport no longer drop non-chat models at write time), and chat selectability moved to read time (auto-pool expansion in autoStrategy applies filterChatSelectableModels; the models-route projection already had its chatOnly filter). Your discovery test now passes end-to-end (3/3): /api/show capabilities persist per connection and image/embedding requests route through the advertising host. Reconciliation notes: conflicted areas merged onto the current tip (adobe discovery import, requestedModel preflight signature, resolvedProvider fast-path coexists with the synced-route override — explicit resolution wins); carried base-red drains (#10055 memoization, #11071 test variants) dropped as already-landed; the managed-model-import exclusion test was propagated to the new contract (image/video models persist; the read filter still hides them from chat pickers — pinned by a new assertion). Full battery: 205/206 focused (the one red is a confirmed periodic-timer timing flake on the loaded devbox — 20/20 isolated), autoCombo vitest 30/30, combo suites 46/46, gates + typecheck clean. Thank you @yourspraveen — the capability probe + routing design was right; it just needed the store contract opened up. Fixes #11087.
91 lines
3.7 KiB
TypeScript
91 lines
3.7 KiB
TypeScript
// tests/unit/combo-context-prefix-resolution.test.ts
|
|
// Regression guard for computeComboContextLength()'s provider-prefix bug.
|
|
//
|
|
// resolveNestedComboTargets() returns target.modelStr in "provider/model" form
|
|
// (e.g. "glm/glm-5.2"), but computeComboContextLength() used to pass that
|
|
// qualified string straight into getCanonicalModelMetadata() without stripping
|
|
// the prefix first — unlike the catalog's own getComboTargetCatalogMetadata(),
|
|
// which strips it via getComboTargetModelId()/getProviderPrefixes() before the
|
|
// lookup. The alias-resolution chain (getResolvedModelCapabilities ->
|
|
// resolveCanonicalProviderModel -> resolveProviderModelAlias) does an exact-match
|
|
// lookup keyed by the BARE registry id, so a "provider/model" string only
|
|
// resolved for the handful of models with a curated MODEL_SPECS alias in that
|
|
// exact qualified form — every other registry-only model (the vast majority)
|
|
// silently fell out of the min() computation, and computed_context_length was
|
|
// dropped from the /api/combos response entirely.
|
|
//
|
|
// This test uses glm-5.2 (open-sse/config/providers/registry/glm — real
|
|
// 1,000,000-token context, no "glm/glm-5.2"-form curated alias) — the exact
|
|
// class of model the bug affected. Confirmed empirically before this fix
|
|
// landed: computeComboContextLength() returned `undefined` for a combo whose
|
|
// only member was "glm/glm-5.2", even though the bare "glm-5.2" resolves fine
|
|
// via getCanonicalModelMetadata() directly.
|
|
import test from "node:test";
|
|
import assert from "node:assert/strict";
|
|
import fs from "node:fs";
|
|
import os from "node:os";
|
|
import path from "node:path";
|
|
|
|
const TEST_DATA_DIR = fs.mkdtempSync(
|
|
path.join(os.tmpdir(), "omniroute-combo-context-prefix-resolution-")
|
|
);
|
|
process.env.DATA_DIR = TEST_DATA_DIR;
|
|
|
|
const core = await import("../../src/lib/db/core.ts");
|
|
const { computeComboContextLength } = await import("../../src/lib/combos/comboContext.ts");
|
|
const { setModelContextOverride, removeModelContextOverride } =
|
|
await import("../../src/lib/db/modelContextOverrides.ts");
|
|
|
|
test.after(() => {
|
|
core.resetDbInstance();
|
|
fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
|
|
});
|
|
|
|
test("computeComboContextLength resolves a registry-known, prefixed member (glm/glm-5.2) to its real context window", () => {
|
|
const combo = {
|
|
name: "prefix-resolution-probe-single",
|
|
models: ["glm/glm-5.2"],
|
|
};
|
|
|
|
const result = computeComboContextLength(combo, []);
|
|
|
|
assert.equal(
|
|
result,
|
|
1000000,
|
|
"glm/glm-5.2 is a real registry model with a 1,000,000-token context window " +
|
|
"(open-sse/config/providers/registry/glm) — the prefix must be stripped " +
|
|
"before the canonical-model lookup so it is not silently excluded"
|
|
);
|
|
});
|
|
|
|
test("computeComboContextLength takes the minimum across multiple prefixed, registry-known members", () => {
|
|
const combo = {
|
|
name: "prefix-resolution-probe-multi",
|
|
// glm-4.5 (128,000) is the smaller of the two known windows.
|
|
models: ["glm/glm-5.2", "glm/glm-4.5"],
|
|
};
|
|
|
|
const result = computeComboContextLength(combo, []);
|
|
|
|
assert.equal(
|
|
result,
|
|
128000,
|
|
"the minimum across all resolved (prefix-stripped) members should win, " +
|
|
"matching the catalog's minKnownNumber semantics"
|
|
);
|
|
});
|
|
|
|
test("computeComboContextLength honors a larger persisted Codex GPT-5.6 window", () => {
|
|
const modelId = "gpt-5.6-terra";
|
|
assert.equal(setModelContextOverride("codex", modelId, 500000, "manual"), true);
|
|
try {
|
|
assert.equal(
|
|
computeComboContextLength({ models: [`codex/${modelId}`] }, []),
|
|
500000,
|
|
"the combo aggregate must use the effective override instead of the Codex registry default"
|
|
);
|
|
} finally {
|
|
removeModelContextOverride("codex", modelId);
|
|
}
|
|
});
|