mirror of
https://github.com/lukaszraczylo/traefik-llmgateway.git
synced 2026-10-21 05:57:53 +00:00
fix(modelmeta): free-tier suffix layer (review IMPORTANT-1)
A ":free"-suffixed model id (a real provider naming convention, e.g. OpenRouter's own live API) is free by naming convention alone, independent of whatever the built-in table does or does not track for that exact id. resolveModelMeta's cost switch gains a new case, ranked above the built-in table and below an operator's own config override — so a provider's free-tier naming is never shadowed by a stale or coincidental paid built-in price under the identical key, and an operator can still override it explicitly if they choose. Context is untouched by this rule: it keeps resolving through the normal config/ discovery/builtin/absent layers, never inherited from an unrelated entry. Also makes resolveModelMeta's own Free case assign 0, 0, true explicitly (matching resolveAliasModelMeta's existing style) instead of relying on the zero-value default — defensive consistency, folded review minor. Independently reviewed and reproduced against the same live LiteLLM fetch: zero ":free"-suffixed keys exist anywhere in the 3111-entry upstream file, confirming the earlier grounding note in the v0.23 report. This layer is a general resolution rule for a naming convention provider APIs use in live discovery, not something the pruned build-time table needs to carry any entries for.
This commit is contained in:
+25
-4
@@ -3,6 +3,7 @@ package traefikllmgateway
|
|||||||
import (
|
import (
|
||||||
"fmt"
|
"fmt"
|
||||||
"sort"
|
"sort"
|
||||||
|
"strings"
|
||||||
)
|
)
|
||||||
|
|
||||||
// microUSDPerUSD scales a micro-USD integer (ModelMetaConfig/
|
// microUSDPerUSD scales a micro-USD integer (ModelMetaConfig/
|
||||||
@@ -15,6 +16,17 @@ import (
|
|||||||
// not be conflated just because they share a value.
|
// not be conflated just because they share a value.
|
||||||
const microUSDPerUSD = 1_000_000
|
const microUSDPerUSD = 1_000_000
|
||||||
|
|
||||||
|
// freeTierSuffix marks a discovered model id as free by naming
|
||||||
|
// convention alone (review addendum, feature v0.23): a real provider
|
||||||
|
// (OpenRouter's own live API is the known case) reports free-tier model
|
||||||
|
// ids with this exact suffix, independent of whatever LiteLLM's static
|
||||||
|
// pricing file does or does not track for that provider. resolveModelMeta
|
||||||
|
// checks this AFTER an operator's own config override and BEFORE the
|
||||||
|
// built-in table, so a ":free"-suffixed id never falls through to a
|
||||||
|
// builtin entry's paid price for that same key, and an operator can
|
||||||
|
// still override it (e.g. to attach a context window) if they choose.
|
||||||
|
const freeTierSuffix = ":free"
|
||||||
|
|
||||||
// resolvedModelMeta is resolveModelMeta's result: a model's context
|
// resolvedModelMeta is resolveModelMeta's result: a model's context
|
||||||
// window and per-token cost, each resolved independently through its own
|
// window and per-token cost, each resolved independently through its own
|
||||||
// layering (feature v0.23). ContextKnown/CostKnown are authoritative — a
|
// layering (feature v0.23). ContextKnown/CostKnown are authoritative — a
|
||||||
@@ -44,12 +56,19 @@ type resolvedModelMeta struct {
|
|||||||
// names, so an exact "provider/model" form makes no sense for that
|
// names, so an exact "provider/model" form makes no sense for that
|
||||||
// table the way it does for cfg; otherwise unknown.
|
// table the way it does for cfg; otherwise unknown.
|
||||||
// - Cost: the config override wins when it sets Free (an explicit,
|
// - Cost: the config override wins when it sets Free (an explicit,
|
||||||
// known zero) or either cost field > 0; otherwise the built-in
|
// known zero) or either cost field > 0; otherwise a ":free"-suffixed
|
||||||
// table. Discovery never reports cost, so there is no discovery
|
// model id (freeTierSuffix — review addendum) wins next, ahead of
|
||||||
// layer for it.
|
// the built-in table, so a provider's own free-tier naming
|
||||||
|
// convention is never shadowed by a paid built-in price under the
|
||||||
|
// identical key; otherwise the built-in table. Discovery never
|
||||||
|
// reports cost, so there is no discovery layer for it.
|
||||||
//
|
//
|
||||||
// Absent at every applicable layer is a genuinely unknown value and is
|
// Absent at every applicable layer is a genuinely unknown value and is
|
||||||
// never invented — ContextKnown/CostKnown both stay false in that case.
|
// never invented — ContextKnown/CostKnown both stay false in that case.
|
||||||
|
// The freeTierSuffix rule affects ONLY cost: a ":free" id's context
|
||||||
|
// still runs through the normal context layering above (config/
|
||||||
|
// discovery/builtin/absent), never inherited from some other, unrelated
|
||||||
|
// entry.
|
||||||
func resolveModelMeta(provider, model string, cfg map[string]*ModelMetaConfig, discoveredContext int, discoveredKnown bool) resolvedModelMeta {
|
func resolveModelMeta(provider, model string, cfg map[string]*ModelMetaConfig, discoveredContext int, discoveredKnown bool) resolvedModelMeta {
|
||||||
var out resolvedModelMeta
|
var out resolvedModelMeta
|
||||||
override := lookupModelMetaConfig(provider, model, cfg)
|
override := lookupModelMetaConfig(provider, model, cfg)
|
||||||
@@ -66,11 +85,13 @@ func resolveModelMeta(provider, model string, cfg map[string]*ModelMetaConfig, d
|
|||||||
|
|
||||||
switch {
|
switch {
|
||||||
case override != nil && override.Free:
|
case override != nil && override.Free:
|
||||||
out.CostKnown = true // both cost fields stay 0 — explicit, known zero
|
out.InputCostPerMTokMicroUSD, out.OutputCostPerMTokMicroUSD, out.CostKnown = 0, 0, true
|
||||||
case override != nil && (override.InputCostPerMTokMicroUSD > 0 || override.OutputCostPerMTokMicroUSD > 0):
|
case override != nil && (override.InputCostPerMTokMicroUSD > 0 || override.OutputCostPerMTokMicroUSD > 0):
|
||||||
out.InputCostPerMTokMicroUSD = override.InputCostPerMTokMicroUSD
|
out.InputCostPerMTokMicroUSD = override.InputCostPerMTokMicroUSD
|
||||||
out.OutputCostPerMTokMicroUSD = override.OutputCostPerMTokMicroUSD
|
out.OutputCostPerMTokMicroUSD = override.OutputCostPerMTokMicroUSD
|
||||||
out.CostKnown = true
|
out.CostKnown = true
|
||||||
|
case strings.HasSuffix(model, freeTierSuffix):
|
||||||
|
out.InputCostPerMTokMicroUSD, out.OutputCostPerMTokMicroUSD, out.CostKnown = 0, 0, true
|
||||||
case builtinOK && (builtin.InputCostPerMTokMicroUSD > 0 || builtin.OutputCostPerMTokMicroUSD > 0):
|
case builtinOK && (builtin.InputCostPerMTokMicroUSD > 0 || builtin.OutputCostPerMTokMicroUSD > 0):
|
||||||
out.InputCostPerMTokMicroUSD = builtin.InputCostPerMTokMicroUSD
|
out.InputCostPerMTokMicroUSD = builtin.InputCostPerMTokMicroUSD
|
||||||
out.OutputCostPerMTokMicroUSD = builtin.OutputCostPerMTokMicroUSD
|
out.OutputCostPerMTokMicroUSD = builtin.OutputCostPerMTokMicroUSD
|
||||||
|
|||||||
@@ -126,6 +126,69 @@ func TestResolveModelMeta_BuiltinTable(t *testing.T) {
|
|||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
// TestResolveModelMeta_FreeTierSuffix covers freeTierSuffix's layer
|
||||||
|
// (review addendum): a ":free"-suffixed id is free by naming convention
|
||||||
|
// alone, ranked above the built-in table but below an operator's own
|
||||||
|
// config override, and never invents a context window.
|
||||||
|
func TestResolveModelMeta_FreeTierSuffix(t *testing.T) {
|
||||||
|
t.Run("free-by-suffix wins over a builtin paid entry under the identical key", func(t *testing.T) {
|
||||||
|
const id = "review-fixture-model:free"
|
||||||
|
t.Cleanup(func() { delete(builtinModelMetaTable, id) })
|
||||||
|
builtinModelMetaTable[id] = builtinModelMeta{ContextTokens: 999, InputCostPerMTokMicroUSD: 5_000_000, OutputCostPerMTokMicroUSD: 5_000_000}
|
||||||
|
|
||||||
|
got := resolveModelMeta("openrouter", id, nil, 0, false)
|
||||||
|
if !got.CostKnown || got.InputCostPerMTokMicroUSD != 0 || got.OutputCostPerMTokMicroUSD != 0 {
|
||||||
|
t.Errorf("cost = (%d, %d, known=%v), want (0, 0, true) — free-by-suffix must beat a builtin paid entry", got.InputCostPerMTokMicroUSD, got.OutputCostPerMTokMicroUSD, got.CostKnown)
|
||||||
|
}
|
||||||
|
// Context still resolves independently — the builtin entry's own
|
||||||
|
// ContextTokens is a real, known value for this exact key, so it
|
||||||
|
// is NOT suppressed by the free-tier cost rule (freeTierSuffix
|
||||||
|
// affects cost only).
|
||||||
|
if !got.ContextKnown || got.ContextTokens != 999 {
|
||||||
|
t.Errorf("context = (%d, known=%v), want (999, true) — freeTierSuffix must not touch context resolution", got.ContextTokens, got.ContextKnown)
|
||||||
|
}
|
||||||
|
})
|
||||||
|
|
||||||
|
t.Run("free-by-suffix never invents context when nothing else knows it", func(t *testing.T) {
|
||||||
|
got := resolveModelMeta("openrouter", "some-model-nobody-prices:free", nil, 0, false)
|
||||||
|
if got.ContextKnown {
|
||||||
|
t.Errorf("context = %+v, want unknown — free-by-suffix must never invent a context window", got)
|
||||||
|
}
|
||||||
|
if !got.CostKnown || got.InputCostPerMTokMicroUSD != 0 || got.OutputCostPerMTokMicroUSD != 0 {
|
||||||
|
t.Errorf("cost = %+v, want known zero", got)
|
||||||
|
}
|
||||||
|
})
|
||||||
|
|
||||||
|
t.Run("an explicit operator override still beats free-by-suffix", func(t *testing.T) {
|
||||||
|
const id = "review-fixture-model-2:free"
|
||||||
|
cfg := map[string]*ModelMetaConfig{id: {InputCostPerMTokMicroUSD: 1, OutputCostPerMTokMicroUSD: 2}}
|
||||||
|
got := resolveModelMeta("openrouter", id, cfg, 0, false)
|
||||||
|
if !got.CostKnown || got.InputCostPerMTokMicroUSD != 1 || got.OutputCostPerMTokMicroUSD != 2 {
|
||||||
|
t.Errorf("cost = (%d, %d, known=%v), want (1, 2, true) — an explicit operator override must win over the free-by-suffix rule", got.InputCostPerMTokMicroUSD, got.OutputCostPerMTokMicroUSD, got.CostKnown)
|
||||||
|
}
|
||||||
|
})
|
||||||
|
|
||||||
|
t.Run("a non-free config override still beats free-by-suffix even at zero context", func(t *testing.T) {
|
||||||
|
const id = "review-fixture-model-3:free"
|
||||||
|
cfg := map[string]*ModelMetaConfig{id: {Free: false, InputCostPerMTokMicroUSD: 0, OutputCostPerMTokMicroUSD: 0}}
|
||||||
|
// A present-but-all-zero, non-Free override entry sets nothing at
|
||||||
|
// this layer (matches ModelMetaConfig's own "zero falls through"
|
||||||
|
// contract) — the free-tier suffix rule still applies underneath
|
||||||
|
// it, not the (absent) override.
|
||||||
|
got := resolveModelMeta("openrouter", id, cfg, 0, false)
|
||||||
|
if !got.CostKnown || got.InputCostPerMTokMicroUSD != 0 {
|
||||||
|
t.Errorf("cost = %+v, want known zero via the free-by-suffix fallback", got)
|
||||||
|
}
|
||||||
|
})
|
||||||
|
|
||||||
|
t.Run("an id without the suffix is unaffected", func(t *testing.T) {
|
||||||
|
got := resolveModelMeta("openrouter", "plain-model-id", nil, 0, false)
|
||||||
|
if got.CostKnown {
|
||||||
|
t.Errorf("cost = %+v, want unknown for a plain id with no :free suffix and no other layer", got)
|
||||||
|
}
|
||||||
|
})
|
||||||
|
}
|
||||||
|
|
||||||
// TestLookupModelMetaConfig covers lookupModelMetaConfig's own
|
// TestLookupModelMetaConfig covers lookupModelMetaConfig's own
|
||||||
// precedence and nil-safety directly.
|
// precedence and nil-safety directly.
|
||||||
func TestLookupModelMetaConfig(t *testing.T) {
|
func TestLookupModelMetaConfig(t *testing.T) {
|
||||||
|
|||||||
Reference in New Issue
Block a user