a993eb3 Avoid unverified Bedrock output limits · zut

M README.md +1 -1


@@ -798,7 +798,7 @@ If both are present, the bearer token wins for inference, while live model disco
798	798		Unlike most providers, Bedrock model pricing, context windows, and capability flags are hand-maintained in zot's baked-in catalog. AWS does not expose usable pricing programmatically for current models — its Price List API returns only stale, input-only rates for a few legacy models — so the catalog is authoritative for cost and metadata. This means:
799	799
800	800		- Bearer-token-only users still get the full curated Bedrock model list with correct pricing from the catalog. No AWS API listing is attempted.
801		-	- Users with SigV4 credentials additionally get live model discovery, even if inference uses a bearer token: on startup, zot lists the foundation models and cross-region inference profiles your account can actually use (via `ListFoundationModels` and `ListInferenceProfiles`) and merges them into the catalog. New Bedrock models therefore appear in `/model` without waiting for a zot release. Discovered IDs already known to the catalog keep their curated metadata. An inference profile the catalog doesn't list under that exact ID (for example `apac.` or `us-gov.` variants) inherits the limits, reasoning flags, and pricing of its base model. Claude models that aren't in the catalog at all get a 64k output / 200k context default; other unknown IDs show up with placeholder pricing and default limits until the catalog is updated.
	801	+	- Users with SigV4 credentials additionally get live model discovery, even if inference uses a bearer token: on startup, zot lists the foundation models and cross-region inference profiles your account can actually use (via `ListFoundationModels` and `ListInferenceProfiles`) and merges them into the catalog. New Bedrock models therefore appear in `/model` without waiting for a zot release. Discovered IDs already known to the catalog keep their curated metadata. An inference profile the catalog doesn't list under that exact ID (for example `apac.` or `us-gov.` variants) inherits the limits, reasoning flags, and pricing of its base model. Models with no matching base catalog entry, including new Claude models, retain placeholder pricing and the client's 4096-token output fallback until their limits are verified and added to the catalog.
802	802
803	803		Live discovery requires SigV4 because the listing APIs live on the Bedrock control plane, which rejects the runtime bearer token. It runs in the background at startup and is cached for 6h like other providers, so `/model` reflects the latest list without any manual refresh. If discovery fails (no SigV4 creds, expired SSO login, missing IAM permissions), zot silently falls back to the catalog. Discovery needs `bedrock:ListFoundationModels` and `bedrock:ListInferenceProfiles`.
804	804


@@ -798,7 +798,7 @@ If both are present, the bearer token wins for inference, while live model disco
798		Unlike most providers, Bedrock model pricing, context windows, and capability flags are hand-maintained in zot's baked-in catalog. AWS does not expose usable pricing programmatically for current models — its Price List API returns only stale, input-only rates for a few legacy models — so the catalog is authoritative for cost and metadata. This means:	798		Unlike most providers, Bedrock model pricing, context windows, and capability flags are hand-maintained in zot's baked-in catalog. AWS does not expose usable pricing programmatically for current models — its Price List API returns only stale, input-only rates for a few legacy models — so the catalog is authoritative for cost and metadata. This means:
799			799
800		- Bearer-token-only users still get the full curated Bedrock model list with correct pricing from the catalog. No AWS API listing is attempted.	800		- Bearer-token-only users still get the full curated Bedrock model list with correct pricing from the catalog. No AWS API listing is attempted.
801	-	- Users with SigV4 credentials additionally get live model discovery, even if inference uses a bearer token: on startup, zot lists the foundation models and cross-region inference profiles your account can actually use (via `ListFoundationModels` and `ListInferenceProfiles`) and merges them into the catalog. New Bedrock models therefore appear in `/model` without waiting for a zot release. Discovered IDs already known to the catalog keep their curated metadata. An inference profile the catalog doesn't list under that exact ID (for example `apac.` or `us-gov.` variants) inherits the limits, reasoning flags, and pricing of its base model. Claude models that aren't in the catalog at all get a 64k output / 200k context default; other unknown IDs show up with placeholder pricing and default limits until the catalog is updated.	801	+	- Users with SigV4 credentials additionally get live model discovery, even if inference uses a bearer token: on startup, zot lists the foundation models and cross-region inference profiles your account can actually use (via `ListFoundationModels` and `ListInferenceProfiles`) and merges them into the catalog. New Bedrock models therefore appear in `/model` without waiting for a zot release. Discovered IDs already known to the catalog keep their curated metadata. An inference profile the catalog doesn't list under that exact ID (for example `apac.` or `us-gov.` variants) inherits the limits, reasoning flags, and pricing of its base model. Models with no matching base catalog entry, including new Claude models, retain placeholder pricing and the client's 4096-token output fallback until their limits are verified and added to the catalog.
802			802
803		Live discovery requires SigV4 because the listing APIs live on the Bedrock control plane, which rejects the runtime bearer token. It runs in the background at startup and is cached for 6h like other providers, so `/model` reflects the latest list without any manual refresh. If discovery fails (no SigV4 creds, expired SSO login, missing IAM permissions), zot silently falls back to the catalog. Discovery needs `bedrock:ListFoundationModels` and `bedrock:ListInferenceProfiles`.	803		Live discovery requires SigV4 because the listing APIs live on the Bedrock control plane, which rejects the runtime bearer token. It runs in the background at startup and is cached for 6h like other providers, so `/model` reflects the latest list without any manual refresh. If discovery fails (no SigV4 creds, expired SSO login, missing IAM permissions), zot silently falls back to the catalog. Discovery needs `bedrock:ListFoundationModels` and `bedrock:ListInferenceProfiles`.
804			804

M packages/provider/discover_bedrock.go +7 -35


@@ -86,16 +86,6 @@ func DiscoverBedrock(ctx context.Context, region string) ([]Model, error) {
86	86		return out, nil
87	87		}
88	88
89		-	// bedrockAnthropicDefaultMaxOutput is the output budget assumed for a
90		-	// discovered Claude model with no catalog entry. It matches Claude Haiku
91		-	// 4.5, the smallest output limit among current Claude models, so it is
92		-	// safe for any Claude 4+ model and far above the client's 4096 fallback.
93		-	const bedrockAnthropicDefaultMaxOutput = 64000
94		-
95		-	// bedrockAnthropicDefaultContextWindow is the context window assumed
96		-	// for an uncatalogued Claude model, the smallest among current models.
97		-	const bedrockAnthropicDefaultContextWindow = 200000
98		-
99	89		// bedrockDiscoveredModel builds the Model for a discovered ID. The
100	90		// control plane returns IDs only, so limits and capabilities come from
101	91		// the static catalog. MergeCatalog already uses an exact catalog match;

@@ -133,7 +123,10 @@ func bedrockDiscoveredModel(id, region string) Model {
133	123		if c.DisplayName != "" {
134	124		m.DisplayName = c.DisplayName
135	125		if cp, _ := bedrockSplitGeoPrefix(c.ID); cp != "" {
136		-	m.DisplayName = strings.TrimSuffix(m.DisplayName, " ("+strings.ToUpper(cp)+")")
	126	+	suffix := " (" + cp + ")"
	127	+	if strings.HasSuffix(strings.ToUpper(m.DisplayName), strings.ToUpper(suffix)) {
	128	+	m.DisplayName = m.DisplayName[:len(m.DisplayName)-len(suffix)]
	129	+	}
137	130		}
138	131		if prefix != "" {
139	132		m.DisplayName += " (" + strings.ToUpper(prefix) + ")"

@@ -141,10 +134,9 @@ func bedrockDiscoveredModel(id, region string) Model {
141	134		}
142	135		return m
143	136		}
144		-	if bedrockIsModernClaude(base) {
145		-	m.ContextWindow = bedrockAnthropicDefaultContextWindow
146		-	m.MaxOutput = bedrockAnthropicDefaultMaxOutput
147		-	}
	137	+	// Unknown models have no verified output or context limits. Keep the
	138	+	// client's 4096-token fallback rather than guessing from the family:
	139	+	// even Claude Opus 4 supports only 32k, not the 64k of newer models.
148	140		return m
149	141		}
150	142

@@ -189,26 +181,6 @@ func bedrockCatalogBase(base string) (Model, bool) {
189	181		return variant, found
190	182		}
191	183
192		-	// bedrockIsModernClaude reports whether a foundation-model ID is a Claude
193		-	// model whose output limit is at least bedrockAnthropicDefaultMaxOutput.
194		-	// Legacy Claude 2 / Instant / 3 / 3.5 models cap output at 4096-8192 and
195		-	// would reject a 64000-token request, so they keep the client fallback.
196		-	func bedrockIsModernClaude(base string) bool {
197		-	if !strings.HasPrefix(base, "anthropic.claude-") {
198		-	return false
199		-	}
200		-	switch {
201		-	case strings.HasPrefix(base, "anthropic.claude-v"),
202		-	strings.HasPrefix(base, "anthropic.claude-instant"),
203		-	strings.HasPrefix(base, "anthropic.claude-3-5-"),
204		-	strings.HasPrefix(base, "anthropic.claude-3-haiku"),
205		-	strings.HasPrefix(base, "anthropic.claude-3-sonnet"),
206		-	strings.HasPrefix(base, "anthropic.claude-3-opus"):
207		-	return false
208		-	}
209		-	return true
210		-	}
211		-
212	184		// bedrockListModelIDs enumerates foundation-model IDs and inference-
213	185		// profile IDs available in the region. Profile IDs are the geo-prefixed
214	186		// forms zot invokes; both are returned so the picker shows every

 	return out, nil
 }
-// bedrockAnthropicDefaultMaxOutput is the output budget assumed for a
-// discovered Claude model with no catalog entry. It matches Claude Haiku
-// 4.5, the smallest output limit among current Claude models, so it is
-// safe for any Claude 4+ model and far above the client's 4096 fallback.
-const bedrockAnthropicDefaultMaxOutput = 64000
-// bedrockAnthropicDefaultContextWindow is the context window assumed
-// for an uncatalogued Claude model, the smallest among current models.
-const bedrockAnthropicDefaultContextWindow = 200000
 // bedrockDiscoveredModel builds the Model for a discovered ID. The
 // control plane returns IDs only, so limits and capabilities come from
 // the static catalog. MergeCatalog already uses an exact catalog match;
 		if c.DisplayName != "" {
 			m.DisplayName = c.DisplayName
 			if cp, _ := bedrockSplitGeoPrefix(c.ID); cp != "" {
-				m.DisplayName = strings.TrimSuffix(m.DisplayName, " ("+strings.ToUpper(cp)+")")
+				suffix := " (" + cp + ")"
+				if strings.HasSuffix(strings.ToUpper(m.DisplayName), strings.ToUpper(suffix)) {
+					m.DisplayName = m.DisplayName[:len(m.DisplayName)-len(suffix)]
+				}
 			}
 			if prefix != "" {
 				m.DisplayName += " (" + strings.ToUpper(prefix) + ")"
 		}
 		return m
 	}
-	if bedrockIsModernClaude(base) {
+	// Unknown models have no verified output or context limits. Keep the
-		m.ContextWindow = bedrockAnthropicDefaultContextWindow
+	// client's 4096-token fallback rather than guessing from the family:
-		m.MaxOutput = bedrockAnthropicDefaultMaxOutput
+	// even Claude Opus 4 supports only 32k, not the 64k of newer models.
-	}
 	return m
 }
 	return variant, found
 }
-// bedrockIsModernClaude reports whether a foundation-model ID is a Claude
-// model whose output limit is at least bedrockAnthropicDefaultMaxOutput.
-// Legacy Claude 2 / Instant / 3 / 3.5 models cap output at 4096-8192 and
-// would reject a 64000-token request, so they keep the client fallback.
-func bedrockIsModernClaude(base string) bool {
-	if !strings.HasPrefix(base, "anthropic.claude-") {
-		return false
-	}
-	switch {
-	case strings.HasPrefix(base, "anthropic.claude-v"),
-		strings.HasPrefix(base, "anthropic.claude-instant"),
-		strings.HasPrefix(base, "anthropic.claude-3-5-"),
-		strings.HasPrefix(base, "anthropic.claude-3-haiku"),
-		strings.HasPrefix(base, "anthropic.claude-3-sonnet"),
-		strings.HasPrefix(base, "anthropic.claude-3-opus"):
-		return false
-	}
-	return true
-}
 // bedrockListModelIDs enumerates foundation-model IDs and inference-
 // profile IDs available in the region. Profile IDs are the geo-prefixed
 // forms zot invokes; both are returned so the picker shows every

M packages/provider/discover_bedrock_metadata_test.go +14 -10


@@ -51,23 +51,27 @@ func TestBedrockDiscoveredProfileInheritsBaseMetadata(t *testing.T) {
51	51		// A base model catalogued only under geo-prefixed IDs still donates its
52	52		// metadata, and the donor's region label is not carried over.
53	53		func TestBedrockDiscoveredProfileUsesPrefixedCatalogVariant(t *testing.T) {
54		-	const id = "apac.anthropic.claude-opus-5-5"
	54	+	const id = "apac.anthropic.claude-opus-5"
55	55		m, wire := discoveredWireMaxTokens(t, id)
56		-	if wire != 128000 \|\| !m.AdaptiveThinking {
57		-	t.Errorf("%s: wire=%d adaptive=%v, want 128000 adaptive", id, wire, m.AdaptiveThinking)
	56	+	if wire != 128000 \|\| !m.AdaptiveThinking \|\| m.PriceOutput != 25 {
	57	+	t.Errorf("%s: wire=%d adaptive=%v price=%v, want 128000 adaptive and $25", id, wire, m.AdaptiveThinking, m.PriceOutput)
58	58		}
59		-	if m.DisplayName != "Claude Opus 5.5 (APAC)" {
	59	+	if m.DisplayName != "Claude Opus 5 (APAC)" {
60	60		t.Errorf("DisplayName = %q", m.DisplayName)
61	61		}
62	62		}
63	63
64		-	// Claude models absent from the catalog get Haiku 4.5's 64k output, the
65		-	// smallest among current Claude models, instead of the 4096 fallback.
66		-	func TestBedrockDiscoveredUnknownClaudeUses64kFloor(t *testing.T) {
67		-	for _, id := range []string{"anthropic.claude-future-9", "us.anthropic.claude-future-9-v1:0"} {
	64	+	// An uncatalogued Claude ID must not assume a 64k output limit. Opus 4
	65	+	// supports only 32k, and an unknown future model has no verified limit.
	66	+	func TestBedrockDiscoveredUnknownClaudeKeepsFallback(t *testing.T) {
	67	+	for _, id := range []string{
	68	+	"anthropic.claude-opus-4-20250514-v1:0",
	69	+	"us.anthropic.claude-opus-4-20250514-v1:0",
	70	+	"anthropic.claude-future-9",
	71	+	} {
68	72		m, wire := discoveredWireMaxTokens(t, id)
69		-	if wire != bedrockAnthropicDefaultMaxOutput \|\| m.ContextWindow != bedrockAnthropicDefaultContextWindow {
70		-	t.Errorf("%s: wire=%d ctx=%d", id, wire, m.ContextWindow)
	73	+	if m.MaxOutput != 0 \|\| m.ContextWindow != 0 \|\| wire != 4096 {
	74	+	t.Errorf("%s: MaxOutput=%d ctx=%d wire=%d, want 0/0/4096", id, m.MaxOutput, m.ContextWindow, wire)
71	75		}
72	76		}
73	77		}

 // A base model catalogued only under geo-prefixed IDs still donates its
 // metadata, and the donor's region label is not carried over.
 func TestBedrockDiscoveredProfileUsesPrefixedCatalogVariant(t *testing.T) {
-	const id = "apac.anthropic.claude-opus-5-5"
+	const id = "apac.anthropic.claude-opus-5"
 	m, wire := discoveredWireMaxTokens(t, id)
-	if wire != 128000 || !m.AdaptiveThinking {
+	if wire != 128000 || !m.AdaptiveThinking || m.PriceOutput != 25 {
-		t.Errorf("%s: wire=%d adaptive=%v, want 128000 adaptive", id, wire, m.AdaptiveThinking)
+		t.Errorf("%s: wire=%d adaptive=%v price=%v, want 128000 adaptive and $25", id, wire, m.AdaptiveThinking, m.PriceOutput)
 	}
-	if m.DisplayName != "Claude Opus 5.5 (APAC)" {
+	if m.DisplayName != "Claude Opus 5 (APAC)" {
 		t.Errorf("DisplayName = %q", m.DisplayName)
 	}
 }
-// Claude models absent from the catalog get Haiku 4.5's 64k output, the
+// An uncatalogued Claude ID must not assume a 64k output limit. Opus 4
-// smallest among current Claude models, instead of the 4096 fallback.
+// supports only 32k, and an unknown future model has no verified limit.
-func TestBedrockDiscoveredUnknownClaudeUses64kFloor(t *testing.T) {
+func TestBedrockDiscoveredUnknownClaudeKeepsFallback(t *testing.T) {
-	for _, id := range []string{"anthropic.claude-future-9", "us.anthropic.claude-future-9-v1:0"} {
+	for _, id := range []string{
+		"anthropic.claude-opus-4-20250514-v1:0",
+		"us.anthropic.claude-opus-4-20250514-v1:0",
+		"anthropic.claude-future-9",
+	} {
 		m, wire := discoveredWireMaxTokens(t, id)
-		if wire != bedrockAnthropicDefaultMaxOutput || m.ContextWindow != bedrockAnthropicDefaultContextWindow {
+		if m.MaxOutput != 0 || m.ContextWindow != 0 || wire != 4096 {
-			t.Errorf("%s: wire=%d ctx=%d", id, wire, m.ContextWindow)
+			t.Errorf("%s: MaxOutput=%d ctx=%d wire=%d, want 0/0/4096", id, m.MaxOutput, m.ContextWindow, wire)
 		}
 	}
 }