Neon: declare structured_output where it does not resolve (#4361)

providers/neon/models/claude-fable-5.toml

@@ -1,6 +1,7 @@
 # No toggle: "thinking.type.disabled" is rejected for this model, which defaults to
 # adaptive thinking, so effort is the only control the gateway exposes.
 base_model = "anthropic/claude-fable-5"
+structured_output = true
 reasoning_options = [
   { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }, # API: {"thinking": {"type": "adaptive"}, "output_config": {"effort": <v>}}
 ]

providers/neon/models/claude-haiku-4-5.toml

@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-haiku-4-5"
+structured_output = true
 reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024, max = 63_999 }]
 
 [cost]

providers/neon/models/claude-opus-4-1.toml

@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-opus-4-1"
+structured_output = true
 reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024, max = 31_999 }]
 
 [cost]

providers/neon/models/claude-opus-4-5.toml

@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-opus-4-5"
+structured_output = true
 reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024, max = 63_999 }]
 
 [cost]

providers/neon/models/claude-opus-4-6.toml

@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-opus-4-6"
+structured_output = true
 reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024, max = 127_999 }]
 
 [cost]

providers/neon/models/claude-opus-4-7.toml

@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-opus-4-7"
+structured_output = true
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
 [cost]

providers/neon/models/claude-opus-4-8.toml

@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-opus-4-8"
+structured_output = true
 reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
 [cost]

providers/neon/models/claude-opus-5.toml

@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-opus-5"
+structured_output = true
 reasoning_options = [
   { type = "toggle" },                                                   # API: {"thinking": {"type": "adaptive"|"disabled"}}
   { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }, # API: {"thinking": {"type": "adaptive"}, "output_config": {"effort": <v>}}

providers/neon/models/claude-sonnet-4-5.toml

@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-sonnet-4-5"
+structured_output = true
 reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024, max = 63_999 }]
 
 [cost]

providers/neon/models/claude-sonnet-4-6.toml

@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-sonnet-4-6"
+structured_output = true
 reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1_024, max = 63_999 }]
 
 [cost]

providers/neon/models/claude-sonnet-5.toml

@@ -1,4 +1,5 @@
 base_model = "anthropic/claude-sonnet-5"
+structured_output = true
 reasoning_options = [
   { type = "toggle" },                                                   # API: {"thinking": {"type": "adaptive"|"disabled"}}
   { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }, # API: {"thinking": {"type": "adaptive"}, "output_config": {"effort": <v>}}

providers/neon/models/inkling.toml

@@ -1,4 +1,8 @@
 base_model = "thinkingmachines/inkling"
+# The gateway does not honour response_format json_schema here: it answers
+# 200 with finish_reason "abort" and empty content, with strict true or
+# false, at any token budget. Plain requests work. Measured 2026-08-08.
+structured_output = false
 reasoning_options = [
   { type = "toggle" },                                                              # API: {"thinking": {"type": "enabled"|"disabled"}}
   { type = "effort", values = ["none", "minimal", "low", "medium", "high", "max"] }, # API: {"reasoning_effort": <v>}

providers/neon/models/llama-4-maverick.toml

@@ -1,4 +1,5 @@
 base_model = "meta/llama-4-maverick-17b-instruct"
+structured_output = true
 
 [cost]
 input = 0.5

providers/neon/models/meta-llama-3-3-70b-instruct.toml

@@ -1,4 +1,5 @@
 base_model = "meta/llama-3.3-70b-instruct"
+structured_output = true
 attachment = false
 
 [cost]

providers/neon/models/qwen3-next-80b-a3b-instruct.toml

@@ -1,4 +1,5 @@
 base_model = "alibaba/qwen3-next-80b-a3b-instruct"
+structured_output = true
 
 [cost]
 input = 0.15