fix: correct Moonshot params to match Kimi's API docs

models/moonshot/kimi-k2.5.yaml

@@ -10,12 +10,20 @@ params:
     range:
       min: 1
     group: generation_length
+  - path: stop
+    type: string
+    label: Stop sequence
+    description: >-
+      Stops generation when this sequence is produced. Moonshot accepts up to 5 sequences of at most
+      32 bytes each.
+    group: generation_length
   - path: thinking.type
     type: enum
     label: Thinking mode
     description: >-
-      Controls whether Kimi reasons step by step before answering, or responds directly when set to
-      disabled.
+      Controls whether Kimi reasons step by step before answering. Thinking is enabled by default;
+      set disabled to respond directly.
+    default: enabled
     values:
       - enabled
       - disabled
@@ -29,3 +37,32 @@ params:
       - text
       - json_object
     group: output_format
+  - path: tool_choice
+    type: enum
+    label: Tool choice
+    description: >-
+      Controls tool calling: auto lets the model decide, none blocks tool calls, and required forces
+      one.
+    default: auto
+    values:
+      - auto
+      - none
+      - required
+    group: tooling
+  - path: logprobs
+    type: boolean
+    label: Log probabilities
+    description: Controls whether the response includes log probabilities for the generated tokens.
+    default: false
+    group: observability
+  - path: top_logprobs
+    type: integer
+    label: Top log probabilities
+    description: Number of most likely tokens to return a log probability for at each position.
+    range:
+      min: 0
+      max: 20
+    group: observability
+    applicability:
+      only:
+        logprobs: true

models/moonshot/kimi-k2.6-subscription.yaml

@@ -21,6 +21,17 @@ params:
       - enabled
       - disabled
     group: reasoning
+  - path: thinking.keep
+    type: enum
+    label: Keep prior reasoning
+    description: >-
+      Set to all to carry reasoning content from earlier assistant turns into the request; leave
+      null to send only the final answers.
+    default: null
+    values:
+      - all
+      - null
+    group: reasoning
   - path: response_format.type
     type: enum
     label: Response format

models/moonshot/kimi-k2.6.yaml

@@ -10,6 +10,13 @@ params:
     range:
       min: 1
     group: generation_length
+  - path: stop
+    type: string
+    label: Stop sequence
+    description: >-
+      Stops generation when this sequence is produced. Moonshot accepts up to 5 sequences of at most
+      32 bytes each.
+    group: generation_length
   - path: thinking.type
     type: enum
     label: Thinking mode
@@ -21,6 +28,17 @@ params:
       - enabled
       - disabled
     group: reasoning
+  - path: thinking.keep
+    type: enum
+    label: Keep prior reasoning
+    description: >-
+      Set to all to carry reasoning content from earlier assistant turns into the request; leave
+      null to send only the final answers.
+    default: null
+    values:
+      - all
+      - null
+    group: reasoning
   - path: response_format.type
     type: enum
     label: Response format
@@ -30,3 +48,32 @@ params:
       - text
       - json_object
     group: output_format
+  - path: tool_choice
+    type: enum
+    label: Tool choice
+    description: >-
+      Controls tool calling: auto lets the model decide, none blocks tool calls, and required forces
+      one.
+    default: auto
+    values:
+      - auto
+      - none
+      - required
+    group: tooling
+  - path: logprobs
+    type: boolean
+    label: Log probabilities
+    description: Controls whether the response includes log probabilities for the generated tokens.
+    default: false
+    group: observability
+  - path: top_logprobs
+    type: integer
+    label: Top log probabilities
+    description: Number of most likely tokens to return a log probability for at each position.
+    range:
+      min: 0
+      max: 20
+    group: observability
+    applicability:
+      only:
+        logprobs: true

models/moonshot/kimi-k2.7-code-highspeed-subscription.yaml

@@ -10,6 +10,16 @@ params:
     range:
       min: 1
     group: generation_length
+  - path: thinking.type
+    type: enum
+    label: Thinking mode
+    description: >-
+      Thinking is always on for this model. Enabled is the only accepted value; disabled returns an
+      error.
+    default: enabled
+    values:
+      - enabled
+    group: reasoning
   - path: response_format.type
     type: enum
     label: Response format

models/moonshot/kimi-k2.7-code-highspeed.yaml

@@ -11,6 +11,23 @@ params:
     range:
       min: 1
     group: generation_length
+  - path: stop
+    type: string
+    label: Stop sequence
+    description: >-
+      Stops generation when this sequence is produced. Moonshot accepts up to 5 sequences of at most
+      32 bytes each.
+    group: generation_length
+  - path: thinking.type
+    type: enum
+    label: Thinking mode
+    description: >-
+      Thinking is always on for this model. Enabled is the only accepted value; disabled returns an
+      error.
+    default: enabled
+    values:
+      - enabled
+    group: reasoning
   - path: response_format.type
     type: enum
     label: Response format
@@ -21,3 +38,32 @@ params:
       - json_object
       - json_schema
     group: output_format
+  - path: tool_choice
+    type: enum
+    label: Tool choice
+    description: >-
+      Controls tool calling: auto lets the model decide, none blocks tool calls, and required forces
+      one.
+    default: auto
+    values:
+      - auto
+      - none
+      - required
+    group: tooling
+  - path: logprobs
+    type: boolean
+    label: Log probabilities
+    description: Controls whether the response includes log probabilities for the generated tokens.
+    default: false
+    group: observability
+  - path: top_logprobs
+    type: integer
+    label: Top log probabilities
+    description: Number of most likely tokens to return a log probability for at each position.
+    range:
+      min: 0
+      max: 20
+    group: observability
+    applicability:
+      only:
+        logprobs: true

models/moonshot/kimi-k2.7-code-subscription.yaml

@@ -10,6 +10,16 @@ params:
     range:
       min: 1
     group: generation_length
+  - path: thinking.type
+    type: enum
+    label: Thinking mode
+    description: >-
+      Thinking is always on for this model. Enabled is the only accepted value; disabled returns an
+      error.
+    default: enabled
+    values:
+      - enabled
+    group: reasoning
   - path: response_format.type
     type: enum
     label: Response format

models/moonshot/kimi-k2.7-code.yaml

@@ -11,6 +11,23 @@ params:
     range:
       min: 1
     group: generation_length
+  - path: stop
+    type: string
+    label: Stop sequence
+    description: >-
+      Stops generation when this sequence is produced. Moonshot accepts up to 5 sequences of at most
+      32 bytes each.
+    group: generation_length
+  - path: thinking.type
+    type: enum
+    label: Thinking mode
+    description: >-
+      Thinking is always on for this model. Enabled is the only accepted value; disabled returns an
+      error.
+    default: enabled
+    values:
+      - enabled
+    group: reasoning
   - path: response_format.type
     type: enum
     label: Response format
@@ -21,3 +38,32 @@ params:
       - json_object
       - json_schema
     group: output_format
+  - path: tool_choice
+    type: enum
+    label: Tool choice
+    description: >-
+      Controls tool calling: auto lets the model decide, none blocks tool calls, and required forces
+      one.
+    default: auto
+    values:
+      - auto
+      - none
+      - required
+    group: tooling
+  - path: logprobs
+    type: boolean
+    label: Log probabilities
+    description: Controls whether the response includes log probabilities for the generated tokens.
+    default: false
+    group: observability
+  - path: top_logprobs
+    type: integer
+    label: Top log probabilities
+    description: Number of most likely tokens to return a log probability for at each position.
+    range:
+      min: 0
+      max: 20
+    group: observability
+    applicability:
+      only:
+        logprobs: true

models/moonshot/kimi-k3.yaml

@@ -10,15 +10,24 @@ params:
     range:
       min: 1
     group: generation_length
-  - path: thinking.type
+  - path: stop
+    type: string
+    label: Stop sequence
+    description: >-
+      Stops generation when this sequence is produced. Moonshot accepts up to 5 sequences of at most
+      32 bytes each.
+    group: generation_length
+  - path: reasoning_effort
     type: enum
-    label: Thinking mode
+    label: Reasoning effort
     description: >-
-      Controls whether Kimi reasons step by step before answering, or responds directly when set to
-      disabled.
+      Controls how much reasoning Kimi performs before answering. Thinking is always on for this
+      model, so it cannot be turned off.
+    default: max
     values:
-      - enabled
-      - disabled
+      - low
+      - high
+      - max
     group: reasoning
   - path: response_format.type
     type: enum
@@ -30,3 +39,32 @@ params:
       - json_object
       - json_schema
     group: output_format
+  - path: tool_choice
+    type: enum
+    label: Tool choice
+    description: >-
+      Controls tool calling: auto lets the model decide, none blocks tool calls, and required forces
+      one.
+    default: auto
+    values:
+      - auto
+      - none
+      - required
+    group: tooling
+  - path: logprobs
+    type: boolean
+    label: Log probabilities
+    description: Controls whether the response includes log probabilities for the generated tokens.
+    default: false
+    group: observability
+  - path: top_logprobs
+    type: integer
+    label: Top log probabilities
+    description: Number of most likely tokens to return a log probability for at each position.
+    range:
+      min: 0
+      max: 20
+    group: observability
+    applicability:
+      only:
+        logprobs: true

models/moonshot/moonshot-v1-128k.yaml

@@ -10,6 +10,13 @@ params:
     range:
       min: 1
     group: generation_length
+  - path: stop
+    type: string
+    label: Stop sequence
+    description: >-
+      Stops generation when this sequence is produced. Moonshot accepts up to 5 sequences of at most
+      32 bytes each.
+    group: generation_length
   - path: temperature
     type: number
     label: Temperature
@@ -73,3 +80,32 @@ params:
       - text
       - json_object
     group: output_format
+  - path: tool_choice
+    type: enum
+    label: Tool choice
+    description: >-
+      Controls tool calling: auto lets the model decide, none blocks tool calls, and required forces
+      one.
+    default: auto
+    values:
+      - auto
+      - none
+      - required
+    group: tooling
+  - path: logprobs
+    type: boolean
+    label: Log probabilities
+    description: Controls whether the response includes log probabilities for the generated tokens.
+    default: false
+    group: observability
+  - path: top_logprobs
+    type: integer
+    label: Top log probabilities
+    description: Number of most likely tokens to return a log probability for at each position.
+    range:
+      min: 0
+      max: 20
+    group: observability
+    applicability:
+      only:
+        logprobs: true