feat: generate the model catalog from live API data

Add a hidden --generate-models flag that rebuilds models.yaml from
models.dev and OpenRouter merged with the hand-owned quirks overlay,
regenerate the committed catalog with it (393 -> 1090 models), and
retire the Python updater. GitHub becomes fully quirks-owned since
upstream only tracks Copilot ids, legacy ids upstream dropped are
preserved as hand-owned quirks entries, and non-chat model families
(image/tts/audio/realtime) are excluded from generation.

quirks.yaml

@@ -9,6 +9,11 @@
 # entries are grouped under a `# legacy — pruning candidates` comment.
 
 - provider: openai
+  exclude:
+    - "text-embedding-*"
+    - "gpt-image-*"
+    - "chatgpt-image-*"
+    - "gpt-realtime*"
   rules:
     - match: "gpt-5.6-*"
       default_reasoning_effort: medium
@@ -118,8 +123,39 @@
           max_tokens: null
           temperature: null
           top_p: null
+    - name: chat-latest
+      max_input_tokens: 400000
+      max_output_tokens: 128000
+      input_price: 5
+      output_price: 30
+      supports_vision: true
+      supports_function_calling: true
+    - name: gpt-5.1-chat-latest
+      max_input_tokens: 400000
+      max_output_tokens: 128000
+      input_price: 1.25
+      output_price: 10
+      supports_vision: true
+      supports_function_calling: true
+      reasoning_levels: [none, low, medium, high]
+      default_reasoning_effort: none
+    - name: gpt-5-chat-latest
+      max_input_tokens: 400000
+      max_output_tokens: 128000
+      input_price: 1.25
+      output_price: 10
+      supports_vision: true
+      supports_function_calling: true
+      reasoning_levels: [minimal, low, medium, high]
+      default_reasoning_effort: medium
 
 - provider: gemini
+  exclude:
+    - "gemini-embedding-*"
+    - "veo-*"
+    - "*-tts"
+    - "*-tts-preview"
+    - "*-image-preview"
   rules:
     - match: "gemini-3.6-flash"
       default_reasoning_effort: medium
@@ -135,6 +171,7 @@
       default_reasoning_effort: high
     - match: "gemini-2.5-pro"
       default_reasoning_effort: high
+      reasoning_levels: [low, medium, high]
   models:
     - name: text-embedding-004
       type: embedding
@@ -142,6 +179,18 @@
       max_tokens_per_chunk: 2048
       default_chunk_size: 1500
       max_batch_size: 100
+    # legacy — pruning candidates
+    - name: gemini-3-pro-preview
+      max_input_tokens: 1048576
+      supports_vision: true
+      supports_function_calling: true
+      reasoning_levels: [low, high]
+      default_reasoning_effort: high
+    - name: gemma-3-27b-it
+      max_input_tokens: 131072
+      max_output_tokens: 8192
+      input_price: 0
+      output_price: 0
 
 - provider: claude
   rules:
@@ -246,8 +295,28 @@
           thinking:
             type: enabled
             budget_tokens: 16000
+  models:
+    # legacy — pruning candidates
+    - name: claude-opus-4-1-20250805
+      max_input_tokens: 200000
+      max_output_tokens: 8192
+      require_max_tokens: true
+      input_price: 15
+      output_price: 75
+      supports_vision: true
+      supports_function_calling: true
+    - name: claude-opus-4-20250514
+      max_input_tokens: 200000
+      max_output_tokens: 8192
+      require_max_tokens: true
+      input_price: 15
+      output_price: 75
+      supports_vision: true
+      supports_function_calling: true
 
 - provider: mistral
+  exclude:
+    - "voxtral-*"
   models:
     - name: mistral-embed
       type: embedding
@@ -255,6 +324,21 @@
       input_price: 0.1
       max_tokens_per_chunk: 8092
       default_chunk_size: 2000
+    # legacy — pruning candidates
+    - name: magistral-small-latest
+      max_input_tokens: 131072
+      input_price: 0.5
+      output_price: 1.5
+    - name: devstral-small-latest
+      max_input_tokens: 262144
+      input_price: 0.1
+      output_price: 0.3
+      supports_function_calling: true
+    - name: ministral-14b-latest
+      max_input_tokens: 262144
+      input_price: 0.2
+      output_price: 0.2
+      supports_function_calling: true
 
 - provider: ai21
   models:
@@ -300,6 +384,8 @@
       max_input_tokens: 4096
 
 - provider: xai
+  exclude:
+    - "grok-imagine-*"
   oauth:
     client_id: b1a00492-073a-47ea-816f-4c329264a828
     authorize_url: https://auth.x.ai/oauth2/authorize
@@ -337,8 +423,71 @@
       output_price: 2.5
       max_input_tokens: 1000000
       supports_function_calling: true
+    - name: grok-4-1-fast-non-reasoning
+      max_input_tokens: 1000000
+      input_price: 0.2
+      output_price: 0.5
+      supports_function_calling: true
+    - name: grok-4-1-fast-reasoning
+      max_input_tokens: 1000000
+      input_price: 0.2
+      output_price: 0.5
+      supports_function_calling: true
+    - name: grok-code-fast-1
+      max_input_tokens: 256000
+      input_price: 0.2
+      output_price: 1.5
+      supports_function_calling: true
+    - name: grok-3
+      max_input_tokens: 131072
+      input_price: 3
+      output_price: 15
+      supports_function_calling: true
+    - name: grok-3-mini
+      max_input_tokens: 131072
+      input_price: 0.3
+      output_price: 0.5
+      supports_function_calling: true
+
+- provider: perplexity
+  models:
+    # legacy — pruning candidates
+    - name: sonar-reasoning
+      max_input_tokens: 128000
+      input_price: 1
+      output_price: 5
+
+- provider: groq
+  exclude:
+    - "whisper-*"
+  models:
+    # legacy — pruning candidates
+    - name: meta-llama/llama-4-maverick-17b-128e-instruct
+      max_input_tokens: 131072
+      input_price: 0
+      output_price: 0
+      supports_vision: true
+      supports_function_calling: true
+    - name: meta-llama/llama-4-scout-17b-16e-instruct
+      max_input_tokens: 131072
+      input_price: 0
+      output_price: 0
+      supports_vision: true
+      supports_function_calling: true
+    - name: moonshotai/kimi-k2-instruct-0905
+      max_input_tokens: 262144
+      input_price: 0
+      output_price: 0
+      supports_function_calling: true
+    - name: qwen/qwen3-32b
+      max_input_tokens: 131072
+      input_price: 0
+      output_price: 0
 
 - provider: vertexai
+  exclude:
+    - "gemini-embedding-*"
+    - "*-tts"
   rules:
     - match: "claude-*"
       require_max_tokens: true
@@ -354,8 +503,10 @@
       default_reasoning_effort: high
     - match: "gemini-2.5-flash"
       default_reasoning_effort: medium
+      reasoning_levels: [low, medium, high]
     - match: "gemini-2.5-pro"
       default_reasoning_effort: high
+      reasoning_levels: [low, medium, high]
     - match: "claude-opus-5"
       default_reasoning_effort: high
     - match: "claude-fable-*"
@@ -449,6 +600,109 @@
       max_tokens_per_chunk: 2048
       default_chunk_size: 1500
       max_batch_size: 5
+    # legacy — pruning candidates
+    - name: gemini-3-pro-preview
+      max_input_tokens: 1048576
+      supports_vision: true
+      supports_function_calling: true
+      reasoning_levels: [low, high]
+      default_reasoning_effort: high
+    - name: claude-opus-5
+      max_input_tokens: 1000000
+      max_output_tokens: 128000
+      require_max_tokens: true
+      input_price: 5
+      output_price: 25
+      supports_function_calling: true
+      supports_vision: true
+      reasoning_levels: [low, medium, high, xhigh, max]
+      default_reasoning_effort: high
+    - name: claude-fable-5-1
+      max_input_tokens: 1000000
+      max_output_tokens: 128000
+      require_max_tokens: true
+      input_price: 10
+      output_price: 50
+      supports_function_calling: true
+      supports_vision: true
+      reasoning_levels: [low, medium, high, xhigh, max]
+      default_reasoning_effort: high
+    - name: claude-fable-5
+      max_input_tokens: 1000000
+      max_output_tokens: 128000
+      require_max_tokens: true
+      input_price: 10
+      output_price: 50
+      supports_function_calling: true
+      supports_vision: true
+      reasoning_levels: [low, medium, high, xhigh, max]
+      default_reasoning_effort: high
+    - name: claude-opus-4-8
+      max_input_tokens: 1000000
+      max_output_tokens: 128000
+      require_max_tokens: true
+      input_price: 5
+      output_price: 25
+      supports_vision: true
+      supports_function_calling: true
+      reasoning_levels: [low, medium, high, xhigh, max]
+      default_reasoning_effort: high
+    - name: claude-opus-4-7
+      max_input_tokens: 1000000
+      max_output_tokens: 128000
+      require_max_tokens: true
+      input_price: 5
+      output_price: 25
+      supports_vision: true
+      supports_function_calling: true
+      reasoning_levels: [low, medium, high, xhigh, max]
+      default_reasoning_effort: high
+    - name: claude-opus-4-6
+      max_input_tokens: 200000
+      max_output_tokens: 8192
+      require_max_tokens: true
+      input_price: 5
+      output_price: 25
+      supports_vision: true
+      supports_function_calling: true
+    - name: claude-sonnet-5
+      max_input_tokens: 1000000
+      max_output_tokens: 128000
+      require_max_tokens: true
+      input_price: 3
+      output_price: 15
+      supports_vision: true
+      supports_function_calling: true
+      reasoning_levels: [low, medium, high, xhigh, max]
+      default_reasoning_effort: high
+    - name: claude-sonnet-4-6
+      max_input_tokens: 200000
+      max_output_tokens: 8192
+      require_max_tokens: true
+      input_price: 3
+      output_price: 15
+      supports_vision: true
+      supports_function_calling: true
+      reasoning_levels: [low, medium, high, max]
+      default_reasoning_effort: high
+    - name: claude-3-5-haiku@20241022
+      max_input_tokens: 200000
+      max_output_tokens: 8192
+      require_max_tokens: true
+      input_price: 0.8
+      output_price: 4
+      supports_vision: true
+      supports_function_calling: true
+    - name: mistral-small-2503
+      max_input_tokens: 32000
+      input_price: 0.1
+      output_price: 0.3
+      supports_function_calling: true
+    - name: codestral-2501
+      max_input_tokens: 256000
+      input_price: 0.3
+      output_price: 0.9
+      supports_function_calling: true
 
 - provider: bedrock
   rules:
@@ -559,6 +813,36 @@
       max_tokens_per_chunk: 512
       default_chunk_size: 1000
       max_batch_size: 96
+    # legacy — pruning candidates
+    - name: us.meta.llama3-3-70b-instruct-v1:0
+      max_input_tokens: 131072
+      max_output_tokens: 8192
+      require_max_tokens: true
+      input_price: 0.72
+      output_price: 0.72
+      supports_function_calling: true
+    - name: us.amazon.nova-premier-v1:0
+      max_input_tokens: 300000
+      max_output_tokens: 5120
+      input_price: 2.5
+      output_price: 12.5
+    - name: us.amazon.nova-pro-v1:0
+      max_input_tokens: 300000
+      max_output_tokens: 5120
+      input_price: 0.8
+      output_price: 3.2
+      supports_vision: true
+    - name: us.amazon.nova-lite-v1:0
+      max_input_tokens: 300000
+      max_output_tokens: 5120
+      input_price: 0.06
+      output_price: 0.24
+      supports_vision: true
+    - name: us.amazon.nova-micro-v1:0
+      max_input_tokens: 128000
+      max_output_tokens: 5120
+      input_price: 0.035
+      output_price: 0.14
 
 - provider: cloudflare
   rules:
@@ -571,6 +855,13 @@
       max_tokens_per_chunk: 512
       default_chunk_size: 1000
       max_batch_size: 100
+    # legacy — pruning candidates
+    - name: '@cf/google/gemma-3-12b-it'
+      max_input_tokens: 131072
+      max_output_tokens: 2048
+      require_max_tokens: true
+      input_price: 0
+      output_price: 0
 
 - provider: ernie
   models:
@@ -609,6 +900,8 @@
       input_price: 0.07
 
 - provider: qianwen
+  exclude:
+    - "*-realtime"
   rules:
     - match: "qwen3.5-plus"
       patch:
@@ -646,6 +939,40 @@
       max_tokens_per_chunk: 8192
       default_chunk_size: 2000
       max_batch_size: 10
+    # legacy — pruning candidates
+    - name: qwen3-vl-flash
+      max_input_tokens: 262144
+      supports_vision: true
+    - name: qwen3-235b-a22b-instruct-2507
+      max_input_tokens: 131072
+      input_price: 0.28
+      output_price: 1.12
+      supports_function_calling: true
+    - name: qwen3-235b-a22b-thinking-2507
+      max_input_tokens: 131072
+      input_price: 0.28
+      output_price: 2.8
+    - name: qwen3-30b-a3b-instruct-2507
+      max_input_tokens: 131072
+      input_price: 0.105
+      output_price: 0.42
+      supports_function_calling: true
+    - name: qwen3-30b-a3b-thinking-2507
+      max_input_tokens: 131072
+      input_price: 0.105
+      output_price: 1.05
+    - name: qwen3-vl-32b-instruct
+      max_input_tokens: 131072
+      input_price: 0.28
+      output_price: 1.12
+      supports_vision: true
+    - name: qwen3-vl-8b-instruct
+      max_input_tokens: 131072
+      input_price: 0.07
+      output_price: 0.28
+      supports_vision: true
+    - name: qwen3-coder-next
+      max_input_tokens: 262144
 
 - provider: hunyuan
   models:
@@ -671,6 +998,21 @@
       default_chunk_size: 1000
       max_batch_size: 100
 
+- provider: deepseek
+  models:
+    # legacy — pruning candidates
+    - name: deepseek-chat
+      max_input_tokens: 64000
+      max_output_tokens: 8192
+      input_price: 0.56
+      output_price: 1.68
+      supports_function_calling: true
+    - name: deepseek-reasoner
+      max_input_tokens: 64000
+      max_output_tokens: 32768
+      input_price: 0.56
+      output_price: 1.68
+
 - provider: zhipuai
   variants:
     - match: "glm-5"
@@ -698,6 +1040,46 @@
       type: reranker
       max_input_tokens: 4096
       input_price: 0.112
+    # legacy — pruning candidates
+    - name: glm-4.6v-flash
+      max_input_tokens: 65536
+      input_price: 0
+      output_price: 0
+      supports_vision: true
+
+- provider: minimax
+  models:
+    # legacy — pruning candidates
+    - name: minimax-m3
+      max_input_tokens: 1000000
+      input_price: 4.2
+      output_price: 16.8
+      supports_function_calling: true
+    - name: minimax-m2.7
+      max_input_tokens: 204800
+      input_price: 0.294
+      output_price: 1.176
+      supports_function_calling: true
+    - name: minimax-m2.5
+      max_input_tokens: 204800
+      input_price: 0.294
+      output_price: 1.176
+      supports_function_calling: true
+    - name: minimax-m2.5-highspeed
+      max_input_tokens: 204800
+      input_price: 0.588
+      output_price: 2.352
+      supports_function_calling: true
+    - name: minimax-m2.1
+      max_input_tokens: 204800
+      input_price: 0.294
+      output_price: 1.176
+      supports_function_calling: true
+    - name: minimax-m2.1-highspeed
+      max_input_tokens: 204800
+      input_price: 0.588
+      output_price: 2.352
+      supports_function_calling: true
 
 - provider: openrouter
   rules:
@@ -705,54 +1087,76 @@
       require_max_tokens: true
     - match: "openai/gpt-5.6-*"
       default_reasoning_effort: medium
+      reasoning_levels: [none, low, medium, high, xhigh, max]
     - match: "openai/gpt-5.5"
       default_reasoning_effort: medium
+      reasoning_levels: [none, low, medium, high, xhigh]
     - match: "openai/gpt-5.5-pro"
       default_reasoning_effort: high
+      reasoning_levels: [medium, high, xhigh]
     - match: "openai/gpt-5.4"
       default_reasoning_effort: none
+      reasoning_levels: [none, low, medium, high, xhigh]
     - match: "openai/gpt-5.4-pro"
       default_reasoning_effort: medium
+      reasoning_levels: [medium, high, xhigh]
     - match: "openai/gpt-5.4-mini"
       default_reasoning_effort: none
+      reasoning_levels: [none, low, medium, high, xhigh]
     - match: "openai/gpt-5.4-nano"
       default_reasoning_effort: none
+      reasoning_levels: [none, low, medium, high, xhigh]
     - match: "openai/gpt-5.3-codex"
       default_reasoning_effort: medium
+      reasoning_levels: [low, medium, high, xhigh]
     - match: "openai/gpt-5.2"
       default_reasoning_effort: none
+      reasoning_levels: [none, low, medium, high, xhigh]
     - match: "openai/gpt-5.2-pro"
       default_reasoning_effort: medium
+      reasoning_levels: [medium, high, xhigh]
     - match: "openai/gpt-5"
       default_reasoning_effort: medium
+      reasoning_levels: [minimal, low, medium, high]
     - match: "google/gemini-3.5-flash"
       default_reasoning_effort: medium
+      reasoning_levels: [minimal, low, medium, high]
     - match: "google/gemini-3-flash-preview"
       default_reasoning_effort: high
+      reasoning_levels: [minimal, low, medium, high]
     - match: "google/gemini-3.1-flash-lite"
       default_reasoning_effort: minimal
+      reasoning_levels: [minimal, low, medium, high]
     - match: "google/gemini-3.1-pro-preview"
       default_reasoning_effort: high
+      reasoning_levels: [low, medium, high]
     - match: "google/gemini-3-pro-preview"
       default_reasoning_effort: high
     - match: "google/gemini-2.5-flash"
       default_reasoning_effort: low
+      reasoning_levels: [low, medium, high]
     - match: "google/gemini-2.5-pro"
       default_reasoning_effort: high
+      reasoning_levels: [low, medium, high]
     - match: "anthropic/claude-opus-5"
       default_reasoning_effort: high
+      reasoning_levels: [low, medium, high, xhigh, max]
     - match: "anthropic/claude-fable-*"
       default_reasoning_effort: high
+      reasoning_levels: [low, medium, high, xhigh, max]
     - match: "anthropic/claude-opus-4-8"
       default_reasoning_effort: high
     - match: "anthropic/claude-opus-4-7"
       default_reasoning_effort: high
     - match: "anthropic/claude-opus-4.6"
       default_reasoning_effort: high
+      reasoning_levels: [low, medium, high, max]
     - match: "anthropic/claude-sonnet-5"
       default_reasoning_effort: high
+      reasoning_levels: [low, medium, high, xhigh, max]
     - match: "anthropic/claude-sonnet-4.6"
       default_reasoning_effort: high
+      reasoning_levels: [low, medium, high, max]
     - match: "perplexity/sonar-reasoning-pro"
       patch:
         body:
@@ -761,9 +1165,105 @@
       patch:
         body:
           include_reasoning: true
+  models:
+    # legacy — pruning candidates
+    - name: google/gemini-3-pro-preview
+      max_input_tokens: 1048576
+      supports_vision: true
+      supports_function_calling: true
+      reasoning_levels: [low, high]
+      default_reasoning_effort: high
+    - name: google/gemini-2.0-flash-001
+      max_input_tokens: 1000000
+      input_price: 0.15
+      output_price: 0.6
+      supports_vision: true
+      supports_function_calling: true
+    - name: google/gemini-2.0-flash-lite-001
+      max_input_tokens: 1048576
+      input_price: 0.075
+      output_price: 0.3
+      supports_vision: true
+      supports_function_calling: true
+    - name: anthropic/claude-fable-5-1
+      max_input_tokens: 1000000
+      max_output_tokens: 128000
+      require_max_tokens: true
+      input_price: 10
+      output_price: 50
+      supports_function_calling: true
+      supports_vision: true
+      reasoning_levels: [low, medium, high, xhigh, max]
+      default_reasoning_effort: high
+    - name: anthropic/claude-opus-4-8
+      max_input_tokens: 1000000
+      max_output_tokens: 128000
+      require_max_tokens: true
+      input_price: 5
+      output_price: 25
+      supports_vision: true
+      supports_function_calling: true
+      reasoning_levels: [low, medium, high, xhigh, max]
+      default_reasoning_effort: high
+    - name: anthropic/claude-opus-4-7
+      max_input_tokens: 1000000
+      max_output_tokens: 128000
+      require_max_tokens: true
+      input_price: 5
+      output_price: 25
+      supports_vision: true
+      supports_function_calling: true
+      reasoning_levels: [low, medium, high, xhigh, max]
+      default_reasoning_effort: high
+    - name: mistralai/devstral-small
+      max_input_tokens: 131072
+      input_price: 0.07
+      output_price: 0.28
+      supports_function_calling: true
+    - name: ai21/jamba-large-1.7
+      max_input_tokens: 256000
+      input_price: 2
+      output_price: 8
+      supports_function_calling: true
+    - name: x-ai/grok-4.1-fast
+      max_input_tokens: 2000000
+      input_price: 0.2
+      output_price: 0.5
+      supports_function_calling: true
+    - name: x-ai/grok-code-fast-1
+      max_input_tokens: 256000
+      input_price: 0.2
+      output_price: 1.5
+      supports_function_calling: true
 
+# Fully hand-owned: upstream catalogs track GitHub Copilot ids, which do not
+# match the GitHub Models API ids coyote uses.
 - provider: github
   models:
+    - name: gpt-5
+      max_input_tokens: 400000
+      max_output_tokens: 128000
+      supports_vision: true
+      supports_function_calling: true
+    - name: gpt-5-mini
+      max_input_tokens: 400000
+      max_output_tokens: 128000
+      supports_vision: true
+      supports_function_calling: true
+    - name: gpt-5-nano
+      max_input_tokens: 400000
+      max_output_tokens: 128000
+      supports_vision: true
+      supports_function_calling: true
+    - name: gpt-4.1
+      max_input_tokens: 1047576
+      max_output_tokens: 32768
+      supports_vision: true
+      supports_function_calling: true
+    - name: gpt-4o
+      max_input_tokens: 128000
+      max_output_tokens: 16384
+      supports_function_calling: true
     - name: text-embedding-3-large
       type: embedding
       max_tokens_per_chunk: 8191
@@ -774,6 +1274,23 @@
       max_tokens_per_chunk: 8191
       default_chunk_size: 2000
       max_batch_size: 100
+    - name: llama-4-maverick-17b-128e-instruct-fp8
+      max_input_tokens: 1048576
+      supports_vision: true
+    - name: llama-4-scout-17b-16e-instruct
+      max_input_tokens: 327680
+      supports_vision: true
+    - name: llama-3.3-70b-instruct
+      max_input_tokens: 131072
+    - name: mistral-medium-2505
+      max_input_tokens: 131072
+      supports_function_calling: true
+    - name: mistral-small-2503
+      max_input_tokens: 131072
+      supports_function_calling: true
+    - name: codestral-2501
+      max_input_tokens: 256000
+      supports_function_calling: true
     - name: cohere-embed-v3-english
       type: embedding
       max_tokens_per_chunk: 512
@@ -784,6 +1301,24 @@
       max_tokens_per_chunk: 512
       default_chunk_size: 1000
       max_batch_size: 96
+    - name: deepseek-r1-0528
+      max_input_tokens: 163840
+    - name: deepseek-v3-0324
+      max_input_tokens: 163840
+    - name: mai-ds-r1
+      max_input_tokens: 163840
+    - name: phi-4
+      max_input_tokens: 16384
+    - name: phi-4-mini-instruct
+      max_input_tokens: 131072
+    - name: phi-4-reasoning
+      max_input_tokens: 33792
+    - name: phi-4-mini-reasoning
+      max_input_tokens: 131072
+    - name: grok-3
+      max_input_tokens: 131072
+    - name: grok-3-mini
+      max_input_tokens: 131072
 
 - provider: deepinfra
   models:
@@ -817,6 +1352,63 @@
       max_tokens_per_chunk: 512
       default_chunk_size: 1000
       max_batch_size: 100
+    # legacy — pruning candidates
+    - name: Qwen/Qwen3-Max-Thinking
+      max_input_tokens: 262144
+      input_price: 1.2
+      output_price: 6
+      supports_function_calling: true
+    - name: Qwen/Qwen3-Next-80B-A3B-Thinking
+      max_input_tokens: 262144
+      input_price: 0.14
+      output_price: 1.4
+    - name: Qwen/Qwen3-235B-A22B-Thinking-2507
+      max_input_tokens: 131072
+      input_price: 0.13
+      output_price: 0.6
+    - name: Qwen/Qwen3-Coder-480B-A35B-Instruct
+      max_input_tokens: 131072
+      input_price: 0.4
+      output_price: 1.6
+      supports_function_calling: true
+    - name: Qwen/Qwen3-Coder-30B-A3B-Instruct
+      max_input_tokens: 262144
+      input_price: 0.07
+      output_price: 0.27
+      supports_function_calling: true
+    - name: Qwen/Qwen3-VL-8B-Instruct
+      max_input_tokens: 262144
+      input_price: 0.18
+      output_price: 0.69
+      supports_vision: true
+    - name: google/gemma-3-27b-it
+      max_input_tokens: 131072
+      input_price: 0.1
+      output_price: 0.2
+    - name: mistralai/Mistral-Small-3.2-24B-Instruct-2506
+      max_input_tokens: 32768
+      input_price: 0.06
+      output_price: 0.12
+    - name: moonshotai/Kimi-K2-Instruct-0905
+      max_input_tokens: 262144
+      input_price: 0.5
+      output_price: 2.0
+      supports_function_calling: true
+    - name: moonshotai/Kimi-K2-Thinking
+      max_input_tokens: 262144
+      input_price: 0.55
+      output_price: 2.5
+      supports_function_calling: true
+    - name: MiniMaxAI/MiniMax-M2.1
+      max_input_tokens: 196608
+      input_price: 0.27
+      output_price: 0.95
+      supports_function_calling: true
+    - name: zai-org/GLM-4.6V
+      max_input_tokens: 131072
+      input_price: 0.3
+      output_price: 0.9
+      supports_vision: true
 
 - provider: jina
   models:

scripts/requirements.txt

@@ -1,2 +0,0 @@
-requests
-ruamel.yaml

scripts/update_models.py

@@ -1,255 +0,0 @@
-import requests
-import sys
-import re
-import json
-
-# Provider mapping from models.yaml to OpenRouter prefixes
-PROVIDER_MAPPING = {
-    "openai": "openai",
-    "claude": "anthropic",
-    "gemini": "google",
-    "mistral": "mistralai",
-    "cohere": "cohere",
-    "perplexity": "perplexity",
-    "xai": "x-ai",
-    "openrouter": "openrouter",
-    "ai21": "ai21",
-    "deepseek": "deepseek",
-    "moonshot": "moonshotai",
-    "qianwen": "qwen",
-    "zhipuai": "zhipuai",
-    "minimax": "minimax",
-    "vertexai": "google",
-    "groq": "groq",
-    "bedrock": "amazon",
-    "hunyuan": "tencent",
-    "ernie": "baidu",
-    "github": "github",
-}
-
-def fetch_openrouter_models():
-    print("Fetching models from OpenRouter...")
-    try:
-        response = requests.get("https://openrouter.ai/api/v1/models")
-        response.raise_for_status()
-        data = response.json()["data"]
-        print(f"Fetched {len(data)} models.")
-        return data
-    except Exception as e:
-        print(f"Error fetching models: {e}")
-        sys.exit(1)
-
-def get_openrouter_model(models_data, provider_prefix, model_name, is_openrouter_provider=False):
-    if is_openrouter_provider:
-        # For openrouter provider, the model_name in yaml is usually the full ID
-        for model in models_data:
-            if model["id"] == model_name:
-                return model
-        return None
-
-    expected_id = f"{provider_prefix}/{model_name}"
-    
-    # 1. Try exact match on ID
-    for model in models_data:
-        if model["id"] == expected_id:
-            return model
-            
-    # 2. Try match by suffix
-    for model in models_data:
-        if model["id"].split("/")[-1] == model_name:
-            if model["id"].startswith(f"{provider_prefix}/"):
-                return model
-
-    return None
-
-def format_price(price_per_token):
-    if price_per_token is None:
-        return None
-    try:
-        price_per_1m = float(price_per_token) * 1_000_000
-        if price_per_1m.is_integer():
-            return str(int(price_per_1m))
-        else:
-            return str(round(price_per_1m, 4))
-    except:
-        return None
-
-def get_indentation(line):
-    return len(line) - len(line.lstrip())
-
-def process_model_block(block_lines, current_provider, or_models):
-    if not block_lines:
-        return []
-
-    # 1. Identify model name and indentation
-    name_line = block_lines[0]
-    name_match = re.match(r"^(\s*)-\s*name:\s*(.+)$", name_line)
-    if not name_match:
-        return block_lines 
-
-    name_indent_str = name_match.group(1)
-    model_name = name_match.group(2).strip()
-    
-    # 2. Find OpenRouter model
-    or_prefix = PROVIDER_MAPPING.get(current_provider)
-    is_openrouter_provider = (current_provider == "openrouter")
-    
-    if not or_prefix and not is_openrouter_provider:
-        return block_lines
-        
-    or_model = get_openrouter_model(or_models, or_prefix, model_name, is_openrouter_provider)
-    if not or_model:
-        return block_lines
-
-    print(f"  Updating {model_name}...")
-
-    # 3. Prepare updates
-    updates = {}
-    
-    # Pricing
-    pricing = or_model.get("pricing", {})
-    p_in = format_price(pricing.get("prompt"))
-    p_out = format_price(pricing.get("completion"))
-    if p_in: updates["input_price"] = p_in
-    if p_out: updates["output_price"] = p_out
-    
-    # Context
-    ctx = or_model.get("context_length")
-    if ctx: updates["max_input_tokens"] = str(ctx)
-    
-    max_out = None
-    if "top_provider" in or_model and or_model["top_provider"]:
-        max_out = or_model["top_provider"].get("max_completion_tokens")
-    if max_out: updates["max_output_tokens"] = str(max_out)
-    
-    # Capabilities
-    arch = or_model.get("architecture", {})
-    modality = arch.get("modality", "")
-    if "image" in modality:
-        updates["supports_vision"] = "true"
-
-    # 4. Detect field indentation
-    field_indent_str = None
-    existing_fields = {} # key -> line_index
-    
-    for i, line in enumerate(block_lines):
-        if i == 0: continue # Skip name line
-        
-        # Skip comments
-        if line.strip().startswith("#"):
-            continue
-            
-        # Look for "key: value"
-        m = re.match(r"^(\s*)([\w_-]+):", line)
-        if m:
-            indent = m.group(1)
-            key = m.group(2)
-            # Must be deeper than name line
-            if len(indent) > len(name_indent_str):
-                if field_indent_str is None:
-                    field_indent_str = indent
-                existing_fields[key] = i
-
-    if field_indent_str is None:
-        field_indent_str = name_indent_str + "  "
-
-    # 5. Apply updates
-    new_block = list(block_lines)
-    
-    # Update existing fields
-    for key, value in updates.items():
-        if key in existing_fields:
-            idx = existing_fields[key]
-            # Preserve original key indentation exactly
-            original_line = new_block[idx]
-            m = re.match(r"^(\s*)([\w_-]+):", original_line)
-            if m:
-                current_indent = m.group(1)
-                new_block[idx] = f"{current_indent}{key}: {value}\n"
-    
-    # Insert missing fields
-    # Insert after the name line
-    insertion_idx = 1
-    
-    for key, value in updates.items():
-        if key not in existing_fields:
-            new_line = f"{field_indent_str}{key}: {value}\n"
-            new_block.insert(insertion_idx, new_line)
-            insertion_idx += 1
-            
-    return new_block
-
-def main():
-    or_models = fetch_openrouter_models()
-    
-    print("Reading models.yaml...")
-    with open("models.yaml", "r") as f:
-        lines = f.readlines()
-        
-    new_lines = []
-    current_provider = None
-    
-    i = 0
-    while i < len(lines):
-        line = lines[i]
-        
-        # Check for provider
-        # - provider: name
-        p_match = re.match(r"^\s*-?\s*provider:\s*(.+)$", line)
-        if p_match:
-            current_provider = p_match.group(1).strip()
-            new_lines.append(line)
-            i += 1
-            continue
-            
-        # Check for model start
-        # - name: ...
-        m_match = re.match(r"^(\s*)-\s*name:\s*.+$", line)
-        if m_match:
-            # Start of a model block
-            start_indent = len(m_match.group(1))
-            
-            # Collect block lines
-            block_lines = [line]
-            j = i + 1
-            while j < len(lines):
-                next_line = lines[j]
-                stripped = next_line.strip()
-                
-                # If empty or comment, include it
-                if not stripped or stripped.startswith("#"):
-                    block_lines.append(next_line)
-                    j += 1
-                    continue
-                
-                # Check indentation
-                next_indent = get_indentation(next_line)
-                
-                # If indentation is greater, it's part of the block (property)
-                if next_indent > start_indent:
-                    block_lines.append(next_line)
-                    j += 1
-                    continue
-                
-                # If indentation is equal or less, it's the end of the block
-                break
-            
-            # Process the block
-            processed_block = process_model_block(block_lines, current_provider, or_models)
-            new_lines.extend(processed_block)
-            
-            # Advance i
-            i = j
-            continue
-            
-        # Otherwise, just a regular line
-        new_lines.append(line)
-        i += 1
-        
-    print("Saving models.yaml...")
-    with open("models.yaml", "w") as f:
-        f.writelines(new_lines)
-    print("Done.")
-
-if __name__ == "__main__":
-    main()

src/cli/mod.rs

@@ -12,6 +12,7 @@ use clap_complete::ArgValueCompleter;
 use is_terminal::IsTerminal;
 use std::collections::HashSet;
 use std::io::{Read, stdin};
+use std::path::PathBuf;
 
 #[derive(clap::ValueEnum, Debug, Clone, Copy, PartialEq, Eq)]
 pub enum McpTransportArg {
@@ -53,7 +54,7 @@ pub enum McpScopeArg {
 				"install_builtins", "sync_models", "list_models", "list_roles",
 				"list_sessions", "list_agents", "list_rags", "list_macros",
 				"list_skills", "list_bundles", "skill", "tail_logs", "completions",
-				"update", "update_bundle", "uninstall",
+				"update", "update_bundle", "uninstall", "generate_models",
 			])
 	),
 	group(
@@ -253,6 +254,15 @@ pub struct Cli {
     /// Sync models updates
     #[arg(long, help_heading = "Installation & Updates")]
     pub sync_models: bool,
+    /// Regenerate the models catalog file from live API data
+    #[arg(
+        long,
+        hide = true,
+        value_name = "PATH",
+        num_args = 0..=1,
+        default_missing_value = "models.yaml"
+    )]
+    pub generate_models: Option<PathBuf>,
     /// Update Coyote to the latest release, or to a specific version
     #[arg(long, value_name = "VERSION", help_heading = "Installation & Updates")]
     pub update: Option<Option<String>>,
@@ -810,6 +820,19 @@ mod tests {
         assert!(cli.sync_models);
     }
 
+    #[test]
+    fn parse_generate_models_flag() {
+        assert_eq!(parse(&[]).generate_models, None);
+        assert_eq!(
+            parse(&["--generate-models"]).generate_models,
+            Some(PathBuf::from("models.yaml"))
+        );
+        assert_eq!(
+            parse(&["--generate-models", "out.yaml"]).generate_models,
+            Some(PathBuf::from("out.yaml"))
+        );
+    }
+
     #[test]
     fn parse_model_with_role() {
         let cli = parse(&["-m", "gpt-4o", "-r", "coder"]);

src/client/catalog.rs

@@ -13,7 +13,6 @@ use std::sync::LazyLock;
 
 pub const QUIRKS_YAML: &str = include_str!("../../quirks.yaml");
 
-#[allow(dead_code)]
 pub static ALL_QUIRKS: LazyLock<Vec<ProviderQuirks>> =
     LazyLock::new(|| serde_yaml::from_str(QUIRKS_YAML).expect("quirks.yaml must be valid"));
 
@@ -133,8 +132,8 @@ const PROVIDER_SOURCES: [SourceIds; 24] = [
     },
     SourceIds {
         provider: "github",
-        modelsdev: Some("github-copilot"),
+        modelsdev: None,
-        openrouter: Some("github"),
+        openrouter: None,
     },
     SourceIds {
         provider: "deepinfra",
@@ -162,7 +161,6 @@ pub struct FetchedSources {
 
 /// Fetches both upstream catalogs concurrently. A failed fetch is logged and
 /// surfaces as `None` rather than an error.
-#[allow(dead_code)]
 pub async fn fetch_sources() -> FetchedSources {
     let (modelsdev, openrouter) = tokio::join!(fetch(MODELSDEV_API_URL), fetch(OPENROUTER_API_URL));
     FetchedSources {
@@ -478,7 +476,6 @@ pub fn build_catalog(sources: &FetchedSources, quirks: &[ProviderQuirks]) -> Cat
 
 /// Like [`build_catalog`], but errors unless every provider was assembled
 /// from live API data (or is quirks-owned by design).
-#[allow(dead_code)]
 pub fn build_catalog_strict(
     sources: &FetchedSources,
     quirks: &[ProviderQuirks],
@@ -494,6 +491,32 @@ pub fn build_catalog_strict(
     Ok(build.providers)
 }
 
+const GENERATED_HEADER: &str = "# GENERATED by 'coyote --generate-models' — DO NOT hand-edit model data.\n\
+# Hand-owned behavior lives in quirks.yaml; this file is the embedded offline fallback.\n";
+
+/// Builds the catalog from live API data and writes it to `path`. Both
+/// upstream sources must respond and every provider must fully resolve;
+/// identical source data always produces identical bytes.
+pub async fn generate_models_file(path: &std::path::Path) -> Result<()> {
+    let sources = fetch_sources().await;
+    if sources.modelsdev.is_none() || sources.openrouter.is_none() {
+        bail!(
+            "generating the models file requires both catalog sources; see the log for the failed fetch"
+        );
+    }
+    let providers = build_catalog_strict(&sources, &ALL_QUIRKS)?;
+    let model_count: usize = providers.iter().map(|provider| provider.models.len()).sum();
+    let yaml = serde_yaml::to_string(&providers)?;
+    std::fs::write(path, format!("{GENERATED_HEADER}{yaml}"))?;
+    println!(
+        "Wrote {} models across {} providers to {}",
+        model_count,
+        providers.len(),
+        path.display()
+    );
+    Ok(())
+}
+
 /// Fraction of `old` names that also appear in `new`; empty `old` yields 1.0.
 /// Used to guard against upstream id-shape drift: a low overlap means the
 /// upstream catalog renamed its models out from under us.
@@ -1333,6 +1356,8 @@ rules:
   models: [{name: e}]
 - provider: hunyuan
   models: [{name: h}]
+- provider: github
+  models: [{name: g}]
 - provider: jina
   models: [{name: j}]
 - provider: voyageai
@@ -1358,6 +1383,96 @@ rules:
         assert_eq!(name_overlap(&old, &[]), 0.0);
     }
 
+    fn embedded_catalog() -> Vec<ProviderModels> {
+        serde_yaml::from_str(crate::client::MODELS_YAML).unwrap()
+    }
+
+    fn embedded_provider<'a>(providers: &'a [ProviderModels], name: &str) -> &'a ProviderModels {
+        providers
+            .iter()
+            .find(|provider| provider.provider == name)
+            .unwrap_or_else(|| panic!("provider '{name}' missing from models.yaml"))
+    }
+
+    #[test]
+    fn embedded_catalog_invariants() {
+        let providers = embedded_catalog();
+        assert!(providers.len() >= 24);
+        let order: Vec<&str> = providers
+            .iter()
+            .map(|provider| provider.provider.as_str())
+            .collect();
+        assert_eq!(order, KNOWN_PROVIDERS);
+
+        assert!(embedded_provider(&providers, "xai").oauth.is_some());
+        for name in ["jina", "voyageai"] {
+            assert!(
+                embedded_provider(&providers, name)
+                    .models
+                    .iter()
+                    .any(|model| model.model_type == "embedding"),
+                "{name} lost its embedding models"
+            );
+        }
+        assert!(
+            embedded_provider(&providers, "cohere")
+                .models
+                .iter()
+                .any(|model| model.model_type == "reranker"),
+            "cohere lost its reranker models"
+        );
+
+        for provider in &providers {
+            for model in &provider.models {
+                // A :thinking model without a patch must at least alias a real
+                // wire name (qianwen twins invert a base-model patch instead).
+                if model.name.ends_with(":thinking") {
+                    assert!(
+                        model.patch.is_some() || model.real_name.is_some(),
+                        "{}:{} has neither patch nor real_name",
+                        provider.provider,
+                        model.name
+                    );
+                }
+                if model.model_type == "chat" {
+                    let serialized = serde_yaml::to_value(model).unwrap();
+                    assert!(
+                        serialized
+                            .as_mapping()
+                            .unwrap()
+                            .keys()
+                            .any(|key| !matches!(key.as_str(), Some("name" | "type"))),
+                        "{}:{} is an all-default chat entry",
+                        provider.provider,
+                        model.name
+                    );
+                }
+            }
+        }
+
+        for model in &embedded_provider(&providers, "claude").models {
+            if model.model_type == "chat" {
+                assert!(
+                    model.require_max_tokens,
+                    "claude:{} is missing require_max_tokens",
+                    model.name
+                );
+            }
+        }
+    }
+
+    #[test]
+    fn embedded_catalog_roundtrip() {
+        let providers = embedded_catalog();
+        let yaml = serde_yaml::to_string(&providers).unwrap();
+        let reparsed: Vec<ProviderModels> = serde_yaml::from_str(&yaml).unwrap();
+        assert_eq!(reparsed.len(), providers.len());
+        for (reparsed, original) in reparsed.iter().zip(&providers) {
+            assert_eq!(reparsed.provider, original.provider);
+            assert_eq!(reparsed.models.len(), original.models.len());
+        }
+    }
+
     /// Guards against upstream id-shape drift by comparing live API model
     /// names against the embedded models.yaml. Run manually:
     /// `cargo test idshape_guard_live -- --ignored --nocapture`

src/client/model.rs

@@ -358,7 +358,7 @@ impl ModelData {
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct ProviderModels {
     pub provider: String,
-    #[serde(default)]
+    #[serde(default, skip_serializing_if = "Option::is_none")]
     pub oauth: Option<OAuthConfig>,
     pub models: Vec<ModelData>,
 }

src/main.rs

@@ -20,7 +20,7 @@ extern crate log;
 
 use crate::cli::Cli;
 use crate::client::{
-    ModelType, call_chat_completions, call_chat_completions_streaming, list_models, oauth,
+    ModelType, call_chat_completions, call_chat_completions_streaming, catalog, list_models, oauth,
 };
 use crate::config::instructions::WORKSPACE_INSTRUCTIONS_FILE_NAME;
 use crate::config::{
@@ -101,6 +101,7 @@ async fn main() -> Result<()> {
 
     let info_flag = cli.info
         || cli.sync_models
+        || cli.generate_models.is_some()
         || cli.list_models
         || cli.list_roles
         || cli.list_agents
@@ -295,6 +296,10 @@ async fn run(
         return sync_models(&url, abort_signal.clone()).await;
     }
 
+    if let Some(path) = &cli.generate_models {
+        return catalog::generate_models_file(path).await;
+    }
+
     if cli.list_models {
         for model in list_models(ctx.app.config.as_ref(), ModelType::Chat) {
             println!("{}", model.id());