test(zen): align provider tests with API

conf/zen_models_live.json

@@ -0,0 +1,696 @@
+{
+  "_README": {
+    "description": "Generated baseline OpenCode Zen catalogue for PAL MCP Server.",
+    "source": "https://opencode.ai/zen/v1/models",
+    "usage": "Generated by scripts/sync_zen_models.py. Curated overrides belong in conf/zen_models.json.",
+    "field_notes": "Entries are conservative discovery data. Curated manifest values override these at runtime."
+  },
+  "models": [
+    {
+      "model_name": "big-pickle",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 32000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Big Pickle via OpenCode Zen - Stealth model for coding tasks",
+      "intelligence_score": 13,
+      "allow_code_generation": true
+    },
+    {
+      "model_name": "claude-3-5-haiku",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model claude-3-5-haiku.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "claude-haiku-4-5",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 5.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Claude Haiku 4.5 via OpenCode Zen - Fast and efficient for coding tasks",
+      "intelligence_score": 16,
+      "allow_code_generation": true
+    },
+    {
+      "model_name": "claude-opus-4-1",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model claude-opus-4-1.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "claude-opus-4-5",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 5.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Claude Opus 4.5 via OpenCode Zen - Anthropic's frontier reasoning model for complex software engineering",
+      "intelligence_score": 18,
+      "allow_code_generation": true
+    },
+    {
+      "model_name": "claude-opus-4-6",
+      "aliases": [],
+      "context_window": 1000000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model claude-opus-4-6.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "claude-sonnet-4",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model claude-sonnet-4.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "claude-sonnet-4-5",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 5.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Claude Sonnet 4.5 via OpenCode Zen - Balanced performance for coding and general tasks",
+      "intelligence_score": 17,
+      "allow_code_generation": true
+    },
+    {
+      "model_name": "claude-sonnet-4-6",
+      "aliases": [],
+      "context_window": 1000000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model claude-sonnet-4-6.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "gemini-3-flash",
+      "aliases": [],
+      "context_window": 1048576,
+      "max_output_tokens": 65536,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gemini-3-flash.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "gemini-3-pro",
+      "aliases": [],
+      "context_window": 1000000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 10.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Gemini 3 Pro via OpenCode Zen - Google's multimodal model with large context",
+      "intelligence_score": 16,
+      "allow_code_generation": true
+    },
+    {
+      "model_name": "gemini-3.1-pro",
+      "aliases": [],
+      "context_window": 1048576,
+      "max_output_tokens": 65536,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gemini-3.1-pro.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "glm-4.6",
+      "aliases": [],
+      "context_window": 205000,
+      "max_output_tokens": 32000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "GLM 4.6 via OpenCode Zen - High-performance model for coding and reasoning",
+      "intelligence_score": 15,
+      "allow_code_generation": true
+    },
+    {
+      "model_name": "glm-4.7",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model glm-4.7.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "glm-5",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model glm-5.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "gpt-5",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5-codex",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5-codex.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5-nano",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 32000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "GPT 5 Nano via OpenCode Zen - Lightweight GPT model",
+      "intelligence_score": 12,
+      "allow_code_generation": true
+    },
+    {
+      "model_name": "gpt-5.1",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "GPT 5.1 via OpenCode Zen - Latest GPT model for general AI tasks",
+      "intelligence_score": 16,
+      "allow_code_generation": true,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5.1-codex",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "GPT 5.1 Codex via OpenCode Zen - Specialized for code generation and understanding",
+      "intelligence_score": 17,
+      "allow_code_generation": true,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5.1-codex-max",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5.1-codex-max.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5.1-codex-mini",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5.1-codex-mini.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5.2",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5.2.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5.2-codex",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5.2-codex.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5.3-codex",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5.3-codex.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5.3-codex-spark",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5.3-codex-spark.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5.4",
+      "aliases": [],
+      "context_window": 1050000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5.4.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5.4-mini",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5.4-mini.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5.4-nano",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5.4-nano.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "gpt-5.4-pro",
+      "aliases": [],
+      "context_window": 1050000,
+      "max_output_tokens": 128000,
+      "supports_extended_thinking": true,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model gpt-5.4-pro.",
+      "intelligence_score": 10,
+      "allow_code_generation": false,
+      "use_openai_response_api": true
+    },
+    {
+      "model_name": "kimi-k2",
+      "aliases": [],
+      "context_window": 400000,
+      "max_output_tokens": 32000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Kimi K2 via OpenCode Zen - Advanced reasoning model",
+      "intelligence_score": 15,
+      "allow_code_generation": true
+    },
+    {
+      "model_name": "kimi-k2-thinking",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model kimi-k2-thinking.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "kimi-k2.5",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model kimi-k2.5.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "mimo-v2-flash-free",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model mimo-v2-flash-free.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "mimo-v2-omni-free",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model mimo-v2-omni-free.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "mimo-v2-pro-free",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model mimo-v2-pro-free.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "minimax-m2.1",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model minimax-m2.1.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "minimax-m2.5",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model minimax-m2.5.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "minimax-m2.5-free",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model minimax-m2.5-free.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "nemotron-3-super-free",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model nemotron-3-super-free.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "qwen3.6-plus-free",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model qwen3.6-plus-free.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    },
+    {
+      "model_name": "trinity-large-preview-free",
+      "aliases": [],
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_json_mode": true,
+      "supports_function_calling": true,
+      "supports_images": false,
+      "max_image_size_mb": 0.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range",
+      "description": "Generated baseline metadata for OpenCode Zen model trinity-large-preview-free.",
+      "intelligence_score": 10,
+      "allow_code_generation": false
+    }
+  ]
+}

docs/configuration.md

@@ -79,6 +79,8 @@ DEFAULT_MODEL=auto  # Claude picks best model for each task (recommended)
   - `conf/openai_models.json` – OpenAI catalogue (can be overridden with `OPENAI_MODELS_CONFIG_PATH`)
   - `conf/gemini_models.json` – Gemini catalogue (`GEMINI_MODELS_CONFIG_PATH`)
   - `conf/xai_models.json` – X.AI / GROK catalogue (`XAI_MODELS_CONFIG_PATH`)
+  - `conf/zen_models.json` – Curated OpenCode Zen overrides (`ZEN_MODELS_CONFIG_PATH`)
+  - `conf/zen_models_live.json` – Generated live OpenCode Zen catalogue (`ZEN_LIVE_MODELS_CONFIG_PATH`)
   - `conf/openrouter_models.json` – Curated OpenRouter overrides (`OPENROUTER_MODELS_CONFIG_PATH`)
   - `conf/openrouter_models_live.json` – Generated live OpenRouter catalogue (`OPENROUTER_LIVE_MODELS_CONFIG_PATH`)
   - `conf/dial_models.json` – DIAL aggregation catalogue (`DIAL_MODELS_CONFIG_PATH`)
@@ -93,10 +95,11 @@ DEFAULT_MODEL=auto  # Claude picks best model for each task (recommended)
   | OpenAI | `gpt-5.2`, `gpt-5.1-codex`, `gpt-5.1-codex-mini`, `gpt-5`, `gpt-5.2-pro`, `gpt-5-mini`, `gpt-5-nano`, `gpt-5-codex`, `gpt-4.1`, `o3`, `o3-mini`, `o3-pro`, `o4-mini` | `gpt5.2`, `gpt-5.2`, `5.2`, `gpt5.1-codex`, `codex-5.1`, `codex-mini`, `gpt5`, `gpt5pro`, `mini`, `nano`, `codex`, `o3mini`, `o3pro`, `o4mini` |
   | Gemini | `gemini-2.5-pro`, `gemini-2.5-flash`, `gemini-2.0-flash`, `gemini-2.0-flash-lite` | `pro`, `gemini-pro`, `flash`, `flash-2.0`, `flashlite` |
   | X.AI | `grok-4`, `grok-4.1-fast` | `grok`, `grok4`, `grok-4.1-fast-reasoning` |
+  | OpenCode Zen | Generated live catalogue plus curated overrides | e.g., `zen-sonnet`, `zen-codex`, plus any curated aliases you add |
   | OpenRouter | Generated live catalogue plus curated overrides | e.g., `opus`, `sonnet`, `flash`, `pro`, `mistral` |
   | Custom | User-managed entries such as `llama3.2` | Define your own aliases per entry |
 
-  Latest OpenAI entries (`gpt-5.2`, `gpt-5.1-codex`, `gpt-5.1-codex-mini`, `gpt-5.2-pro`) expose 400K-token contexts with large outputs, reasoning-token support, and multimodal inputs. `gpt-5.1-codex` and `gpt-5.2-pro` are Responses-only with streaming disabled, while the base `gpt-5.2` and Codex mini support streaming along with full code-generation flags. For OpenRouter, keep PAL-specific metadata in the curated manifest and regenerate the live catalogue when OpenRouter adds or removes models; see [Refreshing the Live OpenRouter Catalogue](custom_models.md#refreshing-the-live-openrouter-catalogue).
+  Latest OpenAI entries (`gpt-5.2`, `gpt-5.1-codex`, `gpt-5.1-codex-mini`, `gpt-5.2-pro`) expose 400K-token contexts with large outputs, reasoning-token support, and multimodal inputs. `gpt-5.1-codex` and `gpt-5.2-pro` are Responses-only with streaming disabled, while the base `gpt-5.2` and Codex mini support streaming along with full code-generation flags. For OpenCode Zen and OpenRouter, keep PAL-specific metadata in the curated manifest and regenerate the live catalogue when the upstream provider adds or removes models; see [Refreshing the Live OpenCode Zen Catalogue](custom_models.md#refreshing-the-live-opencode-zen-catalogue) and [Refreshing the Live OpenRouter Catalogue](custom_models.md#refreshing-the-live-openrouter-catalogue).
 
   > **Tip:** Copy the JSON file you need, customise it, and point the corresponding `*_MODELS_CONFIG_PATH` environment variable to your version. This lets you enable or disable capabilities (JSON mode, function calling, temperature support, code generation) without editing Python.
 

docs/custom_models.md

@@ -221,13 +221,15 @@ CUSTOM_MODEL_NAME=your-loaded-model
 The system automatically routes models to the appropriate provider:
 
 1. Entries in `conf/custom_models.json` → Always routed through the Custom API (requires `CUSTOM_API_URL`)
-2. Entries in `conf/openrouter_models_live.json` and `conf/openrouter_models.json` → Routed through OpenRouter (requires `OPENROUTER_API_KEY`)
+2. Entries in `conf/zen_models_live.json` and `conf/zen_models.json` → Routed through OpenCode Zen (requires `ZEN_API_KEY`)
-3. **Unknown models** → Fallback logic based on model name patterns
+3. Entries in `conf/openrouter_models_live.json` and `conf/openrouter_models.json` → Routed through OpenRouter (requires `OPENROUTER_API_KEY`)
+4. **Unknown models** → Fallback logic based on model name patterns
 
 **Provider Priority Order:**
 1. Native APIs (Google, OpenAI) - if API keys are available
-2. Custom endpoints - for models declared in `conf/custom_models.json`  
+2. OpenCode Zen - curated gateway models
-3. OpenRouter - catch-all for cloud models
+3. Custom endpoints - for models declared in `conf/custom_models.json`
+4. OpenRouter - catch-all for cloud models
 
 This ensures clean separation between local and cloud models while maintaining flexibility for unknown models.
 
@@ -241,7 +243,42 @@ These JSON files define model aliases and capabilities. You can:
 
 ### Adding Custom Models
 
-Edit `conf/openrouter_models.json` to tweak OpenRouter behaviour or `conf/custom_models.json` to add local models. The generated `conf/openrouter_models_live.json` file is discovery data from OpenRouter's `/api/v1/models` endpoint; curated entries in `conf/openrouter_models.json` override those generated defaults. Each entry maps directly onto [`ModelCapabilities`](../providers/shared/model_capabilities.py).
+Edit `conf/zen_models.json` to tweak OpenCode Zen behaviour, `conf/openrouter_models.json` to tweak OpenRouter behaviour, or `conf/custom_models.json` to add local models. The generated `conf/zen_models_live.json` and `conf/openrouter_models_live.json` files are discovery data from the upstream model listing endpoints; curated entries override those generated defaults. Each entry maps directly onto [`ModelCapabilities`](../providers/shared/model_capabilities.py).
+
+### Refreshing the Live OpenCode Zen Catalogue
+
+Run the sync script whenever OpenCode Zen adds or removes models that you want `listmodels` and provider enumeration to expose, or before cutting a release that should include an updated Zen catalogue.
+
+```bash
+source .pal_venv/bin/activate
+python scripts/sync_zen_models.py
+```
+
+By default the script:
+
+- fetches `https://opencode.ai/zen/v1/models`
+- writes conservative discovery data to `conf/zen_models_live.json`
+- leaves `conf/zen_models.json` untouched
+
+Use the optional flags if you need to test against a different endpoint or write to a different file:
+
+```bash
+python scripts/sync_zen_models.py --url https://opencode.ai/zen/v1/models --output conf/zen_models_live.json
+```
+
+Important runtime behavior:
+
+- `conf/zen_models_live.json` is the generated baseline catalogue
+- `conf/zen_models.json` is the curated override layer for aliases and PAL-specific capability flags
+- curated entries win when the same `model_name` appears in both files
+- models missing from the curated file are still available from the generated catalogue
+
+After refreshing the catalogue:
+
+1. Review the diff in `conf/zen_models_live.json`
+2. Add or update curated entries in `conf/zen_models.json` if a new model needs aliases or PAL-specific capability tweaks
+3. Restart the server so the updated manifests are reloaded
+4. Commit the generated JSON alongside any curated overrides so other contributors get the same catalogue state
 
 ### Refreshing the Live OpenRouter Catalogue
 

providers/registries/zen.py

@@ -2,14 +2,48 @@
 
 from __future__ import annotations
 
+import importlib.resources
+import json
+import logging
+from pathlib import Path
+
+from utils.env import get_env
+from utils.file_utils import read_json_file
+
 from ..shared import ModelCapabilities, ProviderType
 from .base import CAPABILITY_FIELD_NAMES, CapabilityModelRegistry
 
 
+logger = logging.getLogger(__name__)
+
+
 class ZenModelRegistry(CapabilityModelRegistry):
     """Capability registry backed by ``conf/zen_models.json``."""
 
-    def __init__(self, config_path: str | None = None) -> None:
+    LIVE_ENV_VAR_NAME = "ZEN_LIVE_MODELS_CONFIG_PATH"
+    LIVE_DEFAULT_FILENAME = "zen_models_live.json"
+
+    def __init__(self, config_path: str | None = None, live_config_path: str | None = None) -> None:
+        self._live_resource = None
+        self._live_config_path: Path | None = None
+        self._live_default_path = Path(__file__).resolve().parents[3] / "conf" / self.LIVE_DEFAULT_FILENAME
+
+        if live_config_path:
+            self._live_config_path = Path(live_config_path)
+        else:
+            env_path = get_env(self.LIVE_ENV_VAR_NAME)
+            if env_path:
+                self._live_config_path = Path(env_path)
+            else:
+                try:
+                    resource = importlib.resources.files("conf").joinpath(self.LIVE_DEFAULT_FILENAME)
+                    if hasattr(resource, "read_text"):
+                        self._live_resource = resource
+                    else:
+                        raise AttributeError("resource accessor not available")
+                except Exception:
+                    self._live_config_path = self._live_default_path
+
         super().__init__(
             env_var_name="ZEN_MODELS_CONFIG_PATH",
             default_filename="zen_models.json",
@@ -18,6 +52,60 @@ class ZenModelRegistry(CapabilityModelRegistry):
             config_path=config_path,
         )
 
+    def reload(self) -> None:
+        live_data = self._load_live_config_data()
+        curated_data = self._load_config_data()
+        merged_data = self._merge_manifest_data(live_data, curated_data)
+
+        self._extras = {}
+        configs = [config for config in self._parse_models(merged_data) if config is not None]
+        self._build_maps(configs)
+
+    def _load_live_config_data(self) -> dict:
+        if self._live_resource is not None:
+            try:
+                if hasattr(self._live_resource, "read_text"):
+                    config_text = self._live_resource.read_text(encoding="utf-8")
+                else:
+                    with self._live_resource.open("r", encoding="utf-8") as handle:
+                        config_text = handle.read()
+                data = json.loads(config_text)
+            except FileNotFoundError:
+                logger.debug("Packaged %s not found", self.LIVE_DEFAULT_FILENAME)
+                return {"models": []}
+            except Exception as exc:
+                logger.warning("Failed to read packaged %s: %s", self.LIVE_DEFAULT_FILENAME, exc)
+                return {"models": []}
+            return data or {"models": []}
+
+        if not self._live_config_path:
+            return {"models": []}
+
+        if not self._live_config_path.exists():
+            logger.debug("Zen live registry config not found at %s", self._live_config_path)
+            return {"models": []}
+
+        data = read_json_file(str(self._live_config_path))
+        return data or {"models": []}
+
+    @staticmethod
+    def _merge_manifest_data(live_data: dict, curated_data: dict) -> dict:
+        merged_models: dict[str, dict] = {}
+
+        for source in (live_data, curated_data):
+            for raw in source.get("models", []):
+                if not isinstance(raw, dict):
+                    continue
+
+                model_name = raw.get("model_name")
+                if not model_name:
+                    continue
+
+                existing = merged_models.get(model_name, {})
+                merged_models[model_name] = {**existing, **dict(raw)}
+
+        return {"models": list(merged_models.values())}
+
     def _finalise_entry(self, entry: dict) -> tuple[ModelCapabilities, dict]:
         provider_override = entry.get("provider")
         if isinstance(provider_override, str):

scripts/sync_zen_models.py

@@ -0,0 +1,237 @@
+#!/usr/bin/env python3
+
+import argparse
+import json
+import sys
+from pathlib import Path
+from urllib.error import HTTPError, URLError
+from urllib.request import Request, urlopen
+
+from utils.env import get_env
+from utils.file_utils import read_json_file
+
+
+ROOT = Path(__file__).resolve().parents[1]
+DEFAULT_OUTPUT = ROOT / "conf" / "zen_models_live.json"
+DEFAULT_CURATED = ROOT / "conf" / "zen_models.json"
+ZEN_MODELS_URL = "https://opencode.ai/zen/v1/models"
+
+
+def fetch_zen_models(url: str, api_key: str) -> dict:
+    request = Request(
+        url,
+        headers={
+            "Accept": "application/json",
+            "Authorization": f"Bearer {api_key}",
+            "User-Agent": "pal-mcp-server/zen-model-sync",
+        },
+    )
+
+    with urlopen(request, timeout=30) as response:
+        charset = response.headers.get_content_charset("utf-8")
+        payload = response.read().decode(charset)
+
+    data = json.loads(payload)
+    if not isinstance(data, dict):
+        raise ValueError("Zen models payload must be a JSON object")
+    return data
+
+
+def load_curated_models(path: Path) -> dict[str, dict]:
+    if not path.exists():
+        return {}
+
+    data = read_json_file(str(path)) or {}
+    models = data.get("models", [])
+    if not isinstance(models, list):
+        return {}
+
+    curated: dict[str, dict] = {}
+    for item in models:
+        if not isinstance(item, dict):
+            continue
+        model_name = item.get("model_name")
+        if isinstance(model_name, str) and model_name:
+            curated[model_name] = dict(item)
+    return curated
+
+
+def _infer_defaults_from_model_name(model_name: str) -> dict:
+    lower_name = model_name.lower()
+
+    defaults = {
+        "aliases": [],
+        "context_window": 200000,
+        "max_output_tokens": 64000,
+        "supports_extended_thinking": False,
+        "supports_json_mode": True,
+        "supports_function_calling": True,
+        "supports_images": False,
+        "max_image_size_mb": 0.0,
+        "supports_temperature": True,
+        "temperature_constraint": "range",
+        "description": f"OpenCode Zen live model: {model_name}",
+        "intelligence_score": 10,
+        "allow_code_generation": False,
+    }
+
+    if lower_name.startswith("claude-"):
+        defaults.update(
+            {
+                "supports_extended_thinking": True,
+                "supports_images": True,
+                "max_image_size_mb": 20.0,
+            }
+        )
+        if "-4-6" in lower_name:
+            defaults.update({"context_window": 1000000, "max_output_tokens": 128000})
+    elif lower_name.startswith("gemini-"):
+        defaults.update(
+            {
+                "context_window": 1048576,
+                "max_output_tokens": 65536,
+                "supports_extended_thinking": True,
+                "supports_images": True,
+                "max_image_size_mb": 20.0,
+            }
+        )
+    elif lower_name.startswith("gpt-"):
+        defaults.update(
+            {
+                "context_window": 400000,
+                "max_output_tokens": 128000,
+                "supports_extended_thinking": True,
+                "supports_images": True,
+                "max_image_size_mb": 20.0,
+            }
+        )
+        if "5.4" in lower_name:
+            defaults["context_window"] = 1050000 if "-pro" in lower_name or lower_name == "gpt-5.4" else 400000
+        if lower_name in {
+            "gpt-5.4",
+            "gpt-5.4-pro",
+            "gpt-5.4-mini",
+            "gpt-5.4-nano",
+            "gpt-5.3-codex",
+            "gpt-5.3-codex-spark",
+            "gpt-5.2",
+            "gpt-5.2-codex",
+            "gpt-5.1",
+            "gpt-5.1-codex",
+            "gpt-5.1-codex-max",
+            "gpt-5.1-codex-mini",
+            "gpt-5",
+            "gpt-5-codex",
+        }:
+            defaults["use_openai_response_api"] = True
+
+    return defaults
+
+
+def convert_model(model_data: dict, curated_models: dict[str, dict]) -> dict | None:
+    model_name = model_data.get("id")
+    if not isinstance(model_name, str) or not model_name:
+        return None
+
+    curated_entry = curated_models.get(model_name, {})
+    defaults = _infer_defaults_from_model_name(model_name)
+
+    return {
+        "model_name": model_name,
+        "aliases": [],
+        "context_window": int(curated_entry.get("context_window", defaults["context_window"])),
+        "max_output_tokens": int(curated_entry.get("max_output_tokens", defaults["max_output_tokens"])),
+        "supports_extended_thinking": bool(
+            curated_entry.get("supports_extended_thinking", defaults["supports_extended_thinking"])
+        ),
+        "supports_json_mode": bool(curated_entry.get("supports_json_mode", defaults["supports_json_mode"])),
+        "supports_function_calling": bool(
+            curated_entry.get("supports_function_calling", defaults["supports_function_calling"])
+        ),
+        "supports_images": bool(curated_entry.get("supports_images", defaults["supports_images"])),
+        "max_image_size_mb": float(curated_entry.get("max_image_size_mb", defaults["max_image_size_mb"])),
+        "supports_temperature": bool(curated_entry.get("supports_temperature", defaults["supports_temperature"])),
+        "temperature_constraint": curated_entry.get("temperature_constraint", defaults["temperature_constraint"]),
+        "description": curated_entry.get("description")
+        or f"Generated baseline metadata for OpenCode Zen model {model_name}.",
+        "intelligence_score": int(curated_entry.get("intelligence_score", defaults["intelligence_score"])),
+        "allow_code_generation": bool(curated_entry.get("allow_code_generation", defaults["allow_code_generation"])),
+        **(
+            {"use_openai_response_api": bool(curated_entry.get("use_openai_response_api", True))}
+            if curated_entry.get("use_openai_response_api") is not None or defaults.get("use_openai_response_api")
+            else {}
+        ),
+    }
+
+
+def build_output_document(source: dict, source_url: str, curated_models: dict[str, dict]) -> dict:
+    models = []
+    for model_data in source.get("data", []):
+        if not isinstance(model_data, dict):
+            continue
+
+        converted = convert_model(model_data, curated_models)
+        if converted:
+            models.append(converted)
+
+    models.sort(key=lambda item: item["model_name"])
+
+    return {
+        "_README": {
+            "description": "Generated baseline OpenCode Zen catalogue for PAL MCP Server.",
+            "source": source_url,
+            "usage": "Generated by scripts/sync_zen_models.py. Curated overrides belong in conf/zen_models.json.",
+            "field_notes": "Entries are conservative discovery data. Curated manifest values override these at runtime.",
+        },
+        "models": models,
+    }
+
+
+def write_output(path: Path, document: dict) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    with path.open("w", encoding="utf-8", newline="\n") as handle:
+        json.dump(document, handle, indent=2, ensure_ascii=False)
+        handle.write("\n")
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description="Sync OpenCode Zen live model catalogue into PAL config.")
+    parser.add_argument("--url", default=ZEN_MODELS_URL, help="Zen models endpoint")
+    parser.add_argument(
+        "--output",
+        default=str(DEFAULT_OUTPUT),
+        help="Path to the generated live Zen manifest",
+    )
+    parser.add_argument(
+        "--curated",
+        default=str(DEFAULT_CURATED),
+        help="Path to the curated Zen manifest used for metadata enrichment",
+    )
+    return parser.parse_args()
+
+
+def main() -> int:
+    args = parse_args()
+    output_path = Path(args.output)
+    curated_path = Path(args.curated)
+    api_key = get_env("ZEN_API_KEY")
+
+    if not api_key:
+        print("Failed to sync Zen models: ZEN_API_KEY is not set", file=sys.stderr)
+        return 1
+
+    try:
+        curated_models = load_curated_models(curated_path)
+        source = fetch_zen_models(args.url, api_key)
+        document = build_output_document(source, args.url, curated_models)
+        write_output(output_path, document)
+    except (HTTPError, URLError, TimeoutError, ValueError, json.JSONDecodeError) as exc:
+        print(f"Failed to sync Zen models: {exc}", file=sys.stderr)
+        return 1
+
+    print(f"Wrote {len(document['models'])} Zen models to {output_path}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

tests/test_sync_zen_models.py

@@ -0,0 +1,61 @@
+import importlib.util
+from pathlib import Path
+
+
+SCRIPT_PATH = Path(__file__).resolve().parents[1] / "scripts" / "sync_zen_models.py"
+SPEC = importlib.util.spec_from_file_location("sync_zen_models", SCRIPT_PATH)
+assert SPEC is not None and SPEC.loader is not None
+MODULE = importlib.util.module_from_spec(SPEC)
+SPEC.loader.exec_module(MODULE)
+
+
+def test_convert_model_applies_family_defaults_for_gpt_5_4():
+    converted = MODULE.convert_model({"id": "gpt-5.4"}, curated_models={})
+
+    assert converted is not None
+    assert converted["model_name"] == "gpt-5.4"
+    assert converted["context_window"] == 1050000
+    assert converted["max_output_tokens"] == 128000
+    assert converted["supports_extended_thinking"] is True
+    assert converted["supports_images"] is True
+    assert converted["use_openai_response_api"] is True
+
+
+def test_convert_model_prefers_curated_metadata_when_available():
+    converted = MODULE.convert_model(
+        {"id": "claude-opus-4-5"},
+        curated_models={
+            "claude-opus-4-5": {
+                "context_window": 200000,
+                "max_output_tokens": 64000,
+                "supports_extended_thinking": False,
+                "description": "Curated Opus 4.5",
+                "intelligence_score": 18,
+                "allow_code_generation": True,
+            }
+        },
+    )
+
+    assert converted is not None
+    assert converted["model_name"] == "claude-opus-4-5"
+    assert converted["context_window"] == 200000
+    assert converted["max_output_tokens"] == 64000
+    assert converted["supports_extended_thinking"] is False
+    assert converted["description"] == "Curated Opus 4.5"
+    assert converted["allow_code_generation"] is True
+
+
+def test_build_output_document_sorts_model_ids():
+    document = MODULE.build_output_document(
+        {
+            "data": [
+                {"id": "gpt-5.4-pro"},
+                {"id": "claude-opus-4-6"},
+            ]
+        },
+        "https://opencode.ai/zen/v1/models",
+        curated_models={},
+    )
+
+    model_names = [item["model_name"] for item in document["models"]]
+    assert model_names == sorted(model_names)

tests/test_zen_model_registry.py

@@ -3,6 +3,7 @@
 import json
 import os
 import tempfile
+from pathlib import Path
 from unittest.mock import patch
 
 from providers.registries.zen import ZenModelRegistry
@@ -43,14 +44,19 @@ class TestZenModelRegistry:
             json.dump(config_data, f)
             temp_path = f.name
 
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".json", delete=False) as live_file:
+            json.dump({"models": []}, live_file)
+            live_path = live_file.name
+
         try:
-            registry = ZenModelRegistry(config_path=temp_path)
+            registry = ZenModelRegistry(config_path=temp_path, live_config_path=live_path)
             assert len(registry.list_models()) == 1
             assert "test/zen-model-1" in registry.list_models()
             assert "zen-test1" in registry.list_aliases()
             assert "zt1" in registry.list_aliases()
         finally:
             os.unlink(temp_path)
+            os.unlink(live_path)
 
     def test_get_capabilities(self):
         """Test capability retrieval."""
@@ -158,9 +164,172 @@ class TestZenModelRegistry:
             json.dump(empty_config, f)
             temp_path = f.name
 
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".json", delete=False) as live_file:
+            json.dump({"models": []}, live_file)
+            live_path = live_file.name
+
         try:
-            registry = ZenModelRegistry(config_path=temp_path)
+            registry = ZenModelRegistry(config_path=temp_path, live_config_path=live_path)
             assert len(registry.list_models()) == 0
             assert len(registry.list_aliases()) == 0
         finally:
             os.unlink(temp_path)
+            os.unlink(live_path)
+
+    def test_live_catalogue_adds_unsynced_model_ids(self):
+        curated_data = {
+            "models": [
+                {
+                    "model_name": "gpt-5.1",
+                    "aliases": ["zen-gpt5.1"],
+                    "context_window": 400000,
+                    "max_output_tokens": 64000,
+                    "intelligence_score": 16,
+                }
+            ]
+        }
+        live_data = {
+            "models": [
+                {
+                    "model_name": "gpt-5.4",
+                    "aliases": [],
+                    "context_window": 1050000,
+                    "max_output_tokens": 128000,
+                    "supports_extended_thinking": True,
+                    "supports_json_mode": True,
+                    "supports_function_calling": True,
+                    "supports_images": True,
+                    "max_image_size_mb": 20.0,
+                    "supports_temperature": True,
+                    "temperature_constraint": "range",
+                    "description": "Live-only GPT-5.4 entry",
+                    "use_openai_response_api": True,
+                }
+            ]
+        }
+
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".json", delete=False) as curated_file:
+            json.dump(curated_data, curated_file)
+            curated_path = curated_file.name
+
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".json", delete=False) as live_file:
+            json.dump(live_data, live_file)
+            live_path = live_file.name
+
+        try:
+            registry = ZenModelRegistry(config_path=curated_path, live_config_path=live_path)
+            assert "gpt-5.4" in registry.list_models()
+            caps = registry.resolve("gpt-5.4")
+            assert caps is not None
+            assert caps.description == "Live-only GPT-5.4 entry"
+            assert caps.use_openai_response_api is True
+        finally:
+            os.unlink(curated_path)
+            os.unlink(live_path)
+
+    def test_curated_manifest_overrides_live_metadata(self):
+        curated_data = {
+            "models": [
+                {
+                    "model_name": "gpt-5.4",
+                    "aliases": ["zen-gpt5.4"],
+                    "context_window": 1050000,
+                    "max_output_tokens": 128000,
+                    "supports_extended_thinking": True,
+                    "supports_json_mode": True,
+                    "supports_function_calling": True,
+                    "supports_images": True,
+                    "max_image_size_mb": 20.0,
+                    "supports_temperature": False,
+                    "temperature_constraint": "fixed",
+                    "description": "Curated override",
+                    "intelligence_score": 19,
+                    "allow_code_generation": True,
+                    "use_openai_response_api": True,
+                }
+            ]
+        }
+        live_data = {
+            "models": [
+                {
+                    "model_name": "gpt-5.4",
+                    "aliases": [],
+                    "context_window": 400000,
+                    "max_output_tokens": 64000,
+                    "supports_extended_thinking": False,
+                    "supports_json_mode": True,
+                    "supports_function_calling": True,
+                    "supports_images": False,
+                    "max_image_size_mb": 0.0,
+                    "supports_temperature": True,
+                    "temperature_constraint": "range",
+                    "description": "Live baseline",
+                    "intelligence_score": 10,
+                    "allow_code_generation": False,
+                }
+            ]
+        }
+
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".json", delete=False) as curated_file:
+            json.dump(curated_data, curated_file)
+            curated_path = curated_file.name
+
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".json", delete=False) as live_file:
+            json.dump(live_data, live_file)
+            live_path = live_file.name
+
+        try:
+            registry = ZenModelRegistry(config_path=curated_path, live_config_path=live_path)
+            caps = registry.resolve("zen-gpt5.4")
+            assert caps is not None
+            assert caps.model_name == "gpt-5.4"
+            assert caps.description == "Curated override"
+            assert caps.context_window == 1050000
+            assert caps.max_output_tokens == 128000
+            assert caps.supports_images is True
+            assert caps.supports_temperature is False
+            assert caps.allow_code_generation is True
+            assert caps.use_openai_response_api is True
+        finally:
+            os.unlink(curated_path)
+            os.unlink(live_path)
+
+    def test_missing_live_catalogue_keeps_curated_models_working(self, monkeypatch):
+        missing_live_path = Path(tempfile.gettempdir()) / "pal-missing-zen-live.json"
+        if missing_live_path.exists():
+            missing_live_path.unlink()
+
+        monkeypatch.setenv("ZEN_LIVE_MODELS_CONFIG_PATH", str(missing_live_path))
+
+        registry = ZenModelRegistry()
+        assert "gpt-5.1" in registry.list_models()
+        assert registry.resolve("zen-gpt5.1") is not None
+
+    def test_invalid_live_json_keeps_curated_models_working(self):
+        curated_data = {
+            "models": [
+                {
+                    "model_name": "gpt-5.1",
+                    "aliases": ["zen-gpt5.1"],
+                    "context_window": 400000,
+                    "max_output_tokens": 64000,
+                    "intelligence_score": 16,
+                }
+            ]
+        }
+
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".json", delete=False) as curated_file:
+            json.dump(curated_data, curated_file)
+            curated_path = curated_file.name
+
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".json", delete=False) as live_file:
+            live_file.write("{ invalid json }")
+            live_path = live_file.name
+
+        try:
+            registry = ZenModelRegistry(config_path=curated_path, live_config_path=live_path)
+            assert "gpt-5.1" in registry.list_models()
+            assert registry.resolve("zen-gpt5.1") is not None
+        finally:
+            os.unlink(curated_path)
+            os.unlink(live_path)

tests/test_zen_provider.py

@@ -288,57 +288,52 @@ class TestZenAPIMocking:
         mock_response.usage.prompt_tokens = 10
         mock_response.usage.completion_tokens = 20
 
-        with patch.object(provider.client.chat.completions, "create", return_value=mock_response):
+        with patch.object(provider.client.chat.completions, "create", return_value=mock_response) as create_mock:
-            # Test the completion method - this will initialize the client
+            response = provider.generate_content(
-            response = provider.complete(
+                prompt="Hello",
-                model="claude-sonnet-4-5", messages=[{"role": "user", "content": "Hello"}], temperature=0.7
+                model_name="claude-sonnet-4-5",
+                temperature=0.7,
             )
 
             assert response.content == "Mocked response from Zen"
+            call_kwargs = create_mock.call_args.kwargs
+            assert call_kwargs["model"] == "claude-sonnet-4-5"
+            assert call_kwargs["messages"] == [{"role": "user", "content": "Hello"}]
+            assert call_kwargs["temperature"] == 0.7
+            assert call_kwargs["stream"] is False
 
-    def test_streaming_completion_mock(self):
+    def test_generate_content_resolves_aliases(self):
-        """Test streaming completion with mocked API."""
+        """Test generate_content resolves Zen aliases before calling the API."""
         provider = ZenProvider(api_key="test-key")
 
-        # Mock streaming response
+        mock_response = Mock()
-        mock_chunk1 = Mock()
+        mock_response.output_text = "Alias response"
-        mock_chunk1.choices = [Mock()]
+        mock_response.model = "gpt-5.1-codex"
-        mock_chunk1.choices[0].delta.content = "Hello"
+        mock_response.id = "resp_123"
-        mock_chunk1.choices[0].finish_reason = None
+        mock_response.created_at = 1234567890
+        mock_response.usage = Mock()
+        mock_response.usage.prompt_tokens = 11
+        mock_response.usage.completion_tokens = 7
+        mock_response.usage.total_tokens = 18
 
-        mock_chunk2 = Mock()
+        with patch.object(provider.client.responses, "create", return_value=mock_response) as create_mock:
-        mock_chunk2.choices = [Mock()]
+            response = provider.generate_content(
-        mock_chunk2.choices[0].delta.content = " world!"
+                prompt="Say hello",
-        mock_chunk2.choices[0].finish_reason = "stop"
+                model_name="zen-codex",
-
-        mock_stream = [mock_chunk1, mock_chunk2]
-
-        # Access client to initialize it first
-        _ = provider.client
-        with patch.object(provider.client.chat.completions, "create", return_value=mock_stream):
-            # Test streaming completion
-            stream = provider.complete_stream(
-                model="gpt-5.1-codex",
-                messages=[{"role": "user", "content": "Say hello"}],
             )
 
-            chunks = list(stream)
+            assert response.content == "Alias response"
-            assert len(chunks) == 2
+            call_kwargs = create_mock.call_args.kwargs
-            assert chunks[0].content == "Hello"
+            assert call_kwargs["model"] == "gpt-5.1-codex"
-            assert chunks[1].content == " world!"
+            assert call_kwargs["input"] == [{"role": "user", "content": [{"type": "input_text", "text": "Say hello"}]}]
 
     def test_api_error_handling(self):
         """Test error handling for API failures."""
         provider = ZenProvider(api_key="test-key")
 
-        # Mock API error
+        with patch.object(provider.client.chat.completions, "create", side_effect=RuntimeError("Mock API error")):
-        from openai import APIError
+            with pytest.raises(RuntimeError, match="OpenCode Zen API error for model claude-sonnet-4-5"):
-
+                provider.generate_content(prompt="Test", model_name="claude-sonnet-4-5")
-        api_error = APIError("Mock API error", request=Mock(), body="error details")
-
-        with patch.object(provider._client.chat.completions, "create", side_effect=api_error):
-            with pytest.raises(APIError):
-                provider.complete(model="claude-sonnet-4-5", messages=[{"role": "user", "content": "Test"}])
 
     def test_invalid_model_error(self):
         """Test error handling for invalid models."""
@@ -351,11 +346,6 @@ class TestZenAPIMocking:
         """Test handling of authentication errors."""
         provider = ZenProvider(api_key="invalid-key")
 
-        # Mock authentication error
+        with patch.object(provider.client.chat.completions, "create", side_effect=RuntimeError("auth failed")):
-        from openai import AuthenticationError
+            with pytest.raises(RuntimeError, match="OpenCode Zen API error for model claude-sonnet-4-5"):
-
+                provider.generate_content(prompt="Test", model_name="claude-sonnet-4-5")
-        auth_error = AuthenticationError("Invalid API key", request=Mock(), body="auth failed")
-
-        with patch.object(provider._client.chat.completions, "create", side_effect=auth_error):
-            with pytest.raises(AuthenticationError):
-                provider.complete(model="claude-sonnet-4-5", messages=[{"role": "user", "content": "Test"}])