@ailib-official/ai-protocol 0.8.4

package/schemas/v2/availability.json

@@ -0,0 +1,66 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://raw.githubusercontent.com/ailib-official/ai-protocol/main/schemas/v2/availability.json",
+  "title": "Availability and Health Check",
+  "description": "Provider availability configuration and health check definition",
+  "type": "object",
+  "required": ["required", "regions", "check"],
+  "properties": {
+    "required": {
+      "type": "boolean",
+      "description": "Whether the provider must be reachable at runtime startup"
+    },
+    "regions": {
+      "$ref": "./regions.json"
+    },
+    "interval_ms": {
+      "type": "integer",
+      "minimum": 1000,
+      "description": "Suggested health check interval in milliseconds (runtime may override)"
+    },
+    "fallback": {
+      "type": "object",
+      "description": "Optional fallback check when primary check fails (e.g., POST ping)",
+      "properties": {
+        "method": { "type": "string", "enum": ["GET", "POST"] },
+        "path": { "type": "string", "pattern": "^/" },
+        "expected_status": { "type": "array", "items": { "type": "integer" } }
+      }
+    },
+    "check": {
+      "type": "object",
+      "required": ["method", "path", "expected_status"],
+      "properties": {
+        "method": {
+          "type": "string",
+          "enum": ["HEAD", "GET"],
+          "description": "HTTP method for health check"
+        },
+        "path": {
+          "type": "string",
+          "pattern": "^/",
+          "description": "Path relative to base_url for health check"
+        },
+        "expected_status": {
+          "type": "array",
+          "minItems": 1,
+          "items": {
+            "type": "integer",
+            "minimum": 100,
+            "maximum": 599
+          },
+          "description": "HTTP status codes that indicate healthy state"
+        },
+        "timeout_ms": {
+          "type": "integer",
+          "minimum": 100,
+          "default": 3000,
+          "description": "Timeout for health check in milliseconds"
+        }
+      },
+      "additionalProperties": false
+    }
+  },
+  "additionalProperties": false
+}
+

package/schemas/v2/capabilities.json

@@ -0,0 +1,182 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://raw.githubusercontent.com/ailib-official/ai-protocol/main/schemas/v2/capabilities.json",
+  "title": "AI-Protocol Capability Declaration",
+  "description": "Structured capability declaration for V2 provider manifests. Defines required and optional capabilities that a provider supports, enabling capability-based loading in runtimes.",
+  "type": "object",
+  "oneOf": [
+    {
+      "description": "V2 structured format with required/optional separation",
+      "type": "object",
+      "required": ["required"],
+      "properties": {
+        "required": {
+          "type": "array",
+          "description": "Capabilities that MUST be available for this provider to function",
+          "items": { "$ref": "#/$defs/capability_name" },
+          "minItems": 1,
+          "uniqueItems": true
+        },
+        "optional": {
+          "type": "array",
+          "description": "Capabilities that MAY be available and can be loaded on demand",
+          "items": { "$ref": "#/$defs/capability_name" },
+          "uniqueItems": true,
+          "default": []
+        },
+        "feature_flags": {
+          "type": "object",
+          "description": "Fine-grained feature toggles within capabilities",
+          "properties": {
+            "structured_output": {
+              "type": "boolean",
+              "default": false,
+              "description": "Supports JSON mode / schema-constrained output"
+            },
+            "parallel_tool_calls": {
+              "type": "boolean",
+              "default": false,
+              "description": "Supports parallel tool invocations in a single response"
+            },
+            "extended_thinking": {
+              "type": "boolean",
+              "default": false,
+              "description": "Supports extended reasoning / thinking blocks"
+            },
+            "streaming_usage": {
+              "type": "boolean",
+              "default": false,
+              "description": "Reports token usage during streaming"
+            },
+            "system_messages": {
+              "type": "boolean",
+              "default": true,
+              "description": "Supports system role messages"
+            },
+            "image_generation": {
+              "type": "boolean",
+              "default": false,
+              "description": "Supports image generation"
+            },
+            "json_mode": {
+              "type": "boolean",
+              "default": false,
+              "description": "Supports JSON mode for constrained output generation"
+            },
+            "json_schema_mode": {
+              "type": "boolean",
+              "default": false,
+              "description": "Supports JSON schema-constrained output generation"
+            },
+            "reasoning_tokens": {
+              "type": "boolean",
+              "default": false,
+              "description": "Reports reasoning/thinking tokens separately in usage metadata"
+            },
+            "streaming_tool_calls": {
+              "type": "boolean",
+              "default": false,
+              "description": "Supports tool call streaming with incremental argument chunks"
+            },
+            "recursive_tool_calls": {
+              "type": "boolean",
+              "default": false,
+              "description": "Supports multi-turn recursive tool calling without client re-prompting"
+            }
+          },
+          "additionalProperties": {
+            "type": "boolean"
+          }
+        }
+      },
+      "additionalProperties": false
+    },
+    {
+      "description": "V1 legacy format - simple boolean flags (backward compatible)",
+      "type": "object",
+      "properties": {
+        "streaming": {
+          "type": "boolean",
+          "description": "Supports streaming responses"
+        },
+        "tools": {
+          "type": "boolean",
+          "description": "Supports tool / function calling"
+        },
+        "vision": {
+          "type": "boolean",
+          "description": "Supports image or multimodal inputs"
+        },
+        "agentic": {
+          "type": "boolean",
+          "default": false,
+          "description": "Supports agentic reasoning and multi-turn tool sequences"
+        },
+        "reasoning": {
+          "type": "boolean",
+          "default": false,
+          "description": "Supports extended reasoning / thinking blocks"
+        },
+        "parallel_tools": {
+          "type": "boolean",
+          "default": false,
+          "description": "Supports parallel tool invocations"
+        }
+      },
+      "additionalProperties": false
+    }
+  ],
+  "$defs": {
+    "capability_name": {
+      "type": "string",
+      "description": "Standard capability identifier",
+      "enum": [
+        "text",
+        "streaming",
+        "vision",
+        "audio",
+        "video",
+        "tools",
+        "parallel_tools",
+        "agentic",
+        "reasoning",
+        "embeddings",
+        "structured_output",
+        "batch",
+        "image_generation",
+        "computer_use",
+        "mcp_client",
+        "mcp_server",
+        "stt",
+        "tts",
+        "rerank"
+      ]
+    },
+    "capability_module_mapping": {
+      "type": "object",
+      "description": "Reference: Maps capability names to runtime modules",
+      "properties": {
+        "text": { "const": "Core (always loaded)" },
+        "streaming": { "const": "streaming.*" },
+        "vision": { "const": "multimodal.vision" },
+        "audio": { "const": "multimodal.audio" },
+        "video": { "const": "multimodal.video" },
+        "tools": { "const": "tools.*" },
+        "parallel_tools": { "const": "tools.parallel" },
+        "agentic": { "const": "agentic.*" },
+        "reasoning": { "const": "reasoning.*" },
+        "embeddings": { "const": "embeddings.*" },
+        "structured_output": { "const": "structured.*" },
+        "batch": { "const": "batch.*" },
+        "image_generation": { "const": "generation.image" },
+        "computer_use": { "const": "computer_use.*" },
+        "mcp_client": { "const": "mcp.client" },
+        "mcp_server": { "const": "mcp.server" },
+        "stt": { "const": "stt.*" },
+        "tts": { "const": "tts.*" },
+        "rerank": { "const": "rerank.*" }
+      }
+    }
+  }
+}
+

package/schemas/v2/capability-profile.json

@@ -0,0 +1,174 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://raw.githubusercontent.com/ailib-official/ai-protocol/main/schemas/v2/capability-profile.json",
+  "title": "AI-Protocol Capability Profile (Staged IOS/IOSPC Phase)",
+  "description": "Optional staged capability profile for governance. ios_v1 limits scope to Inputs/Outcomes/Systems. iospc_v1 introduces Process/Contract in report-first mode. Metadata-only and MUST NOT imply runtime behavioral changes by itself.",
+  "type": "object",
+  "additionalProperties": false,
+  "required": ["phase"],
+  "allOf": [
+    {
+      "if": { "properties": { "phase": { "const": "ios_v1" } }, "required": ["phase"] },
+      "then": {
+        "anyOf": [
+          { "required": ["inputs"] },
+          { "required": ["outcomes"] },
+          { "required": ["systems"] }
+        ],
+        "not": {
+          "anyOf": [
+            { "required": ["process"] },
+            { "required": ["contract"] }
+          ]
+        }
+      }
+    },
+    {
+      "if": { "properties": { "phase": { "const": "iospc_v1" } }, "required": ["phase"] },
+      "then": {
+        "allOf": [
+          {
+            "anyOf": [
+              { "required": ["inputs"] },
+              { "required": ["outcomes"] },
+              { "required": ["systems"] }
+            ]
+          },
+          {
+            "anyOf": [
+              { "required": ["process"] },
+              { "required": ["contract"] }
+            ]
+          }
+        ]
+      }
+    }
+  ],
+  "properties": {
+    "phase": {
+      "type": "string",
+      "enum": ["ios_v1", "iospc_v1"],
+      "default": "ios_v1",
+      "description": "Staged rollout marker. ios_v1 enables Inputs/Outcomes/Systems only. iospc_v1 allows Process/Contract in report-first mode."
+    },
+    "inputs": {
+      "type": "object",
+      "description": "Input constraints and limits (I)",
+      "additionalProperties": false,
+      "properties": {
+        "modalities": {
+          "type": "array",
+          "items": {
+            "type": "string",
+            "enum": ["text", "image", "audio", "video", "tool_result", "other"]
+          },
+          "uniqueItems": true
+        },
+        "max_references": {
+          "type": "integer",
+          "minimum": 0,
+          "maximum": 64
+        },
+        "max_payload_mb": {
+          "type": "number",
+          "minimum": 0,
+          "maximum": 2048
+        }
+      }
+    },
+    "outcomes": {
+      "type": "object",
+      "description": "Semantic output declaration (O)",
+      "additionalProperties": false,
+      "properties": {
+        "types": {
+          "type": "array",
+          "items": {
+            "type": "string",
+            "enum": [
+              "text_completion",
+              "structured_json",
+              "image_generation",
+              "video_generation",
+              "reasoning_trace",
+              "tool_call_sequence",
+              "other"
+            ]
+          },
+          "uniqueItems": true
+        },
+        "quality_tier": {
+          "type": "string",
+          "enum": ["best_effort", "standard", "high"]
+        }
+      }
+    },
+    "systems": {
+      "type": "object",
+      "description": "External system dependencies (S)",
+      "additionalProperties": false,
+      "properties": {
+        "requires": {
+          "type": "array",
+          "items": {
+            "type": "string",
+            "enum": ["mcp", "search", "filesystem", "browser", "gis", "other"]
+          },
+          "uniqueItems": true
+        }
+      }
+    },
+    "process": {
+      "type": "object",
+      "description": "Execution process declaration (P) for iospc_v1 only.",
+      "additionalProperties": false,
+      "properties": {
+        "mode": {
+          "type": "string",
+          "enum": ["sync", "async", "hybrid"]
+        },
+        "supports_callbacks": {
+          "type": "boolean"
+        },
+        "max_async_jobs": {
+          "type": "integer",
+          "minimum": 1,
+          "maximum": 100000
+        }
+      }
+    },
+    "contract": {
+      "type": "object",
+      "description": "Execution contract declaration (C) for iospc_v1 only.",
+      "additionalProperties": false,
+      "properties": {
+        "safety_profile": {
+          "type": "string",
+          "enum": ["baseline", "standard", "strict"]
+        },
+        "cost_class": {
+          "type": "string",
+          "enum": ["low", "medium", "high", "premium"]
+        },
+        "error_model_ref": {
+          "type": "string",
+          "minLength": 1
+        },
+        "sla_tier": {
+          "type": "string",
+          "enum": ["best_effort", "standard", "enterprise"]
+        }
+      }
+    },
+    "future_reserved": {
+      "type": "array",
+      "description": "Reserved names for future staged expansion.",
+      "items": {
+        "type": "string",
+        "enum": ["process", "contract"]
+      },
+      "default": ["process", "contract"],
+      "uniqueItems": true
+    }
+  }
+}

package/schemas/v2/computer-use.json

@@ -0,0 +1,222 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://raw.githubusercontent.com/ailib-official/ai-protocol/main/schemas/v2/computer-use.json",
+  "title": "AI-Protocol V2 Computer Use",
+  "description": "Computer Use capability configuration for V2 provider manifests. Standardizes how providers expose GUI automation and desktop interaction capabilities. Validated against Anthropic (screen_based, Beta), Google (tool_based, GA), and OpenAI (screen_based, Preview) implementations.",
+  "type": "object",
+  "properties": {
+    "supported": {
+      "type": "boolean",
+      "default": false,
+      "description": "Whether the provider supports computer use / GUI automation"
+    },
+    "status": {
+      "type": "string",
+      "enum": ["ga", "beta", "preview", "experimental"],
+      "description": "Maturity status of computer use support"
+    },
+    "implementation": {
+      "type": "string",
+      "enum": ["screen_based", "tool_based", "hybrid"],
+      "description": "Implementation approach: screen_based (screenshot loop with coordinates), tool_based (structured semantic actions), hybrid (both approaches available)"
+    },
+    "actions": {
+      "type": "object",
+      "description": "Supported computer use actions —normalized across all implementations",
+      "properties": {
+        "screenshot": {
+          "type": "object",
+          "properties": {
+            "supported": { "type": "boolean", "default": true },
+            "formats": {
+              "type": "array",
+              "items": { "type": "string", "enum": ["png", "jpeg", "webp"] },
+              "minItems": 1,
+              "description": "Supported screenshot output formats"
+            },
+            "max_resolution": {
+              "type": "string",
+              "pattern": "^\\d+x\\d+$",
+              "description": "Maximum screenshot resolution (e.g. '4096x4096')"
+            }
+          },
+          "additionalProperties": false
+        },
+        "mouse": {
+          "type": "object",
+          "properties": {
+            "supported": { "type": "boolean", "default": true },
+            "operations": {
+              "type": "array",
+              "items": {
+                "type": "string",
+                "enum": ["click", "double_click", "right_click", "drag", "move", "scroll"]
+              },
+              "minItems": 1,
+              "uniqueItems": true
+            }
+          },
+          "additionalProperties": false
+        },
+        "keyboard": {
+          "type": "object",
+          "properties": {
+            "supported": { "type": "boolean", "default": true },
+            "operations": {
+              "type": "array",
+              "items": {
+                "type": "string",
+                "enum": ["type", "key_press", "key_release", "shortcut"]
+              },
+              "minItems": 1,
+              "uniqueItems": true
+            }
+          },
+          "additionalProperties": false
+        },
+        "browser": {
+          "type": "object",
+          "properties": {
+            "supported": { "type": "boolean", "default": false },
+            "operations": {
+              "type": "array",
+              "items": {
+                "type": "string",
+                "enum": ["navigate", "fill_form", "click_element", "scroll", "back", "forward", "refresh"]
+              },
+              "uniqueItems": true
+            }
+          },
+          "additionalProperties": false
+        },
+        "zoom": {
+          "type": "object",
+          "properties": {
+            "supported": { "type": "boolean", "default": false },
+            "description": { "type": "string" }
+          },
+          "description": "Screen region zoom for detailed inspection (Anthropic computer_20251124)",
+          "additionalProperties": false
+        },
+        "file": {
+          "type": "object",
+          "properties": {
+            "supported": { "type": "boolean", "default": false },
+            "operations": {
+              "type": "array",
+              "items": {
+                "type": "string",
+                "enum": ["read", "write", "list", "delete", "move"]
+              },
+              "uniqueItems": true
+            }
+          },
+          "description": "File system operations (primarily Google tool_based implementation)",
+          "additionalProperties": false
+        }
+      },
+      "additionalProperties": false
+    },
+    "safety": {
+      "type": "object",
+      "description": "Safety configuration for computer use —mandatory safeguards",
+      "properties": {
+        "confirmation_required": {
+          "type": "boolean",
+          "default": true,
+          "description": "Require human confirmation for consequential actions"
+        },
+        "sandbox_mode": {
+          "type": "string",
+          "enum": ["required", "recommended", "optional"],
+          "default": "recommended",
+          "description": "Whether sandbox/VM environment is required"
+        },
+        "action_logging": {
+          "type": "boolean",
+          "default": true,
+          "description": "Whether all actions are logged for audit trail"
+        },
+        "domain_allowlist": {
+          "type": "boolean",
+          "default": false,
+          "description": "Whether domain allowlisting is supported for web browsing"
+        },
+        "sensitive_data_protection": {
+          "type": "boolean",
+          "default": true,
+          "description": "Whether the system avoids accessing sensitive data (passwords, PII)"
+        },
+        "max_actions_per_turn": {
+          "type": "integer",
+          "minimum": 1,
+          "description": "Maximum number of actions allowed per model turn"
+        },
+        "action_timeout_ms": {
+          "type": "integer",
+          "minimum": 1000,
+          "default": 30000,
+          "description": "Timeout for individual action execution"
+        }
+      },
+      "additionalProperties": false
+    },
+    "environment": {
+      "type": "object",
+      "description": "Supported execution environments",
+      "properties": {
+        "browser": {
+          "type": "boolean",
+          "default": true,
+          "description": "Browser-based environment (Playwright, Selenium, Puppeteer)"
+        },
+        "desktop": {
+          "type": "boolean",
+          "default": false,
+          "description": "Full desktop environment (VM, Docker, native OS)"
+        },
+        "display_config": {
+          "type": "object",
+          "properties": {
+            "width": { "type": "integer", "minimum": 320, "description": "Display width in pixels" },
+            "height": { "type": "integer", "minimum": 240, "description": "Display height in pixels" }
+          },
+          "additionalProperties": false
+        }
+      },
+      "additionalProperties": false
+    },
+    "provider_mapping": {
+      "type": "object",
+      "description": "Provider-specific API configuration for computer use",
+      "properties": {
+        "tool_type": {
+          "type": "string",
+          "description": "Provider's tool type identifier (e.g. 'computer_20251124', 'computer_use_preview', 'computer_use')"
+        },
+        "beta_header": {
+          "type": "string",
+          "description": "Required beta header for the feature"
+        },
+        "api_endpoint": {
+          "type": "string",
+          "description": "Specific API endpoint for computer use (if different from main chat)"
+        },
+        "model_requirement": {
+          "type": "string",
+          "description": "Specific model required for computer use (e.g. 'computer-use-preview')"
+        },
+        "config_method": {
+          "type": "string",
+          "enum": ["tool_parameter", "sdk_config", "api_header"],
+          "description": "How computer use is activated in the provider API"
+        }
+      },
+      "additionalProperties": true
+    }
+  },
+  "if": { "properties": { "supported": { "const": true } } },
+  "then": { "required": ["implementation", "actions", "safety"] },
+  "additionalProperties": false
+}
+