@ailib-official/ai-protocol 0.8.4 → 1.0.0

package/v2/providers/doubao.yaml

@@ -1,4 +1,5 @@
 # Doubao V2 provider manifest — P0 multimodal generative coverage
+# Last Updated: 2026-06-26 (add tool_calling)
 $schema: "https://raw.githubusercontent.com/ailib-official/ai-protocol/main/schemas/v2/provider.json"
 
 id: doubao
@@ -29,6 +30,16 @@ error_classification:
     "500": "server_error"
     "503": "overloaded"
 
+availability:
+  required: false
+  regions:
+    - cn
+  check:
+    method: GET
+    path: "/models"
+    expected_status: [200, 401]
+    timeout_ms: 3000
+
 capabilities:
   required:
     - text
@@ -50,6 +61,13 @@ capabilities:
     system_messages: true
     image_generation: true
 
+  tool_calling:
+    native:
+      supported: true
+      reliability: "full"
+      parallel: true
+      streaming: true
+
 capability_profile:
   phase: "ios_v1"
   inputs:
@@ -65,14 +83,6 @@ parameters:
   top_p: { type: float, range: [0.0, 1.0] }
   stream: { type: boolean }
 
-parameter_mappings:
-  temperature: "temperature"
-  max_tokens: "max_tokens"
-  stream: "stream"
-  top_p: "top_p"
-  tools: "tools"
-  tool_choice: "tool_choice"
-
 streaming:
   decoder:
     format: "sse"

package/v2/providers/google.yaml

@@ -1,7 +1,6 @@
-# Google Gemini V2 正式提供商清单 — 全模态支持（音频/视频/CU），合并 v2-alpha 与深度调研数据
-# AI-Protocol V2 Provider Manifest
-# Provider: Google | Models: Gemini 3 Pro / Flash
-# Last Updated: 2026-02-16
+# Google Gemini V2 Provider Manifest — 全模态（音频/视频/CU）
+# Provider: Google | Models: Gemini 3 Pro / Flash / 3.1 Flash
+# Last Updated: 2026-06-26 (add tool_calling)
 $schema: "https://raw.githubusercontent.com/ailib-official/ai-protocol/main/schemas/v2/provider.json"
 
 # === Ring 1: Core Skeleton ===
@@ -42,6 +41,16 @@ error_classification:
     "UNAVAILABLE": "overloaded"
     "DEADLINE_EXCEEDED": "timeout"
 
+availability:
+  required: false
+  regions:
+    - global
+  check:
+    method: GET
+    path: "/models"
+    expected_status: [200, 401]
+    timeout_ms: 3000
+
 # === Ring 2: Capability Mapping ===
 capabilities:
   required:
@@ -65,6 +74,13 @@ capabilities:
     streaming_usage: false
     system_messages: true
 
+  tool_calling:
+    native:
+      supported: true
+      reliability: "full"
+      parallel: true
+      streaming: true
+
 capability_profile:
   phase: "ios_v1"
   inputs:
@@ -236,6 +252,18 @@ metadata:
       max_output_tokens: 64000
       status: public_preview
       pricing: { input_per_1m: 0.15, output_per_1m: 0.60 }
+    gemini-2.5-flash-lite:
+      context_window: 1048576
+      max_output_tokens: 65536
+    gemini-2.5-flash:
+      context_window: 1048576
+      max_output_tokens: 65536
+    gemini-2.5-pro:
+      context_window: 2097152
+      max_output_tokens: 65536
+    gemini-3.1-flash-lite-preview:
+      context_window: 1048576
+      max_output_tokens: 65536
   context_caching:
     supported: true
     pricing: "$0.20-0.40 per 1M tokens/hour"

package/v2/providers/groq.yaml

@@ -0,0 +1,159 @@
+# Groq V2 provider — OpenAI-compatible fast inference
+# Last Updated: 2026-06-26 (add tool_calling)
+$schema: "https://raw.githubusercontent.com/ailib-official/ai-protocol/main/schemas/v2/provider.json"
+
+id: groq
+protocol_version: "2.0"
+name: "Groq"
+version: "1.0.0"
+status: stable
+category: ai_provider
+official_url: "https://console.groq.com/docs"
+support_contact: "https://support.groq.com"
+
+endpoint:
+  base_url: "https://api.groq.com/openai/v1"
+  chat: "/chat/completions"
+  auth:
+    type: "bearer"
+    header: "Authorization"
+    prefix: "Bearer"
+    token_env: "GROQ_API_KEY"
+
+error_classification:
+  by_http_status:
+    "400": "invalid_request"
+    "401": "authentication"
+    "403": "permission_denied"
+    "404": "not_found"
+    "429": "rate_limited"
+    "500": "server_error"
+    "503": "overloaded"
+  by_error_code:
+    "context_length_exceeded": "request_too_large"
+    "model_not_found": "not_found"
+    "rate_limit_exceeded": "rate_limited"
+
+availability:
+  required: false
+  regions:
+    - global
+  check:
+    method: GET
+    path: "/models"
+    expected_status: [200, 401]
+    timeout_ms: 3000
+
+capabilities:
+  required:
+    - text
+    - streaming
+    - tools
+  optional:
+    - parallel_tools
+    - agentic
+  feature_flags:
+    parallel_tool_calls: true
+    streaming_usage: true
+    system_messages: true
+
+  tool_calling:
+    native:
+      supported: true
+      reliability: "full"
+      parallel: true
+
+parameters:
+  temperature: { type: float, range: [0.0, 2.0], default: 1.0 }
+  max_tokens: { type: integer, min: 1, max: 8192 }
+  top_p: { type: float, range: [0.0, 1.0] }
+  stream: { type: boolean }
+
+streaming:
+  decoder:
+    format: "sse"
+    strategy: "openai_chat"
+    done_signal: "[DONE]"
+    prefix: "data: "
+  content_path: "$.choices[0].delta.content"
+  tool_call_path: "$.choices[0].delta.tool_calls"
+  usage_path: "$.usage"
+  event_map:
+    - match: "exists($.choices[*].delta.content)"
+      emit: "PartialContentDelta"
+      fields:
+        content: "$.choices[*].delta.content"
+    - match: "exists($.choices[*].delta.tool_calls)"
+      emit: "PartialToolCall"
+      fields:
+        tool_calls: "$.choices[*].delta.tool_calls"
+    - match: "exists($.usage)"
+      emit: "Metadata"
+      fields:
+        usage: "$.usage"
+    - match: "$.choices[*].finish_reason != null"
+      emit: "StreamEnd"
+      fields:
+        finish_reason: "$.choices[*].finish_reason"
+  stop_condition: "$.choices[0].finish_reason != null"
+
+multimodal:
+  input:
+    vision:
+      supported: false
+    audio:
+      supported: false
+    video:
+      supported: false
+  output:
+    text: true
+    audio:
+      supported: false
+    image:
+      supported: false
+
+api_families: ["chat_completions"]
+default_api_family: "chat_completions"
+
+endpoints:
+  chat:
+    path: "/chat/completions"
+    method: "POST"
+    adapter: "openai"
+
+services:
+  list_models:
+    path: "/models"
+    method: "GET"
+    response_binding: "data"
+
+retry_policy:
+  strategy: "exponential_backoff"
+  max_retries: 3
+  min_delay_ms: 1000
+  max_delay_ms: 30000
+  jitter: "full"
+  retry_on_http_status: [429, 500, 502, 503]
+
+termination:
+  source_field: "finish_reason"
+  mapping:
+    stop: "end_turn"
+    length: "max_tokens"
+    tool_calls: "tool_use"
+
+metadata:
+  api_compatibility: openai
+  models:
+    llama-3.1-8b-instant:
+      context_window: 128000
+      max_output_tokens: 8192
+      verification:
+        status: verified
+        source: official_documentation
+        verified_at: "2026-05-29"
+        notes: "https://console.groq.com/docs/models"
+  sdk:
+    note: "OpenAI-compatible — use openai SDK with base_url override"
+    python: openai
+    typescript: openai

package/v2/providers/jina.yaml

@@ -32,6 +32,16 @@ error_classification:
     "429": "rate_limited"
     "500": "server_error"
 
+availability:
+  required: false
+  regions:
+    - global
+  check:
+    method: GET
+    path: "/v1/models"
+    expected_status: [200, 401]
+    timeout_ms: 3000
+
 # === Ring 2: Capability Mapping ===
 capabilities:
   required:

package/v2/providers/moonshot.yaml

@@ -1,7 +1,6 @@
-# Moonshot Kimi V2 正式提供商清单 — Agent Swarm 先驱，OpenAI 兼容 API
-# AI-Protocol V2 Provider Manifest
-# Provider: Moonshot AI | Models: Kimi K2.5 / K2-Thinking
-# Last Updated: 2026-02-16
+# Moonshot Kimi V2 Provider Manifest — Agent Swarm 先驱
+# Provider: Moonshot AI | Models: Kimi K2.6 / K2.5 / K2-Thinking
+# Last Updated: 2026-06-26 (add tool_calling)
 $schema: "https://raw.githubusercontent.com/ailib-official/ai-protocol/main/schemas/v2/provider.json"
 
 # === Ring 1: Core Skeleton ===
@@ -38,6 +37,16 @@ error_classification:
     "rate_limit_exceeded": "rate_limited"
     "insufficient_quota": "quota_exhausted"
 
+availability:
+  required: false
+  regions:
+    - cn
+  check:
+    method: GET
+    path: "/models"
+    expected_status: [200, 401]
+    timeout_ms: 3000
+
 # === Ring 2: Capability Mapping ===
 capabilities:
   required:
@@ -57,6 +66,13 @@ capabilities:
     streaming_usage: true
     system_messages: true
 
+  tool_calling:
+    native:
+      supported: true
+      reliability: "full"
+      parallel: true
+      streaming: true
+
 capability_profile:
   phase: "ios_v1"
   inputs:
@@ -72,14 +88,6 @@ parameters:
   top_p: { type: float, range: [0.0, 1.0] }
   stream: { type: boolean }
 
-parameter_mappings:
-  temperature: "temperature"
-  max_tokens: "max_tokens"
-  stream: "stream"
-  top_p: "top_p"
-  tools: "tools"
-  tool_choice: "tool_choice"
-
 streaming:
   decoder:
     format: "sse"