Merge pull request #311 from getcompanion-ai/hari/migrate-chat-lossless-convex

hari/migrate chat lossless convex

packages/ai/src/models.generated.ts

@@ -3010,7 +3010,7 @@ export const MODELS = {
         cacheRead: 0,
         cacheWrite: 0,
       },
-      contextWindow: 128000,
+      contextWindow: 264000,
       maxTokens: 64000,
     } satisfies Model<"openai-responses">,
     "gpt-5.2-codex": {
@@ -3033,7 +3033,7 @@ export const MODELS = {
         cacheRead: 0,
         cacheWrite: 0,
       },
-      contextWindow: 272000,
+      contextWindow: 400000,
       maxTokens: 128000,
     } satisfies Model<"openai-responses">,
     "gpt-5.3-codex": {
@@ -6486,6 +6486,23 @@ export const MODELS = {
       contextWindow: 204800,
       maxTokens: 131072,
     } satisfies Model<"anthropic-messages">,
+    "nemotron-3-super-free": {
+      id: "nemotron-3-super-free",
+      name: "Nemotron 3 Super Free",
+      api: "openai-completions",
+      provider: "opencode",
+      baseUrl: "https://opencode.ai/zen/v1",
+      reasoning: true,
+      input: ["text"],
+      cost: {
+        input: 0,
+        output: 0,
+        cacheRead: 0,
+        cacheWrite: 0,
+      },
+      contextWindow: 1000000,
+      maxTokens: 128000,
+    } satisfies Model<"openai-completions">,
   },
   "opencode-go": {
     "glm-5": {
@@ -7230,13 +7247,13 @@ export const MODELS = {
       reasoning: true,
       input: ["text"],
       cost: {
-        input: 0.25,
+        input: 0.26,
-        output: 0.39999999999999997,
+        output: 0.38,
-        cacheRead: 0,
+        cacheRead: 0.13,
         cacheWrite: 0,
       },
       contextWindow: 163840,
-      maxTokens: 65536,
+      maxTokens: 4096,
     } satisfies Model<"openai-completions">,
     "deepseek/deepseek-v3.2-exp": {
       id: "deepseek/deepseek-v3.2-exp",
@@ -8360,6 +8377,23 @@ export const MODELS = {
       contextWindow: 256000,
       maxTokens: 4096,
     } satisfies Model<"openai-completions">,
+    "nvidia/nemotron-3-super-120b-a12b:free": {
+      id: "nvidia/nemotron-3-super-120b-a12b:free",
+      name: "NVIDIA: Nemotron 3 Super (free)",
+      api: "openai-completions",
+      provider: "openrouter",
+      baseUrl: "https://openrouter.ai/api/v1",
+      reasoning: true,
+      input: ["text"],
+      cost: {
+        input: 0,
+        output: 0,
+        cacheRead: 0,
+        cacheWrite: 0,
+      },
+      contextWindow: 262144,
+      maxTokens: 262144,
+    } satisfies Model<"openai-completions">,
     "nvidia/nemotron-nano-12b-v2-vl:free": {
       id: "nvidia/nemotron-nano-12b-v2-vl:free",
       name: "NVIDIA: Nemotron Nano 12B 2 VL (free)",
@@ -9346,6 +9380,40 @@ export const MODELS = {
       contextWindow: 200000,
       maxTokens: 4096,
     } satisfies Model<"openai-completions">,
+    "openrouter/healer-alpha": {
+      id: "openrouter/healer-alpha",
+      name: "Healer Alpha",
+      api: "openai-completions",
+      provider: "openrouter",
+      baseUrl: "https://openrouter.ai/api/v1",
+      reasoning: true,
+      input: ["text", "image"],
+      cost: {
+        input: 0,
+        output: 0,
+        cacheRead: 0,
+        cacheWrite: 0,
+      },
+      contextWindow: 262144,
+      maxTokens: 65536,
+    } satisfies Model<"openai-completions">,
+    "openrouter/hunter-alpha": {
+      id: "openrouter/hunter-alpha",
+      name: "Hunter Alpha",
+      api: "openai-completions",
+      provider: "openrouter",
+      baseUrl: "https://openrouter.ai/api/v1",
+      reasoning: true,
+      input: ["text"],
+      cost: {
+        input: 0,
+        output: 0,
+        cacheRead: 0,
+        cacheWrite: 0,
+      },
+      contextWindow: 1048576,
+      maxTokens: 65536,
+    } satisfies Model<"openai-completions">,
     "prime-intellect/intellect-3": {
       id: "prime-intellect/intellect-3",
       name: "Prime Intellect: INTELLECT-3",
@@ -9848,13 +9916,13 @@ export const MODELS = {
       reasoning: true,
       input: ["text"],
       cost: {
-        input: 0.15,
+        input: 0.0975,
-        output: 1.2,
+        output: 0.78,
         cacheRead: 0,
         cacheWrite: 0,
       },
-      contextWindow: 128000,
+      contextWindow: 131072,
-      maxTokens: 4096,
+      maxTokens: 32768,
     } satisfies Model<"openai-completions">,
     "qwen/qwen3-vl-235b-a22b-instruct": {
       id: "qwen/qwen3-vl-235b-a22b-instruct",
@@ -9882,8 +9950,8 @@ export const MODELS = {
       reasoning: true,
       input: ["text", "image"],
       cost: {
-        input: 0,
+        input: 0.26,
-        output: 0,
+        output: 2.6,
         cacheRead: 0,
         cacheWrite: 0,
       },
@@ -9916,8 +9984,8 @@ export const MODELS = {
       reasoning: true,
       input: ["text", "image"],
       cost: {
-        input: 0,
+        input: 0.13,
-        output: 0,
+        output: 1.56,
         cacheRead: 0,
         cacheWrite: 0,
       },
@@ -10649,13 +10717,13 @@ export const MODELS = {
       reasoning: true,
       input: ["text"],
       cost: {
-        input: 0.09999999999999999,
+        input: 0.29,
-        output: 0.3,
+        output: 0.59,
         cacheRead: 0,
         cacheWrite: 0,
       },
-      contextWindow: 40960,
+      contextWindow: 131072,
-      maxTokens: 16384,
+      maxTokens: 40960,
     } satisfies Model<"anthropic-messages">,
     "alibaba/qwen3-235b-a22b-thinking": {
       id: "alibaba/qwen3-235b-a22b-thinking",
@@ -10700,13 +10768,13 @@ export const MODELS = {
       reasoning: true,
       input: ["text"],
       cost: {
-        input: 0.07,
+        input: 0.15,
-        output: 0.27,
+        output: 0.6,
         cacheRead: 0,
         cacheWrite: 0,
       },
-      contextWindow: 160000,
+      contextWindow: 262144,
-      maxTokens: 32768,
+      maxTokens: 8192,
     } satisfies Model<"anthropic-messages">,
     "alibaba/qwen3-coder-next": {
       id: "alibaba/qwen3-coder-next",
@@ -10742,6 +10810,23 @@ export const MODELS = {
       contextWindow: 1000000,
       maxTokens: 65536,
     } satisfies Model<"anthropic-messages">,
+    "alibaba/qwen3-max": {
+      id: "alibaba/qwen3-max",
+      name: "Qwen3 Max",
+      api: "anthropic-messages",
+      provider: "vercel-ai-gateway",
+      baseUrl: "https://ai-gateway.vercel.sh",
+      reasoning: false,
+      input: ["text"],
+      cost: {
+        input: 1.2,
+        output: 6,
+        cacheRead: 0.24,
+        cacheWrite: 0,
+      },
+      contextWindow: 262144,
+      maxTokens: 32768,
+    } satisfies Model<"anthropic-messages">,
     "alibaba/qwen3-max-preview": {
       id: "alibaba/qwen3-max-preview",
       name: "Qwen3 Max Preview",
@@ -10889,8 +10974,8 @@ export const MODELS = {
       cost: {
         input: 3,
         output: 15,
-        cacheRead: 0,
+        cacheRead: 0.3,
-        cacheWrite: 0,
+        cacheWrite: 3.75,
       },
       contextWindow: 200000,
       maxTokens: 8192,
@@ -11099,6 +11184,23 @@ export const MODELS = {
       contextWindow: 256000,
       maxTokens: 8000,
     } satisfies Model<"anthropic-messages">,
+    "deepseek/deepseek-r1": {
+      id: "deepseek/deepseek-r1",
+      name: "DeepSeek-R1",
+      api: "anthropic-messages",
+      provider: "vercel-ai-gateway",
+      baseUrl: "https://ai-gateway.vercel.sh",
+      reasoning: true,
+      input: ["text"],
+      cost: {
+        input: 1.35,
+        output: 5.4,
+        cacheRead: 0,
+        cacheWrite: 0,
+      },
+      contextWindow: 128000,
+      maxTokens: 8192,
+    } satisfies Model<"anthropic-messages">,
     "deepseek/deepseek-v3": {
       id: "deepseek/deepseek-v3",
       name: "DeepSeek V3 0324",
@@ -11125,13 +11227,13 @@ export const MODELS = {
       reasoning: true,
       input: ["text"],
       cost: {
-        input: 0.21,
+        input: 0.5,
-        output: 0.7899999999999999,
+        output: 1.5,
         cacheRead: 0,
         cacheWrite: 0,
       },
       contextWindow: 163840,
-      maxTokens: 128000,
+      maxTokens: 16384,
     } satisfies Model<"anthropic-messages">,
     "deepseek/deepseek-v3.1-terminus": {
       id: "deepseek/deepseek-v3.1-terminus",
@@ -11184,6 +11286,40 @@ export const MODELS = {
       contextWindow: 128000,
       maxTokens: 64000,
     } satisfies Model<"anthropic-messages">,
+    "google/gemini-2.0-flash": {
+      id: "google/gemini-2.0-flash",
+      name: "Gemini 2.0 Flash",
+      api: "anthropic-messages",
+      provider: "vercel-ai-gateway",
+      baseUrl: "https://ai-gateway.vercel.sh",
+      reasoning: false,
+      input: ["text", "image"],
+      cost: {
+        input: 0.15,
+        output: 0.6,
+        cacheRead: 0,
+        cacheWrite: 0,
+      },
+      contextWindow: 1048576,
+      maxTokens: 8192,
+    } satisfies Model<"anthropic-messages">,
+    "google/gemini-2.0-flash-lite": {
+      id: "google/gemini-2.0-flash-lite",
+      name: "Gemini 2.0 Flash Lite",
+      api: "anthropic-messages",
+      provider: "vercel-ai-gateway",
+      baseUrl: "https://ai-gateway.vercel.sh",
+      reasoning: false,
+      input: ["text", "image"],
+      cost: {
+        input: 0.075,
+        output: 0.3,
+        cacheRead: 0,
+        cacheWrite: 0,
+      },
+      contextWindow: 1048576,
+      maxTokens: 8192,
+    } satisfies Model<"anthropic-messages">,
     "google/gemini-2.5-flash": {
       id: "google/gemini-2.5-flash",
       name: "Gemini 2.5 Flash",
@@ -11191,11 +11327,11 @@ export const MODELS = {
       provider: "vercel-ai-gateway",
       baseUrl: "https://ai-gateway.vercel.sh",
       reasoning: true,
-      input: ["text"],
+      input: ["text", "image"],
       cost: {
         input: 0.3,
         output: 2.5,
-        cacheRead: 0,
+        cacheRead: 0.03,
         cacheWrite: 0,
       },
       contextWindow: 1000000,
@@ -11259,11 +11395,11 @@ export const MODELS = {
       provider: "vercel-ai-gateway",
       baseUrl: "https://ai-gateway.vercel.sh",
       reasoning: true,
-      input: ["text"],
+      input: ["text", "image"],
       cost: {
         input: 1.25,
         output: 10,
-        cacheRead: 0,
+        cacheRead: 0.125,
         cacheWrite: 0,
       },
       contextWindow: 1048576,
@@ -11284,7 +11420,7 @@ export const MODELS = {
         cacheWrite: 0,
       },
       contextWindow: 1000000,
-      maxTokens: 64000,
+      maxTokens: 65000,
     } satisfies Model<"anthropic-messages">,
     "google/gemini-3-pro-preview": {
       id: "google/gemini-3-pro-preview",
@@ -11386,7 +11522,7 @@ export const MODELS = {
         cacheWrite: 0,
       },
       contextWindow: 128000,
-      maxTokens: 8192,
+      maxTokens: 100000,
     } satisfies Model<"anthropic-messages">,
     "meituan/longcat-flash-thinking": {
       id: "meituan/longcat-flash-thinking",
@@ -11414,13 +11550,13 @@ export const MODELS = {
       reasoning: false,
       input: ["text"],
       cost: {
-        input: 0.39999999999999997,
+        input: 0.72,
-        output: 0.39999999999999997,
+        output: 0.72,
         cacheRead: 0,
         cacheWrite: 0,
       },
-      contextWindow: 131072,
+      contextWindow: 128000,
-      maxTokens: 16384,
+      maxTokens: 8192,
     } satisfies Model<"anthropic-messages">,
     "meta/llama-3.1-8b": {
       id: "meta/llama-3.1-8b",
@@ -11431,12 +11567,12 @@ export const MODELS = {
       reasoning: false,
       input: ["text"],
       cost: {
-        input: 0.03,
+        input: 0.09999999999999999,
-        output: 0.049999999999999996,
+        output: 0.09999999999999999,
         cacheRead: 0,
         cacheWrite: 0,
       },
-      contextWindow: 131072,
+      contextWindow: 128000,
       maxTokens: 16384,
     } satisfies Model<"anthropic-messages">,
     "meta/llama-3.2-11b": {
@@ -11499,12 +11635,12 @@ export const MODELS = {
       reasoning: false,
       input: ["text", "image"],
       cost: {
-        input: 0.15,
+        input: 0.24,
-        output: 0.6,
+        output: 0.9700000000000001,
         cacheRead: 0,
         cacheWrite: 0,
       },
-      contextWindow: 131072,
+      contextWindow: 128000,
       maxTokens: 8192,
     } satisfies Model<"anthropic-messages">,
     "meta/llama-4-scout": {
@@ -11516,12 +11652,12 @@ export const MODELS = {
       reasoning: false,
       input: ["text", "image"],
       cost: {
-        input: 0.08,
+        input: 0.16999999999999998,
-        output: 0.3,
+        output: 0.66,
         cacheRead: 0,
         cacheWrite: 0,
       },
-      contextWindow: 131072,
+      contextWindow: 128000,
       maxTokens: 8192,
     } satisfies Model<"anthropic-messages">,
     "minimax/minimax-m2": {
@@ -11552,8 +11688,8 @@ export const MODELS = {
       cost: {
         input: 0.3,
         output: 1.2,
-        cacheRead: 0.15,
+        cacheRead: 0.03,
-        cacheWrite: 0,
+        cacheWrite: 0.375,
       },
       contextWindow: 204800,
       maxTokens: 131072,
@@ -11788,14 +11924,31 @@ export const MODELS = {
       reasoning: false,
       input: ["text"],
       cost: {
-        input: 0.5,
+        input: 0.6,
-        output: 2,
+        output: 2.5,
         cacheRead: 0,
         cacheWrite: 0,
       },
       contextWindow: 131072,
       maxTokens: 16384,
     } satisfies Model<"anthropic-messages">,
+    "moonshotai/kimi-k2-0905": {
+      id: "moonshotai/kimi-k2-0905",
+      name: "Kimi K2 0905",
+      api: "anthropic-messages",
+      provider: "vercel-ai-gateway",
+      baseUrl: "https://ai-gateway.vercel.sh",
+      reasoning: false,
+      input: ["text"],
+      cost: {
+        input: 0.6,
+        output: 2.5,
+        cacheRead: 0,
+        cacheWrite: 0,
+      },
+      contextWindow: 256000,
+      maxTokens: 16384,
+    } satisfies Model<"anthropic-messages">,
     "moonshotai/kimi-k2-thinking": {
       id: "moonshotai/kimi-k2-thinking",
       name: "Kimi K2 Thinking",
@@ -11805,13 +11958,13 @@ export const MODELS = {
       reasoning: true,
       input: ["text"],
       cost: {
-        input: 0.47,
+        input: 0.6,
-        output: 2,
+        output: 2.5,
-        cacheRead: 0.14100000000000001,
+        cacheRead: 0.15,
         cacheWrite: 0,
       },
-      contextWindow: 216144,
+      contextWindow: 262114,
-      maxTokens: 216144,
+      maxTokens: 262114,
     } satisfies Model<"anthropic-messages">,
     "moonshotai/kimi-k2-thinking-turbo": {
       id: "moonshotai/kimi-k2-thinking-turbo",
@@ -11856,13 +12009,13 @@ export const MODELS = {
       reasoning: true,
       input: ["text", "image"],
       cost: {
-        input: 0.5,
+        input: 0.6,
-        output: 2.8,
+        output: 3,
-        cacheRead: 0,
+        cacheRead: 0.09999999999999999,
         cacheWrite: 0,
       },
-      contextWindow: 256000,
+      contextWindow: 262114,
-      maxTokens: 256000,
+      maxTokens: 262114,
     } satisfies Model<"anthropic-messages">,
     "nvidia/nemotron-nano-12b-v2-vl": {
       id: "nvidia/nemotron-nano-12b-v2-vl",
@@ -11890,8 +12043,8 @@ export const MODELS = {
       reasoning: true,
       input: ["text"],
       cost: {
-        input: 0.04,
+        input: 0.06,
-        output: 0.16,
+        output: 0.22999999999999998,
         cacheRead: 0,
         cacheWrite: 0,
       },
@@ -11960,7 +12113,7 @@ export const MODELS = {
       cost: {
         input: 0.09999999999999999,
         output: 0.39999999999999997,
-        cacheRead: 0.03,
+        cacheRead: 0.024999999999999998,
         cacheWrite: 0,
       },
       contextWindow: 1047576,
@@ -12011,7 +12164,7 @@ export const MODELS = {
       cost: {
         input: 1.25,
         output: 10,
-        cacheRead: 0.13,
+        cacheRead: 0.125,
         cacheWrite: 0,
       },
       contextWindow: 400000,
@@ -12041,11 +12194,11 @@ export const MODELS = {
       provider: "vercel-ai-gateway",
       baseUrl: "https://ai-gateway.vercel.sh",
       reasoning: true,
-      input: ["text", "image"],
+      input: ["text"],
       cost: {
         input: 1.25,
         output: 10,
-        cacheRead: 0.13,
+        cacheRead: 0.125,
         cacheWrite: 0,
       },
       contextWindow: 400000,
@@ -12062,7 +12215,7 @@ export const MODELS = {
       cost: {
         input: 0.25,
         output: 2,
-        cacheRead: 0.03,
+        cacheRead: 0.024999999999999998,
         cacheWrite: 0,
       },
       contextWindow: 400000,
@@ -12079,7 +12232,7 @@ export const MODELS = {
       cost: {
         input: 0.049999999999999996,
         output: 0.39999999999999997,
-        cacheRead: 0.01,
+        cacheRead: 0.005,
         cacheWrite: 0,
       },
       contextWindow: 400000,
@@ -12113,7 +12266,7 @@ export const MODELS = {
       cost: {
         input: 1.25,
         output: 10,
-        cacheRead: 0.13,
+        cacheRead: 0.125,
         cacheWrite: 0,
       },
       contextWindow: 400000,
@@ -12164,7 +12317,7 @@ export const MODELS = {
       cost: {
         input: 1.25,
         output: 10,
-        cacheRead: 0.13,
+        cacheRead: 0.125,
         cacheWrite: 0,
       },
       contextWindow: 128000,
@@ -12181,7 +12334,7 @@ export const MODELS = {
       cost: {
         input: 1.25,
         output: 10,
-        cacheRead: 0.13,
+        cacheRead: 0.125,
         cacheWrite: 0,
       },
       contextWindow: 400000,
@@ -12198,7 +12351,7 @@ export const MODELS = {
       cost: {
         input: 1.75,
         output: 14,
-        cacheRead: 0.18,
+        cacheRead: 0.175,
         cacheWrite: 0,
       },
       contextWindow: 400000,
@@ -12303,7 +12456,7 @@ export const MODELS = {
         cacheRead: 0.25,
         cacheWrite: 0,
       },
-      contextWindow: 200000,
+      contextWindow: 1050000,
       maxTokens: 128000,
     } satisfies Model<"anthropic-messages">,
     "openai/gpt-5.4-pro": {
@@ -12320,26 +12473,9 @@ export const MODELS = {
         cacheRead: 0,
         cacheWrite: 0,
       },
-      contextWindow: 200000,
+      contextWindow: 1050000,
       maxTokens: 128000,
     } satisfies Model<"anthropic-messages">,
-    "openai/gpt-oss-120b": {
-      id: "openai/gpt-oss-120b",
-      name: "gpt-oss-120b",
-      api: "anthropic-messages",
-      provider: "vercel-ai-gateway",
-      baseUrl: "https://ai-gateway.vercel.sh",
-      reasoning: true,
-      input: ["text"],
-      cost: {
-        input: 0.09999999999999999,
-        output: 0.5,
-        cacheRead: 0,
-        cacheWrite: 0,
-      },
-      contextWindow: 131072,
-      maxTokens: 131072,
-    } satisfies Model<"anthropic-messages">,
     "openai/gpt-oss-20b": {
       id: "openai/gpt-oss-20b",
       name: "gpt-oss-20b",
@@ -12757,9 +12893,9 @@ export const MODELS = {
       reasoning: true,
       input: ["text"],
       cost: {
-        input: 0.09,
+        input: 0.09999999999999999,
-        output: 0.29,
+        output: 0.3,
-        cacheRead: 0,
+        cacheRead: 0.02,
         cacheWrite: 0,
       },
       contextWindow: 262144,
@@ -12776,11 +12912,11 @@ export const MODELS = {
       cost: {
         input: 0.6,
         output: 2.2,
-        cacheRead: 0,
+        cacheRead: 0.11,
         cacheWrite: 0,
       },
-      contextWindow: 131072,
+      contextWindow: 128000,
-      maxTokens: 131072,
+      maxTokens: 96000,
     } satisfies Model<"anthropic-messages">,
     "zai/glm-4.5-air": {
       id: "zai/glm-4.5-air",
@@ -12805,16 +12941,16 @@ export const MODELS = {
       api: "anthropic-messages",
       provider: "vercel-ai-gateway",
       baseUrl: "https://ai-gateway.vercel.sh",
-      reasoning: true,
+      reasoning: false,
       input: ["text", "image"],
       cost: {
         input: 0.6,
         output: 1.7999999999999998,
-        cacheRead: 0,
+        cacheRead: 0.11,
         cacheWrite: 0,
       },
-      contextWindow: 65536,
+      contextWindow: 66000,
-      maxTokens: 16384,
+      maxTokens: 16000,
     } satisfies Model<"anthropic-messages">,
     "zai/glm-4.6": {
       id: "zai/glm-4.6",
@@ -12876,12 +13012,12 @@ export const MODELS = {
       reasoning: true,
       input: ["text"],
       cost: {
-        input: 0.43,
+        input: 0.6,
-        output: 1.75,
+        output: 2.2,
-        cacheRead: 0.08,
+        cacheRead: 0,
         cacheWrite: 0,
       },
-      contextWindow: 202752,
+      contextWindow: 200000,
       maxTokens: 120000,
     } satisfies Model<"anthropic-messages">,
     "zai/glm-4.7-flash": {
@@ -12927,9 +13063,9 @@ export const MODELS = {
       reasoning: true,
       input: ["text"],
       cost: {
-        input: 0.7999999999999999,
+        input: 1,
-        output: 2.56,
+        output: 3.1999999999999997,
-        cacheRead: 0.16,
+        cacheRead: 0.19999999999999998,
         cacheWrite: 0,
       },
       contextWindow: 202800,

packages/coding-agent/src/core/gateway/internal-types.ts

@@ -55,7 +55,29 @@ export type GatewayEvent =
     }
   | { type: "message_complete"; sessionKey: string; text: string }
   | { type: "error"; sessionKey: string; error: string }
-  | { type: "aborted"; sessionKey: string };
+  | { type: "aborted"; sessionKey: string }
+  | {
+      type: "structured_part";
+      sessionKey: string;
+      partType: "teamActivity";
+      payload: {
+        teamId: string;
+        status: string;
+        members: Array<{ id: string; name: string; role?: string; status: string; message?: string }>;
+      };
+    }
+  | {
+      type: "structured_part";
+      sessionKey: string;
+      partType: "media";
+      payload: { url: string; mimeType?: string };
+    }
+  | {
+      type: "structured_part";
+      sessionKey: string;
+      partType: "error";
+      payload: { code: string; message: string };
+    };
 
 export interface ManagedGatewaySession {
   sessionKey: string;

packages/coding-agent/src/core/gateway/runtime.ts

@@ -7,6 +7,7 @@ import {
 import { rm } from "node:fs/promises";
 import { join } from "node:path";
 import { URL } from "node:url";
+import type { AgentMessage } from "@mariozechner/companion-agent-core";
 import type { AgentSession, AgentSessionEvent } from "../agent-session.js";
 import type { Settings } from "../settings-manager.js";
 import { extractMessageText, getLastAssistantText } from "./helpers.js";
@@ -29,6 +30,7 @@ import type {
   ModelInfo,
 } from "./types.js";
 import {
+  createGatewayStructuredPartListener,
   createVercelStreamListener,
   errorVercelStream,
   extractUserText,
@@ -265,7 +267,7 @@ export class GatewayRuntime {
     sessionKey: string,
     listener: (event: GatewayEvent) => void,
   ): Promise<() => void> {
-    const managedSession = await this.requireExistingSession(sessionKey);
+    const managedSession = await this.ensureSession(sessionKey);
     managedSession.listeners.add(listener);
     listener({
       type: "hello",
@@ -567,6 +569,7 @@ export class GatewayRuntime {
             sessionKey: managedSession.sessionKey,
             text: extractMessageText(event.message),
           });
+          this.emitStructuredParts(managedSession, event.message);
           return;
         }
         if (event.message.role === "toolResult") {
@@ -654,6 +657,76 @@ export class GatewayRuntime {
     });
   }
 
+  private emitStructuredParts(
+    managedSession: ManagedGatewaySession,
+    message: AgentMessage,
+  ): void {
+    const content = message.content;
+    if (!Array.isArray(content)) return;
+
+    for (const part of content) {
+      if (typeof part !== "object" || part === null) continue;
+      const p = part as Record<string, unknown>;
+
+      if (p.type === "teamActivity") {
+        const teamId = typeof p.teamId === "string" ? p.teamId : "";
+        const status = typeof p.status === "string" ? p.status : "running";
+        if (!teamId) continue;
+        const rawMembers = Array.isArray(p.members) ? p.members : [];
+        const members = rawMembers
+          .filter(
+            (m): m is Record<string, unknown> =>
+              typeof m === "object" && m !== null,
+          )
+          .map((m) => ({
+            id: typeof m.id === "string" ? m.id : "",
+            name: typeof m.name === "string" ? m.name : "Teammate",
+            ...(typeof m.role === "string" ? { role: m.role } : {}),
+            status: typeof m.status === "string" ? m.status : "running",
+            ...(typeof m.message === "string" ? { message: m.message } : {}),
+          }))
+          .filter((m) => m.id.length > 0);
+        this.emit(managedSession, {
+          type: "structured_part",
+          sessionKey: managedSession.sessionKey,
+          partType: "teamActivity",
+          payload: { teamId, status, members },
+        });
+        continue;
+      }
+
+      if (p.type === "image") {
+        const url = typeof p.url === "string" ? p.url : "";
+        if (!url) continue;
+        this.emit(managedSession, {
+          type: "structured_part",
+          sessionKey: managedSession.sessionKey,
+          partType: "media",
+          payload: {
+            url,
+            ...(typeof p.mimeType === "string" ? { mimeType: p.mimeType } : {}),
+          },
+        });
+        continue;
+      }
+
+      if (p.type === "error") {
+        const errorMessage = typeof p.message === "string" ? p.message : "";
+        if (!errorMessage) continue;
+        this.emit(managedSession, {
+          type: "structured_part",
+          sessionKey: managedSession.sessionKey,
+          partType: "error",
+          payload: {
+            code: typeof p.code === "string" ? p.code : "unknown",
+            message: errorMessage,
+          },
+        });
+        continue;
+      }
+    }
+  }
+
   private createSessionState(
     managedSession: ManagedGatewaySession,
   ): GatewaySessionState {
@@ -740,7 +813,28 @@ export class GatewayRuntime {
     request: IncomingMessage,
     response: ServerResponse,
   ): Promise<void> {
+    const origin = request.headers.origin;
+    if (origin) {
+      response.setHeader("Access-Control-Allow-Origin", origin);
+      response.setHeader(
+        "Access-Control-Allow-Methods",
+        "GET, POST, PUT, PATCH, DELETE, OPTIONS",
+      );
+      response.setHeader(
+        "Access-Control-Allow-Headers",
+        "Content-Type, Authorization",
+      );
+      response.setHeader("Access-Control-Allow-Credentials", "true");
+    }
+
     const method = request.method ?? "GET";
+
+    if (method === "OPTIONS") {
+      response.writeHead(204);
+      response.end();
+      return;
+    }
+
     const url = new URL(
       request.url ?? "/",
       `http://${request.headers.host ?? `${this.config.bind}:${this.config.port}`}`,
@@ -919,7 +1013,7 @@ export class GatewayRuntime {
     const action = sessionMatch[2];
 
     if (!action && method === "GET") {
-      const session = await this.requireExistingSession(sessionKey);
+      const session = await this.ensureSession(sessionKey);
       this.writeJson(response, 200, { session: this.createSnapshot(session) });
       return;
     }
@@ -1106,7 +1200,10 @@ export class GatewayRuntime {
     response.write("\n");
 
     const listener = createVercelStreamListener(response);
+    const structuredPartListener =
+      createGatewayStructuredPartListener(response);
     let unsubscribe: (() => void) | undefined;
+    let unsubscribeStructured: (() => void) | undefined;
     let streamingActive = false;
 
     const stopStreaming = () => {
@@ -1114,6 +1211,8 @@ export class GatewayRuntime {
       streamingActive = false;
       unsubscribe?.();
       unsubscribe = undefined;
+      unsubscribeStructured?.();
+      unsubscribeStructured = undefined;
     };
 
     // Clean up on client disconnect
@@ -1135,6 +1234,10 @@ export class GatewayRuntime {
         onStart: () => {
           if (clientDisconnected || streamingActive) return;
           unsubscribe = managedSession.session.subscribe(listener);
+          managedSession.listeners.add(structuredPartListener);
+          unsubscribeStructured = () => {
+            managedSession.listeners.delete(structuredPartListener);
+          };
           streamingActive = true;
         },
         onFinish: () => {
@@ -1283,7 +1386,7 @@ export class GatewayRuntime {
     provider: string,
     modelId: string,
   ): Promise<{ ok: true; model: { provider: string; modelId: string } }> {
-    const managed = await this.requireExistingSession(sessionKey);
+    const managed = await this.ensureSession(sessionKey);
     const found = managed.session.modelRegistry.find(provider, modelId);
     if (!found) {
       throw new HttpError(404, `Model not found: ${provider}/${modelId}`);
@@ -1389,7 +1492,8 @@ export class GatewayRuntime {
   }
 
   private getCompanionChannelsSettings(): CompanionChannelsSettings {
-    const globalSettings = this.primarySession.settingsManager.getGlobalSettings();
+    const globalSettings =
+      this.primarySession.settingsManager.getGlobalSettings();
     const projectSettings =
       this.primarySession.settingsManager.getProjectSettings();
     const mergedSettings = mergeRecords(
@@ -1397,7 +1501,9 @@ export class GatewayRuntime {
       isRecord(projectSettings) ? projectSettings : {},
     );
     const piChannels = mergedSettings["companion-channels"];
-    return isRecord(piChannels) ? (piChannels as CompanionChannelsSettings) : {};
+    return isRecord(piChannels)
+      ? (piChannels as CompanionChannelsSettings)
+      : {};
   }
 
   private buildSlackChannelStatus(
@@ -1419,7 +1525,8 @@ export class GatewayRuntime {
 
     if (hasConfig) {
       if (!adapter) {
-        error = 'Slack requires `companion-channels.adapters.slack = { "type": "slack" }`.';
+        error =
+          'Slack requires `companion-channels.adapters.slack = { "type": "slack" }`.';
       } else if (adapterType !== "slack") {
         error = 'Slack adapter type must be "slack".';
       } else if (!appToken) {
@@ -1468,7 +1575,8 @@ export class GatewayRuntime {
       } else if (adapterType !== "telegram") {
         error = 'Telegram adapter type must be "telegram".';
       } else if (!botToken) {
-        error = "Telegram requires companion-channels.adapters.telegram.botToken.";
+        error =
+          "Telegram requires companion-channels.adapters.telegram.botToken.";
       } else if (!pollingEnabled) {
         error =
           "Telegram requires companion-channels.adapters.telegram.polling = true.";

packages/coding-agent/src/core/gateway/session-state.ts

@@ -85,6 +85,41 @@ export function messageContentToHistoryParts(msg: AgentMessage): HistoryPart[] {
           args: toolCall.arguments,
           state: "call",
         });
+      } else if (contentPart.type === "teamActivity") {
+        const activity = contentPart as {
+          type: "teamActivity";
+          teamId: string;
+          status: string;
+          members?: Array<{ id: string; name: string; role?: string; status: string; message?: string }>;
+        };
+        parts.push({
+          type: "teamActivity",
+          teamId: activity.teamId,
+          status: activity.status,
+          members: Array.isArray(activity.members) ? activity.members : [],
+        });
+      } else if (contentPart.type === "image") {
+        const image = contentPart as {
+          type: "image";
+          url: string;
+          mimeType?: string;
+        };
+        parts.push({
+          type: "media",
+          url: image.url,
+          mimeType: image.mimeType,
+        });
+      } else if (contentPart.type === "error") {
+        const error = contentPart as {
+          type: "error";
+          code?: string;
+          message: string;
+        };
+        parts.push({
+          type: "error",
+          code: typeof error.code === "string" ? error.code : "unknown",
+          message: error.message,
+        });
       }
     }
     return parts;

packages/coding-agent/src/core/gateway/types.ts

@@ -77,7 +77,15 @@ export type HistoryPart =
       args: unknown;
       state: string;
       result?: unknown;
-    };
+    }
+  | {
+      type: "teamActivity";
+      teamId: string;
+      status: string;
+      members: Array<{ id: string; name: string; role?: string; status: string; message?: string }>;
+    }
+  | { type: "media"; url: string; mimeType?: string }
+  | { type: "error"; code: string; message: string };
 
 export interface ChannelStatus {
   id: string;

packages/coding-agent/src/core/gateway/vercel-ai-stream.ts

@@ -1,6 +1,7 @@
 import { randomUUID } from "node:crypto";
 import type { ServerResponse } from "node:http";
 import type { AgentSessionEvent } from "../agent-session.js";
+import type { GatewayEvent } from "./internal-types.js";
 
 type TextStreamState = {
   started: boolean;
@@ -324,3 +325,22 @@ export function errorVercelStream(
   writeChunk(response, "[DONE]");
   response.end();
 }
+
+/**
+ * Create a GatewayEvent listener that forwards `structured_part` events to the
+ * response as custom SSE chunks. Returns the listener function so the caller
+ * can subscribe it to managedSession.listeners and unsubscribe on cleanup.
+ */
+export function createGatewayStructuredPartListener(
+  response: ServerResponse,
+): (event: GatewayEvent) => void {
+  return (event: GatewayEvent) => {
+    if (response.writableEnded) return;
+    if (event.type !== "structured_part") return;
+    writeChunk(response, {
+      type: "structured-part",
+      partType: event.partType,
+      payload: event.payload,
+    });
+  };
+}