fix: обновить Anthropic default и диагностику 410

README.md

@@ -206,7 +206,7 @@ The `teleton setup` wizard generates a fully configured `~/.teleton/config.yaml`
 agent:
   provider: "anthropic"              # anthropic | claude-code | openai | google | xai | groq | openrouter | moonshot | mistral | cerebras | zai | minimax | huggingface | nvidia | gocoon | local
   api_key: "sk-ant-api03-..."
-  model: "claude-opus-4-6"
+  model: "claude-opus-4-8"
   utility_model: "claude-haiku-4-5-20251001"  # optional: summarization, compaction, vision
   max_agentic_iterations: 5
   session_reset_policy:

config.example.yaml

@@ -5,8 +5,8 @@
 agent:
   provider: "anthropic"              # anthropic | openai | google | xai | groq | openrouter | nvidia
   api_key: "YOUR_API_KEY"
-  model: "claude-haiku-4-5-20251001" # Model ID (varies by provider)
-  # utility_model: "claude-3-5-haiku-20241022"  # Optional: cheap model for summarization
+  model: "claude-opus-4-8" # Model ID (varies by provider)
+  # utility_model: "claude-haiku-4-5-20251001"  # Optional: cheap model for summarization
   max_tokens: 4096
   temperature: 0.7
   # system_prompt: null              # Custom system prompt override (null = use SOUL.md)

docs/configuration.md

@@ -40,7 +40,7 @@ LLM provider and agentic loop configuration.
 |-----|------|---------|-------------|
 | `agent.provider` | `enum` | `"anthropic"` | LLM provider. One of: `anthropic`, `claude-code`, `openai`, `google`, `xai`, `groq`, `openrouter`, `moonshot`, `mistral`, `cerebras`, `zai`, `minimax`, `huggingface`, `nvidia`, `gocoon`, `local`. |
 | `agent.api_key` | `string` | `""` | API key for the chosen provider. Can be overridden with `TELETON_API_KEY` env var. |
-| `agent.model` | `string` | `"claude-haiku-4-5-20251001"` | Primary model ID. Auto-detected from provider if not set (only for non-Anthropic providers). |
+| `agent.model` | `string` | `"claude-opus-4-8"` | Primary model ID. Auto-detected from provider if not set (only for non-Anthropic providers). |
 | `agent.utility_model` | `string` | *auto-detected* | Cheap/fast model used for summarization and compaction. If omitted, the platform selects one based on the provider (e.g., `claude-haiku-4-5-20251001` for Anthropic, `gpt-4o-mini` for OpenAI). |
 | `agent.base_url` | `string` | *optional* | Base URL for local LLM server (e.g., `http://localhost:11434/v1`). Must be a valid URL. |
 | `agent.max_tokens` | `number` | `4096` | Maximum tokens in each LLM response. |
@@ -65,7 +65,7 @@ Controls when conversation sessions are cleared, giving the agent a fresh memory
 agent:
   provider: "anthropic"
   api_key: "sk-ant-..."
-  model: "claude-haiku-4-5-20251001"
+  model: "claude-opus-4-8"
   utility_model: "claude-haiku-4-5-20251001"
   max_tokens: 4096
   temperature: 0.7
@@ -83,7 +83,7 @@ When you change the `provider` and omit `model`, the platform auto-selects:
 
 | Provider | Default Model | Default Utility Model |
 |----------|--------------|----------------------|
-| `anthropic` | `claude-haiku-4-5-20251001` | `claude-haiku-4-5-20251001` |
+| `anthropic` | `claude-opus-4-8` | `claude-haiku-4-5-20251001` |
 | `codex` | `gpt-5.5` | `gpt-5.1-codex-mini` |
 | `openai` | `gpt-5.5` | `gpt-4o-mini` |
 | `google` | `gemini-2.5-flash` | `gemini-2.0-flash-lite` |
@@ -737,7 +737,7 @@ meta:
 agent:
   provider: "anthropic"
   api_key: "sk-ant-..."
-  model: "claude-haiku-4-5-20251001"
+  model: "claude-opus-4-8"
   max_tokens: 4096
   temperature: 0.7
   max_agentic_iterations: 5

src/agent/__tests__/client-abort.test.ts

@@ -88,4 +88,23 @@ describe("chatWithContext abort signal handling", () => {
     expect(completeSignal?.aborted).toBe(true);
     expect(completeSignal?.reason).toBe(reason);
   });
+
+  it("omits temperature for Claude Opus 4.8 requests", async () => {
+    const { chatWithContext } = await import("../client.js");
+
+    await chatWithContext(
+      {
+        ...agentConfig(),
+        provider: "anthropic",
+        api_key: "sk-ant-test-key",
+        model: "claude-opus-4-8",
+        temperature: 0.7,
+      },
+      {
+        context: { messages: [], systemPrompt: "test" },
+      }
+    );
+
+    expect(getCapturedCompleteOptions()).not.toHaveProperty("temperature");
+  });
 });

src/agent/__tests__/runtime-retry.test.ts

@@ -223,6 +223,26 @@ describe("AgentRuntime retry backoff", () => {
     expect(chatWithContextMock).toHaveBeenCalledTimes(2);
   });
 
+  it("reports Anthropic 410 responses as an unavailable model with recovery guidance", async () => {
+    const runtime = await createRuntime();
+    chatWithContextMock.mockResolvedValueOnce(errorResponse("410 status code (no body)"));
+
+    await expect(
+      runtime.processMessage({
+        chatId: "1001",
+        userMessage: "hello",
+        userName: "Owner",
+        toolContext: {
+          senderId: 1001,
+          config: makeConfig(),
+        },
+      })
+    ).rejects.toThrow(
+      'Provider anthropic rejected model "claude-opus-4-6" with 410 status code (no body)'
+    );
+    expect(chatWithContextMock).toHaveBeenCalledTimes(1);
+  });
+
   it("stops a pending rate-limit backoff when the signal is aborted", async () => {
     vi.useFakeTimers();
     const runtime = await createRuntime();

src/agent/__tests__/runtime-utils.test.ts

@@ -6,6 +6,8 @@ import {
   sleepWithAbort,
   isNetworkError,
   isNetworkErrorMessage,
+  isModelUnavailableError,
+  getModelUnavailableDiagnostic,
   getEmptyResponseDiagnostic,
   getEmptyResponseRecoveryPrompt,
   trimRagContext,
@@ -379,6 +381,35 @@ describe("isNetworkErrorMessage", () => {
   });
 });
 
+describe("model unavailable diagnostics", () => {
+  it("detects Anthropic 410 responses as unavailable model errors", () => {
+    expect(isModelUnavailableError("410 status code (no body)")).toBe(true);
+  });
+
+  it("returns update guidance when the default model differs from the configured model", () => {
+    const result = getModelUnavailableDiagnostic({
+      provider: "anthropic",
+      model: "claude-opus-4-6",
+      defaultModel: "claude-opus-4-8",
+      errorMessage: "410 status code (no body)",
+    });
+
+    expect(result).toContain('Provider anthropic rejected model "claude-opus-4-6"');
+    expect(result).toContain('Update agent.model to "claude-opus-4-8"');
+  });
+
+  it("does not produce guidance for unrelated API errors", () => {
+    expect(
+      getModelUnavailableDiagnostic({
+        provider: "anthropic",
+        model: "claude-opus-4-8",
+        defaultModel: "claude-opus-4-8",
+        errorMessage: "500 internal server error",
+      })
+    ).toBeNull();
+  });
+});
+
 // ─── T14b: getEmptyResponseDiagnostic ──────────────────────────────────
 
 describe("getEmptyResponseDiagnostic", () => {

src/agent/client.ts

@@ -58,6 +58,12 @@ function normalizeProviderTemperature(provider: SupportedProvider, temperature:
   return Math.min(Math.max(temperature, 0), NVIDIA_MAX_TEMPERATURE);
 }
 
+function shouldOmitProviderTemperature(provider: SupportedProvider, modelId: string): boolean {
+  if (provider !== "anthropic" && provider !== "claude-code") return false;
+
+  return /^claude-opus-4-(?:7|8)(?:$|-)/.test(modelId.toLowerCase());
+}
+
 export interface ChatOptions {
   systemPrompt?: string;
   context: Context;
@@ -134,13 +140,15 @@ export async function chatWithContext(
   const completeOptions: Record<string, unknown> = {
     apiKey: getEffectiveApiKey(provider, config.api_key),
     maxTokens: options.maxTokens ?? config.max_tokens,
-    temperature,
     sessionId: options.sessionId,
     cacheRetention: provider === "nvidia" ? NVIDIA_CACHE_RETENTION : DEFAULT_CACHE_RETENTION,
     signal: options.signal
       ? AbortSignal.any([options.signal, requestTimeoutSignal])
       : requestTimeoutSignal,
   };
+  if (!shouldOmitProviderTemperature(provider, config.model)) {
+    completeOptions.temperature = temperature;
+  }
 
   let response = await complete(model, context, completeOptions as ProviderStreamOptions);
 
@@ -191,10 +199,12 @@ export function streamWithContext(config: AgentConfig, options: ChatOptions): St
   const streamOptions: Record<string, unknown> = {
     apiKey: getEffectiveApiKey(provider, config.api_key),
     maxTokens: options.maxTokens ?? config.max_tokens,
-    temperature,
     sessionId: options.sessionId,
     cacheRetention: provider === "nvidia" ? NVIDIA_CACHE_RETENTION : DEFAULT_CACHE_RETENTION,
   };
+  if (!shouldOmitProviderTemperature(provider, config.model)) {
+    streamOptions.temperature = temperature;
+  }
 
   const eventStream = stream(model, context, streamOptions as ProviderStreamOptions);
 

src/agent/runtime-utils.ts

@@ -97,6 +97,44 @@ export function isNetworkErrorMessage(message: string): boolean {
   );
 }
 
+export interface ModelUnavailableDiagnosticInput {
+  provider: string;
+  model: string;
+  errorMessage: string;
+  defaultModel?: string;
+}
+
+export function isModelUnavailableError(errorMessage: string): boolean {
+  const lower = errorMessage.toLowerCase();
+  return (
+    /\b410\b/.test(lower) ||
+    lower.includes("gone") ||
+    lower.includes("model not found") ||
+    lower.includes("model_not_found") ||
+    lower.includes("model unavailable") ||
+    lower.includes("model_unavailable") ||
+    lower.includes("not available or recognized") ||
+    lower.includes("does not exist") ||
+    (lower.includes("selected model") && lower.includes("not available"))
+  );
+}
+
+export function getModelUnavailableDiagnostic(
+  input: ModelUnavailableDiagnosticInput
+): string | null {
+  if (!isModelUnavailableError(input.errorMessage)) return null;
+
+  const base =
+    `Provider ${input.provider} rejected model "${input.model}" with ${input.errorMessage}. ` +
+    "This usually means the configured model ID or provider endpoint is no longer available.";
+
+  if (input.defaultModel && input.defaultModel !== input.model) {
+    return `${base} Update agent.model to "${input.defaultModel}" or run teleton setup again.`;
+  }
+
+  return `${base} Choose a currently supported model for this provider and update agent.model.`;
+}
+
 export interface EmptyResponseDiagnosticInput {
   provider: string;
   model: string;

src/agent/runtime.ts

@@ -87,6 +87,7 @@ import {
   isNetworkErrorMessage,
   getEmptyResponseDiagnostic,
   getEmptyResponseRecoveryPrompt,
+  getModelUnavailableDiagnostic,
   trimRagContext,
   LoopStallDetector,
   sleepWithAbort,
@@ -1098,6 +1099,17 @@ export class AgentRuntime {
               `Network error after ${NETWORK_ERROR_MAX_RETRIES} retries. Please check your connection and try again.`
             );
           } else {
+            const modelDiagnostic = getModelUnavailableDiagnostic({
+              provider,
+              model: this.config.agent.model,
+              errorMessage: errorMsg,
+              defaultModel: providerMeta.defaultModel,
+            });
+            if (modelDiagnostic) {
+              log.error(`🚨 ${modelDiagnostic}`);
+              throw new Error(modelDiagnostic);
+            }
+
             log.error(`🚨 API error: ${errorMsg}`);
             throw new Error(`API error: ${errorMsg || "Unknown error"}`);
           }

src/config/__tests__/loader.test.ts

@@ -420,7 +420,7 @@ describe("Config Loader", () => {
       const config = loadConfig(TEST_CONFIG_PATH);
 
       // Agent defaults
-      expect(config.agent.model).toBe("claude-opus-4-6");
+      expect(config.agent.model).toBe("claude-opus-4-8");
       expect(config.agent.max_tokens).toBe(4096);
       expect(config.agent.temperature).toBe(0.7);
       expect(config.agent.max_agentic_iterations).toBe(5);

src/config/model-catalog.ts

@@ -96,15 +96,20 @@ export const GROQ_TTS_MODELS: ModelOption[] = [
 
 export const MODEL_OPTIONS: Record<string, ModelOption[]> = {
   anthropic: [
+    {
+      value: "claude-opus-4-8",
+      name: "Claude Opus 4.8",
+      description: "Current default, 1M ctx, $5/$25",
+    },
     {
       value: "claude-opus-4-7",
       name: "Claude Opus 4.7",
-      description: "Most capable available, 1M ctx, reasoning, $5/$25",
+      description: "Previous gen, 1M ctx, reasoning, $5/$25",
     },
     {
       value: "claude-opus-4-6",
       name: "Claude Opus 4.6",
-      description: "Previous gen, 1M ctx, reasoning, $5/$25",
+      description: "Legacy, may be retired, 1M ctx, $5/$25",
     },
     {
       value: "claude-opus-4-5-20251101",
@@ -119,7 +124,7 @@ export const MODEL_OPTIONS: Record<string, ModelOption[]> = {
     {
       value: "claude-haiku-4-5-20251001",
       name: "Claude Haiku 4.5",
-      description: "Fast & cheap, 200K ctx, $1/$5 (default)",
+      description: "Fast & cheap, 200K ctx, $1/$5",
     },
   ],
   openai: [

src/config/providers.ts

@@ -37,7 +37,7 @@ const PROVIDER_REGISTRY: Record<SupportedProvider, ProviderMetadata> = {
     keyPrefix: "sk-ant-",
     keyHint: "sk-ant-api03-...",
     consoleUrl: "https://console.anthropic.com/",
-    defaultModel: "claude-opus-4-6",
+    defaultModel: "claude-opus-4-8",
     utilityModel: "claude-haiku-4-5-20251001",
     toolLimit: null,
     piAiProvider: "anthropic",
@@ -49,7 +49,7 @@ const PROVIDER_REGISTRY: Record<SupportedProvider, ProviderMetadata> = {
     keyPrefix: "sk-ant-",
     keyHint: "Auto-detected from Claude Code",
     consoleUrl: "https://console.anthropic.com/",
-    defaultModel: "claude-opus-4-6",
+    defaultModel: "claude-opus-4-8",
     utilityModel: "claude-haiku-4-5-20251001",
     toolLimit: null,
     piAiProvider: "anthropic",

src/config/schema.ts

@@ -72,7 +72,7 @@ export const AgentConfigSchema = z.object({
     .url()
     .optional()
     .describe("Base URL for local LLM server (e.g. http://localhost:11434/v1)"),
-  model: z.string().default("claude-opus-4-6"),
+  model: z.string().default("claude-opus-4-8"),
   utility_model: z
     .string()
     .optional()

src/providers/__tests__/anthropic-provider.test.ts

@@ -0,0 +1,36 @@
+import { describe, expect, it } from "vitest";
+import { getModelsForProvider } from "../../config/model-catalog.js";
+import { getProviderMetadata } from "../../config/providers.js";
+import { AgentConfigSchema } from "../../config/schema.js";
+import { getProviderModel } from "../model-resolver.js";
+
+describe("Anthropic provider registration", () => {
+  it("defaults to the current Opus model for primary calls", () => {
+    const meta = getProviderMetadata("anthropic");
+
+    expect(meta.defaultModel).toBe("claude-opus-4-8");
+    expect(meta.utilityModel).toBe("claude-haiku-4-5-20251001");
+  });
+
+  it("resolves Claude Opus 4.8 even before pi-ai ships it in the generated registry", () => {
+    const model = getProviderModel("anthropic", "claude-opus-4-8");
+
+    expect(model.id).toBe("claude-opus-4-8");
+    expect(model.api).toBe("anthropic-messages");
+    expect(model.provider).toBe("anthropic");
+    expect(model.contextWindow).toBe(1_000_000);
+    expect(model.maxTokens).toBe(128_000);
+  });
+
+  it("exposes Claude Opus 4.8 first in the setup model catalog", () => {
+    const models = getModelsForProvider("anthropic");
+
+    expect(models[0]?.value).toBe("claude-opus-4-8");
+  });
+
+  it("uses Claude Opus 4.8 in AgentConfigSchema defaults", () => {
+    const result = AgentConfigSchema.parse({});
+
+    expect(result.model).toBe("claude-opus-4-8");
+  });
+});

src/providers/__tests__/claude-code-provider.test.ts

@@ -14,7 +14,7 @@ describe("claude-code provider registration", () => {
     expect(meta.displayName).toBe("Claude Code (Auto)");
     expect(meta.piAiProvider).toBe("anthropic");
     expect(meta.toolLimit).toBeNull();
-    expect(meta.defaultModel).toBe("claude-opus-4-6");
+    expect(meta.defaultModel).toBe("claude-opus-4-8");
     expect(meta.utilityModel).toBe("claude-haiku-4-5-20251001");
     expect(meta.keyPrefix).toBe("sk-ant-");
   });