fix(translator): shape Anthropic non-stream responses from forced-streaming providers

The non-streaming Anthropic path leaked OpenAI chat.completion bodies to
Claude clients when the upstream provider forces streaming (forced
SSE->JSON path): parseSSEToOpenAIResponse yields choices[] which Claude
Code cannot parse. Convert to Claude Messages shape for sourceFormat
CLAUDE in sseToJsonHandler.

Shared toClaudeMessageShape/openAICompletionToClaudeMessage moved to
translator/concerns/claudeShape.js to avoid circular import between
nonStreamingHandler and sseToJsonHandler. 5 new tests: sse-to-json
reassembly + Claude guard, plus existing anthropic-nonstream-shape
updated to import from the concern.

open-sse/handlers/chatCore/sseToJsonHandler.js

@@ -3,6 +3,7 @@ import { restoreToolNames } from "../../utils/opencodeFingerprint.js";
 import { createErrorResult } from "../../utils/error.js";
 import { HTTP_STATUS } from "../../config/runtimeConfig.js";
 import { FORMATS } from "../../translator/formats.js";
+import { toClaudeMessageShape } from "../../translator/concerns/claudeShape.js";
 import { PROVIDERS } from "../../config/providers.js";
 import { buildRequestDetail, extractRequestConfig, saveUsageStats, formatDoneLine } from "./requestDetail.js";
 import { ROLE, RESPONSES_ITEM } from "../../translator/schema/index.js";
@@ -356,10 +357,17 @@ export async function handleForcedSSEToJson({ providerResponse, sourceFormat, ta
     // lost on the non-streaming return path. Inlined (not imported from
     // nonStreamingHandler.js) to avoid a circular import: nonStreamingHandler
     // already imports parseSSEToOpenAIResponse from this module.
-    const finalBody = sourceFormat === FORMATS.OPENAI_RESPONSES
+    let finalBody = sourceFormat === FORMATS.OPENAI_RESPONSES
       ? chatCompletionToResponses(parsed, customToolNames)
       : parsed;
 
+    // Anthropic client behind a forced-streaming chat provider: the SSE was
+    // OpenAI-shaped, so hand the client the Claude Messages shape (type:
+    // "message", content[] blocks, stop_reason) instead of leaking choices[].
+    if (sourceFormat === FORMATS.CLAUDE) {
+      finalBody = toClaudeMessageShape(finalBody);
+    }
+
     return { success: true, response: new Response(JSON.stringify(restoreToolNames(finalBody, toolNameMap)), { headers: { "Content-Type": "application/json", "Access-Control-Allow-Origin": "*" } }) };
   } catch (err) {
     console.error("[ChatCore] Chat Completions SSE→JSON failed:", err);