Settle the last concrete review regressions on the release branch

A few review threads still pointed at real behavior mismatches: Anthropic tool
JSON deltas were still trimming whitespace, Responses conversion still raised
low token caps to the floor, and the bare /responses GET alias returned upgrade
messaging that did not match the actual websocket entrypoint. This commit makes
those behaviors explicit and locks them with focused tests.

Constraint: Keep the release branch green while fixing only the review findings that have a clear, testable contract
Rejected: Leave these for a follow-up cleanup PR | they were low-diff and directly blocked comment resolution on the current release PR
Confidence: high
Scope-risk: narrow
Directive: Preserve raw stream delta text unless a downstream protocol explicitly requires normalization; whitespace can be semantic in incremental payloads
Tested: Focused vitest for anthropic stream bridge, responses conversion, route aliases; typecheck; git diff --check
Not-tested: Full npm test; the remaining review threads that still need broader product decisions

src/server/routes/proxy/responses.ts

@@ -34,7 +34,7 @@ export async function responsesProxyRoute(app: FastifyInstance) {
   app.get('/responses', async (_request: FastifyRequest, reply: FastifyReply) =>
     reply.code(426).send({
       error: {
-        message: 'WebSocket upgrade required for GET /responses',
+        message: 'WebSocket upgrade required for GET /v1/responses',
         type: 'invalid_request_error',
       },
     }));

src/server/routes/proxy/routeAliases.test.ts

@@ -89,7 +89,7 @@ describe('proxy route aliases', () => {
     expect(handleOpenAiResponsesSurfaceRequestMock).not.toHaveBeenCalled();
     expect(response.json()).toEqual({
       error: {
-        message: 'WebSocket upgrade required for GET /responses',
+        message: 'WebSocket upgrade required for GET /v1/responses',
         type: 'invalid_request_error',
       },
     });

src/server/transformers/anthropic/messages/streamBridge.test.ts

@@ -109,6 +109,30 @@ describe('anthropic messages stream bridge', () => {
     expect(thinkingResult.lines.join('')).toContain('  padded thinking  ');
   });
 
+  it('preserves partial_json spacing for streaming tool-call deltas', () => {
+    const streamContext = anthropicMessagesStream.createContext('claude-test');
+    const downstreamContext = anthropicMessagesStream.createDownstreamContext();
+
+    const result = consumeAnthropicSseEvent(
+      {
+        event: 'content_block_delta',
+        data: JSON.stringify({
+          type: 'content_block_delta',
+          index: 0,
+          delta: {
+            type: 'input_json_delta',
+            partial_json: '{ "city": "Paris" }',
+          },
+        }),
+      },
+      streamContext,
+      downstreamContext,
+      'claude-test',
+    );
+
+    expect(result.lines.join('')).toContain('partial_json":"{ \\"city\\": \\"Paris\\" }"');
+  });
+
   it('preserves responses output-item ordering when streaming anthropic fallback blocks', () => {
     const streamContext = anthropicMessagesStream.createContext('claude-test');
     const downstreamContext = anthropicMessagesStream.createDownstreamContext();

src/server/transformers/anthropic/messages/streamBridge.ts

@@ -867,7 +867,7 @@ function normalizeAnthropicRawEvent(
       return {
         toolCallDeltas: [{
           index,
-          argumentsDelta: asTrimmedString(delta.partial_json),
+          argumentsDelta: asUntouchedString(delta.partial_json),
         }],
       };
     }

src/server/transformers/openai/responses/conversion.test.ts

@@ -719,7 +719,7 @@ describe('convertOpenAiBodyToResponsesBody', () => {
     expect(result.max_output_tokens).toBeUndefined();
   });
 
-  it('clamps explicit chat token limits to the sub2api floor when converting to Responses bodies', () => {
+  it('preserves explicit chat token limits when converting to Responses bodies', () => {
     const result = convertOpenAiBodyToResponsesBody(
       {
         model: 'gpt-5',
@@ -730,7 +730,7 @@ describe('convertOpenAiBodyToResponsesBody', () => {
       false,
     );
 
-    expect(result.max_output_tokens).toBe(128);
+    expect(result.max_output_tokens).toBe(64);
   });
 
   it('uses sub2api-compatible placeholders for empty tool arguments and empty tool outputs', () => {

src/server/transformers/openai/responses/conversion.ts

@@ -618,10 +618,7 @@ export function convertOpenAiBodyToResponsesBody(
   };
 
   if (requestedMaxOutputTokens !== null && requestedMaxOutputTokens > 0) {
-    body.max_output_tokens = Math.max(
-      MIN_RESPONSES_MAX_OUTPUT_TOKENS,
-      Math.trunc(requestedMaxOutputTokens),
-    );
+    body.max_output_tokens = Math.trunc(requestedMaxOutputTokens);
   }
 
   if (systemContents.length > 0) {