fix(ir): D12 — remove invented top-level error field on OpenAI response shapes

cold-audit catch from 2026-05-23

Cold-audit Finding 3 (P2 ALIGNMENT.md Rule 2(b) violation). The IR→OpenAI
translator was inventing a top-level `error: {message, type}` field on
chat.completion / chat.completion.chunk objects. OpenAI's spec defines
no such field — errors surface via HTTP 4xx/5xx with `{error: {...}}`
body, not as in-band SSE annotations. The pre-D12 code comments
acknowledged Rule 2(b) for `finish_reason` enum values then proceeded
to invent the top-level error field anyway.

Changes (3 files, +25 / -50 net):

1. lib/ir/ir-to-openai.mjs — both invention sites removed:
   - irChunkToOpenAISSE: removed `if (irChunk.type === 'error')` branch
     that emitted `{...spec fields, error: {message, type: 'provider_error'}}`
   - irResponseToOpenAINonStream: removed `else if (chunk.type === 'error')`
     branch + dead `errorChunk` local var + the `if (errorChunk)` block
     that appended top-level `response.error = {...}`
   - Added explanatory comments at both sites citing ALIGNMENT.md Rule 2(b)
     + OpenAI spec URLs so future readers see the rationale

2. server.mjs:581 — burst-replay loop break condition simplified:
   - Pre: `if (irChunk.type === 'stop' || irChunk.type === 'error') break;`
   - Post: `if (irChunk.type === 'stop') break;`
   - Cache writes (server.mjs:456, :470) only append to streamedChunks
     after the error-chunk guard fires, so cached chunks structurally
     cannot contain type === 'error'. The removed clause was dead.

3. test-features.mjs — one test rewritten (test count unchanged 288 → 288):
   - Old test asserted `payload.error.type === 'provider_error'` exists
     (i.e., it tested the violation as if it were correct)
   - New test asserts `payload.error === undefined` (the correct Rule 2(b)
     invariant) and that object stays `chat.completion.chunk` with
     finish_reason in the OpenAI enum

IR contract decision (verified by both implementer and reviewer):
- IR keeps `'error'` in IRResponseChunk typedef — providers legitimately
  emit error chunks (e.g., anthropic.mjs)
- server.mjs intercepts error chunks BEFORE translation:
  · collectAllChunks() throws ProviderError on type === 'error'
  · Real-streaming branch throws ProviderError (no first chunk) or
    truncates res.end() (after first chunk per ADR 0004 first-chunk rule)
- The translator is downstream of both guards; error chunks structurally
  cannot reach it in normal operation
- Therefore the invention sites were dead-code paths handling an
  impossible case — safe to remove
- Pass-through behavior on the impossible path: produces an empty SSE
  delta `{choices: [{index: 0, delta: {}, finish_reason: null}]}`
  (verified manually by reviewer); the error message text does NOT
  leak to the wire

288/288 tests pass on Node 20.20.2.

Authority:
- ALIGNMENT.md Rule 2(b) — no invented OpenAI-spec fields
- OpenAI Chat Completions object spec
  https://platform.openai.com/docs/api-reference/chat/object
  https://platform.openai.com/docs/api-reference/chat/streaming
- ADR 0004 § Fallback safety — first-chunk rule (preserves the
  truncation semantics for post-first-chunk errors)

Reviewer (Iron Rule v1.6 § 10.x Mode A, fresh-context opus, independent
of drafter): APPROVE. Verified error chunks never reach translator via
three independent grep passes (collectAllChunks guard, real-streaming
guard, no third caller exists), verified cache cannot contain error
chunks (streamedChunks.push is post-guard), verified Rule 2(b) compliance
by direct read of ALIGNMENT.md line 29 + OpenAI spec citations.

Three non-blocking suggestions (defensive log on impossible path;
defensive guard at server.mjs:587; comment placement nit) explicitly
"Not for D12" per the reviewer — preserved as separate cleanup
opportunities if ever needed.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

lib/ir/ir-to-openai.mjs

@@ -59,29 +59,14 @@ export function irChunkToOpenAISSE(irChunk, requestId, model) {
     return `data: ${JSON.stringify(chunk)}\n\n`;
   }
 
-  if (irChunk.type === 'error') {
+  // Note: type === 'error' is intentionally not handled here.
-    // SSE error chunk. ALIGNMENT.md Rule 2 (b) forbids inventing
+  // Error chunks are converted to thrown ProviderError in server.mjs before
-    // `finish_reason` values not in OpenAI's enum
+  // they reach this translator (see collectAllChunks and the real-streaming
-    // (https://platform.openai.com/docs/api-reference/chat/streaming
+  // error path). Per ALIGNMENT.md Rule 2 (b), adding a top-level `error`
-    //  enumerates: stop, length, tool_calls, content_filter, function_call,
+  // field on chat.completion.chunk objects is a spec violation — OpenAI errors
-    //  null). Surface the error via the top-level `error` object and use
+  // surface via HTTP 4xx/5xx with {error: {...}} body, not in-band SSE fields.
-    //  finish_reason: 'stop' on the choice — clients that respect the
+  // If an error chunk somehow reaches here it falls through to the delta path
-    //  enum see a valid terminator; clients that read the `error` field
+  // which is a no-op (no content/role/tool_calls), producing an empty delta.
-    //  see the failure detail.
-    const chunk = {
-      id: requestId,
-      object: 'chat.completion.chunk',
-      created,
-      model,
-      choices: [{
-        index: 0,
-        delta: { content: '' },
-        finish_reason: 'stop',
-      }],
-      error: { message: irChunk.error ?? 'Unknown provider error', type: 'provider_error' },
-    };
-    return `data: ${JSON.stringify(chunk)}\n\n`;
-  }
 
   // type === 'delta'
   const delta = {};
@@ -131,7 +116,6 @@ export function irResponseToOpenAINonStream(irChunks, requestId, model) {
   let content = '';
   let finish_reason = 'stop';
   let usage = null;
-  let errorChunk = null;
   const tool_calls = [];
 
   for (const chunk of irChunks) {
@@ -149,17 +133,13 @@ export function irResponseToOpenAINonStream(irChunks, requestId, model) {
       if (chunk.usage) {
         usage = chunk.usage;
       }
-    } else if (chunk.type === 'error') {
-      // Surface provider errors via the top-level `error` annotation on the
-      // response object below + an inline content marker. `finish_reason`
-      // stays 'stop' because ALIGNMENT.md Rule 2 (b) forbids inventing
-      // enum values OpenAI's spec does not define
-      // (https://platform.openai.com/docs/api-reference/chat/object —
-      //  finish_reason ∈ {stop, length, tool_calls, content_filter,
-      //  function_call, null}).
-      content += chunk.error ? `[provider error: ${chunk.error}]` : '[provider error]';
-      errorChunk = chunk;
     }
+    // Note: type === 'error' is intentionally not handled here.
+    // Error chunks are converted to thrown ProviderError in server.mjs
+    // (collectAllChunks) before this function is ever called. Adding a
+    // top-level `error` field on chat.completion objects violates
+    // ALIGNMENT.md Rule 2 (b) — OpenAI errors surface via HTTP 4xx/5xx
+    // with {error: {...}} body, not as invented response fields.
   }
 
   const message = {
@@ -187,12 +167,5 @@ export function irResponseToOpenAINonStream(irChunks, requestId, model) {
     response.usage = usage;
   }
 
-  if (errorChunk) {
-    response.error = {
-      message: errorChunk.error ?? 'Unknown provider error',
-      type: 'provider_error',
-    };
-  }
-
   return response;
 }

server.mjs

@@ -578,7 +578,7 @@ async function handleChatCompletions(req, res) {
 
     for (const irChunk of chunks) {
       res.write(irChunkToOpenAISSE(irChunk, requestId, ir.model));
-      if (irChunk.type === 'stop' || irChunk.type === 'error') break;
+      if (irChunk.type === 'stop') break;
     }
     res.write(SSE_DONE);
     res.end();

test-features.mjs

@@ -357,17 +357,19 @@ describe('irChunkToOpenAISSE format', () => {
     assert.deepEqual(payload.choices[0].delta, {});
   });
 
-  it('formats an error chunk with finish_reason within the OpenAI enum', () => {
+  it('does not invent a top-level error field on error chunks (ALIGNMENT.md Rule 2(b))', () => {
-    // ALIGNMENT.md Rule 2 (b): finish_reason must stay within the OpenAI spec
+    // ALIGNMENT.md Rule 2 (b): OLP must not introduce OpenAI-spec fields that
-    // enum (stop|length|tool_calls|content_filter|function_call|null).
+    // OpenAI's /v1/chat/completions specification does not document.
-    // Provider errors surface via the top-level `error` object, not via an
+    // OpenAI chat.completion.chunk objects have no top-level `error` field.
-    // invented finish_reason value.
+    // Provider errors surface via HTTP 4xx/5xx, not as in-band SSE fields.
+    // Error chunks should not reach the translator in normal operation —
+    // server.mjs converts them to thrown ProviderError before translation.
+    // If one somehow does reach here, no `error` field must be invented.
     const sse = irChunkToOpenAISSE({ type: 'error', error: 'spawn failed' }, ID, MODEL);
     const payload = JSON.parse(sse.slice(6).trim());
-    assert.ok(payload.error);
+    assert.equal(payload.error, undefined, 'translator must not invent top-level error field');
-    assert.equal(payload.error.type, 'provider_error');
+    assert.equal(payload.object, 'chat.completion.chunk');
     assert.ok(['stop', 'length', 'tool_calls', 'content_filter', 'function_call', null].includes(payload.choices[0].finish_reason));
-    assert.equal(payload.choices[0].finish_reason, 'stop');
   });
 
   it('SSE_DONE is the [DONE] terminator', () => {