feat(rag): retry a failed LLM request before the first token lands — BOR_LLM_RETRIES/BOR_LLM_RETRY_DELAY with a live 'retrying' status

app/schemas.py

@@ -108,6 +108,30 @@ class ChatErrorEvent(BaseModel):
     detail: str
 
 
+class ChatRetryEvent(BaseModel):
+    """SSE retry event: an LLM request is restarted before the first token
+    (phase 67, owner-locked 2026-09-01).
+
+    Sibling of :class:`ChatErrorEvent`, but transient — the client shows a
+    live status on the existing ``#send-status`` line (locked A4:
+    "Communication interrupted — retrying (n of N)…") and the send button
+    stays the Stop control; it never flips the state machine to error. It
+    is only ever sent when the failed attempt had NOT streamed a single
+    output frame yet (locked A2: no thinking/tool/delta emitted) — once
+    tokens are flowing, a failure is terminal (the ``error`` frame) and
+    this event cannot appear.
+
+    ``attempt`` is the 1-based number of the attempt the endpoint is about
+    to try next (what the endpoint sends — the first failure of a
+    4-attempt budget carries ``attempt=2``); ``max_attempts`` is the total
+    attempt budget (``llm_retries + 1``).
+    """
+
+    type: Literal["retry"] = "retry"
+    attempt: int
+    max_attempts: int
+
+
 class DocSummary(BaseModel):
     """One indexed document as shown on the Sources page / API."""