fix(providers): retry before falling back

nanobot/providers/base.py

@@ -7,8 +7,9 @@ import json
 import os
 import re
 from abc import ABC, abstractmethod
-from collections.abc import Awaitable, Callable
-from contextlib import suppress
+from collections.abc import Awaitable, Callable, Generator
+from contextlib import contextmanager, suppress
+from contextvars import ContextVar
 from copy import deepcopy
 from dataclasses import dataclass, field
 from datetime import datetime, timezone
@@ -25,6 +26,26 @@ DEFAULT_STREAM_IDLE_TIMEOUT_S = 90.0
 MAX_STREAM_IDLE_TIMEOUT_S = 3600.0
 RETRY_AFTER_BUFFER = 1
 
+RetryEventCallback = Callable[[str], Awaitable[None]]
+_RETRY_EXHAUSTED_CALLBACK: ContextVar[RetryEventCallback | None] = ContextVar(
+    "nanobot_retry_exhausted_callback",
+    default=None,
+)
+
+
+@contextmanager
+def retry_exhaustion_callback(callback: RetryEventCallback) -> Generator[None, None, None]:
+    """Redirect terminal retry events within one async call context.
+
+    Provider wrappers use this internal scope to defer a candidate's terminal
+    notification without changing the public retry-method signatures.
+    """
+    token = _RETRY_EXHAUSTED_CALLBACK.set(callback)
+    try:
+        yield
+    finally:
+        _RETRY_EXHAUSTED_CALLBACK.reset(token)
+
 
 def resolve_stream_idle_timeout_s(
     *,
@@ -910,12 +931,16 @@ class LLMProvider(ABC):
             kw["provider_context"] = provider_context
         if on_stream_recover and getattr(self, "supports_stream_recover_callback", False):
             kw["on_stream_recover"] = _recover_stream
+        on_retry_exhausted = _RETRY_EXHAUSTED_CALLBACK.get()
         return await self._run_with_retry(
             self._safe_chat_stream,
             kw,
             messages,
             retry_mode=retry_mode,
             on_retry_wait=on_retry_wait,
+            on_retry_exhausted=(
+                on_retry_exhausted if on_retry_exhausted is not None else on_retry_wait
+            ),
             should_retry_guard=lambda: not has_streamed_content,
             on_stream_recover=_recover_stream if on_stream_recover else None,
         )
@@ -956,12 +981,16 @@ class LLMProvider(ABC):
         )
         if provider_context is not None:
             kw["provider_context"] = provider_context
+        on_retry_exhausted = _RETRY_EXHAUSTED_CALLBACK.get()
         return await self._run_with_retry(
             self._safe_chat,
             kw,
             messages,
             retry_mode=retry_mode,
             on_retry_wait=on_retry_wait,
+            on_retry_exhausted=(
+                on_retry_exhausted if on_retry_exhausted is not None else on_retry_wait
+            ),
         )
 
     @classmethod
@@ -1067,6 +1096,7 @@ class LLMProvider(ABC):
         *,
         retry_mode: str,
         on_retry_wait: Callable[[str], Awaitable[None]] | None,
+        on_retry_exhausted: Callable[[str], Awaitable[None]] | None,
         should_retry_guard: Callable[[], bool] | None = None,
         on_stream_recover: Callable[[], Awaitable[None]] | None = None,
     ) -> LLMResponse:
@@ -1154,21 +1184,21 @@ class LLMProvider(ABC):
                     identical_error_count,
                     (response.content or "")[:120].lower(),
                 )
-                if on_retry_wait:
-                    await on_retry_wait(
+                if on_retry_exhausted:
+                    await on_retry_exhausted(
                         f"Persistent retry stopped after {identical_error_count} identical errors."
                     )
                 return response
 
             if not persistent and attempt > len(delays):
                 logger.warning(
-                    "LLM request failed after {} retries, giving up: {}",
+                    "LLM request failed after {} attempts, giving up: {}",
                     attempt,
                     (response.content or "")[:120].lower(),
                 )
-                if on_retry_wait:
-                    await on_retry_wait(
-                        f"Model request failed after {attempt} retries, giving up."
+                if on_retry_exhausted:
+                    await on_retry_exhausted(
+                        f"Model request failed after {attempt} attempts, giving up."
                     )
                 break
 

nanobot/providers/fallback_provider.py

@@ -17,6 +17,7 @@ from nanobot.providers.base import (
     LLMResponse,
     ProviderCallContext,
     ProviderConversationState,
+    retry_exhaustion_callback,
 )
 
 # Circuit breaker tuned to match OpenAICompatProvider's Responses API breaker.
@@ -91,6 +92,7 @@ _FALLBACK_ERROR_TOKENS = (
 
 
 FallbackModelObserver = Callable[[str], Awaitable[None]]
+_UNSET = object()
 
 
 class FallbackProvider(LLMProvider):
@@ -105,6 +107,7 @@ class FallbackProvider(LLMProvider):
 
     Key design:
     - Failover is request-scoped (the wrapper itself is stateless between turns).
+    - Retrying entry points exhaust one provider's retry policy before failover.
     - Skipped when content was already streamed to avoid duplicate output,
       except timeout recovery can resume in a new stream segment.
     - Recursive failover is prevented by the factory returning plain providers.
@@ -193,6 +196,47 @@ class FallbackProvider(LLMProvider):
             lambda p, kw: p.chat(**kw), kwargs, has_streamed=None
         )
 
+    async def chat_with_retry(
+        self,
+        messages: list[dict[str, Any]],
+        tools: list[dict[str, Any]] | None = None,
+        model: str | None = None,
+        max_tokens: object = _UNSET,
+        temperature: object = _UNSET,
+        reasoning_effort: object = _UNSET,
+        tool_choice: str | dict[str, Any] | None = None,
+        retry_mode: str = "standard",
+        on_retry_wait: Callable[[str], Awaitable[None]] | None = None,
+        provider_context: ProviderCallContext | None = None,
+    ) -> LLMResponse:
+        """Exhaust each provider's retries before moving to the next fallback."""
+        call_kwargs: dict[str, Any] = {
+            "messages": messages,
+            "tools": tools,
+            "model": model,
+            "tool_choice": tool_choice,
+            "retry_mode": retry_mode,
+            "on_retry_wait": on_retry_wait,
+        }
+        if max_tokens is not _UNSET:
+            call_kwargs["max_tokens"] = max_tokens
+        if temperature is not _UNSET:
+            call_kwargs["temperature"] = temperature
+        if reasoning_effort is not _UNSET:
+            call_kwargs["reasoning_effort"] = reasoning_effort
+        if provider_context is not None:
+            call_kwargs["provider_context"] = self._primary_call_context(
+                provider_context,
+                model,
+            )
+        if not self._has_fallbacks:
+            return await self._primary.chat_with_retry(**call_kwargs)
+        return await self._route_with_retry_fallback(
+            lambda p, kw: p.chat_with_retry(**kw),
+            call_kwargs,
+            has_streamed=None,
+        )
+
     async def chat_with_context(
         self,
         *,
@@ -234,6 +278,154 @@ class FallbackProvider(LLMProvider):
             on_stream_recover=on_stream_recover,
         )
 
+    async def chat_stream_with_retry(
+        self,
+        messages: list[dict[str, Any]],
+        tools: list[dict[str, Any]] | None = None,
+        model: str | None = None,
+        max_tokens: object = _UNSET,
+        temperature: object = _UNSET,
+        reasoning_effort: object = _UNSET,
+        tool_choice: str | dict[str, Any] | None = None,
+        on_content_delta: Callable[[str], Awaitable[None]] | None = None,
+        on_thinking_delta: Callable[[str], Awaitable[None]] | None = None,
+        on_tool_call_delta: Callable[[dict[str, Any]], Awaitable[None]] | None = None,
+        on_stream_recover: Callable[[], Awaitable[None]] | None = None,
+        retry_mode: str = "standard",
+        on_retry_wait: Callable[[str], Awaitable[None]] | None = None,
+        provider_context: ProviderCallContext | None = None,
+    ) -> LLMResponse:
+        """Exhaust streaming retries on one provider before failing over."""
+        call_kwargs: dict[str, Any] = {
+            "messages": messages,
+            "tools": tools,
+            "model": model,
+            "tool_choice": tool_choice,
+            "on_content_delta": on_content_delta,
+            "on_thinking_delta": on_thinking_delta,
+            "on_tool_call_delta": on_tool_call_delta,
+            "retry_mode": retry_mode,
+            "on_retry_wait": on_retry_wait,
+        }
+        if max_tokens is not _UNSET:
+            call_kwargs["max_tokens"] = max_tokens
+        if temperature is not _UNSET:
+            call_kwargs["temperature"] = temperature
+        if reasoning_effort is not _UNSET:
+            call_kwargs["reasoning_effort"] = reasoning_effort
+        if provider_context is not None:
+            call_kwargs["provider_context"] = self._primary_call_context(
+                provider_context,
+                model,
+            )
+        if not self._has_fallbacks:
+            if on_stream_recover is not None:
+                call_kwargs["on_stream_recover"] = on_stream_recover
+            return await self._primary.chat_stream_with_retry(**call_kwargs)
+
+        has_streamed: list[bool] = [False]
+        has_unrecovered_stream: list[bool] = [False]
+        original_delta = call_kwargs.get("on_content_delta")
+
+        async def _tracking_delta(text: str) -> None:
+            if text:
+                has_streamed[0] = True
+                has_unrecovered_stream[0] = True
+            if original_delta:
+                await original_delta(text)
+
+        async def _recover_stream() -> None:
+            has_streamed[0] = False
+            has_unrecovered_stream[0] = False
+            if on_stream_recover:
+                await on_stream_recover()
+
+        if original_delta is not None:
+            call_kwargs["on_content_delta"] = _tracking_delta
+        if on_stream_recover is not None:
+            call_kwargs["on_stream_recover"] = _recover_stream
+        return await self._route_with_retry_fallback(
+            lambda p, kw: p.chat_stream_with_retry(**kw),
+            call_kwargs,
+            has_streamed=has_streamed,
+            on_stream_recover=_recover_stream if on_stream_recover is not None else None,
+            persistent_retry_guard=lambda: not has_unrecovered_stream[0],
+        )
+
+    async def _route_with_retry_fallback(
+        self,
+        call: Callable[[LLMProvider, dict[str, Any]], Awaitable[LLMResponse]],
+        kwargs: dict[str, Any],
+        has_streamed: list[bool] | None,
+        on_stream_recover: Callable[[], Awaitable[None]] | None = None,
+        persistent_retry_guard: Callable[[], bool] | None = None,
+    ) -> LLMResponse:
+        """Apply finite retries per provider and persistence to the whole chain."""
+        on_retry_wait: Callable[[str], Awaitable[None]] | None = kwargs.get("on_retry_wait")
+        if kwargs.get("retry_mode", "standard") != "persistent":
+            return await self._try_with_retry_fallback(
+                call,
+                kwargs,
+                has_streamed=has_streamed,
+                on_stream_recover=on_stream_recover,
+                on_retry_exhausted=on_retry_wait,
+            )
+
+        async def _call_chain(**chain_kwargs: Any) -> LLMResponse:
+            chain_kwargs["retry_mode"] = "standard"
+            return await self._try_with_retry_fallback(
+                call,
+                chain_kwargs,
+                has_streamed=has_streamed,
+                on_stream_recover=on_stream_recover,
+                on_retry_exhausted=None,
+            )
+
+        return await self._run_with_retry(
+            _call_chain,
+            dict(kwargs),
+            kwargs["messages"],
+            retry_mode="persistent",
+            on_retry_wait=on_retry_wait,
+            on_retry_exhausted=on_retry_wait,
+            should_retry_guard=persistent_retry_guard,
+            on_stream_recover=on_stream_recover,
+        )
+
+    async def _try_with_retry_fallback(
+        self,
+        call: Callable[[LLMProvider, dict[str, Any]], Awaitable[LLMResponse]],
+        kwargs: dict[str, Any],
+        has_streamed: list[bool] | None,
+        on_stream_recover: Callable[[], Awaitable[None]] | None,
+        on_retry_exhausted: Callable[[str], Awaitable[None]] | None,
+    ) -> LLMResponse:
+        """Defer a provider's terminal retry event until the chain fails."""
+        last_exhausted_message: str | None = None
+
+        async def _capture_exhaustion(message: str) -> None:
+            nonlocal last_exhausted_message
+            last_exhausted_message = message
+
+        async def _call_with_deferred_exhaustion(
+            provider: LLMProvider,
+            call_kwargs: dict[str, Any],
+        ) -> LLMResponse:
+            nonlocal last_exhausted_message
+            last_exhausted_message = None
+            with retry_exhaustion_callback(_capture_exhaustion):
+                return await call(provider, call_kwargs)
+
+        response = await self._try_with_fallback(
+            _call_with_deferred_exhaustion,
+            kwargs,
+            has_streamed=has_streamed,
+            on_stream_recover=on_stream_recover,
+        )
+        if response.finish_reason == "error" and last_exhausted_message and on_retry_exhausted:
+            await on_retry_exhausted(last_exhausted_message)
+        return response
+
     async def chat_stream_with_context(
         self,
         *,
@@ -275,6 +467,7 @@ class FallbackProvider(LLMProvider):
     ) -> LLMResponse:
         primary_model = kwargs.get("model") or self._primary.get_default_model()
         primary_was_attempted = False
+        primary_response: LLMResponse | None = None
         primary_error = "unknown error"
         # A primary error eligible for failover did not return a replacement
         # continuation, so the incoming primary state remains reusable.
@@ -287,6 +480,7 @@ class FallbackProvider(LLMProvider):
                 self._primary_failures = 0
                 self._primary_tripped_at = None
                 return response
+            primary_response = response
             primary_error = (response.content or primary_error)[:120]
 
             if has_streamed is not None and has_streamed[0]:
@@ -326,7 +520,7 @@ class FallbackProvider(LLMProvider):
         else:
             logger.debug("Primary model '{}' circuit open; skipping", primary_model)
 
-        last_response: LLMResponse | None = None
+        last_response = primary_response
         primary_skipped = not primary_was_attempted
         for idx, fallback in enumerate(self._fallback_presets):
             fallback_model = fallback.model
@@ -423,11 +617,22 @@ class FallbackProvider(LLMProvider):
                 last_response,
                 preserve_provider_state_on_error=preserve_primary_state,
             )
-        # Primary was tripped and we have no fallbacks — synthesize an error.
+        # Primary was skipped and no fallback returned a response. Keep the result
+        # transient until the primary circuit is eligible for another probe.
+        retry_after_s = (
+            max(
+                0.1,
+                _PRIMARY_COOLDOWN_S - (time.monotonic() - self._primary_tripped_at),
+            )
+            if self._primary_tripped_at is not None
+            else None
+        )
         return LLMResponse(
             content=f"Primary model '{primary_model}' circuit open and no fallbacks available",
             finish_reason="error",
             preserve_provider_state_on_error=preserve_primary_state,
+            error_retry_after_s=retry_after_s,
+            error_should_retry=True,
         )
 
     async def _notify_fallback_model(self, model: str) -> None:

tests/agent/test_runner_fallback.py

@@ -3,7 +3,7 @@
 from __future__ import annotations
 
 from typing import Any
-from unittest.mock import MagicMock, patch
+from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
 from loguru import logger
@@ -743,6 +743,565 @@ class TestFailoverOnTransientError:
         factory.assert_called_once_with(_fallback("fallback-a"))
 
 
+class TestRetryBeforeFailover:
+    @pytest.mark.asyncio
+    async def test_retry_entrypoints_accept_positional_messages(self) -> None:
+        primary = _FakeProvider("primary", _make_response("primary ok"))
+        factory = MagicMock()
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+        messages = [{"role": "user", "content": "hi"}]
+        provider_context = ProviderCallContext()
+
+        chat_result = await fb.chat_with_retry(
+            messages,
+            None,
+            None,
+            4096,
+            0.7,
+            None,
+            None,
+            "standard",
+            None,
+            provider_context,
+        )
+        stream_result = await fb.chat_stream_with_retry(messages)
+
+        assert chat_result.content == "primary ok"
+        assert stream_result.content == "primary ok"
+        assert primary.context_calls[0] is not None
+        factory.assert_not_called()
+
+    @pytest.mark.asyncio
+    async def test_primary_retries_and_recovers_before_fallback(self) -> None:
+        primary = _FakeProvider("primary")
+        fallback = _FakeProvider("fallback", _make_response("fallback ok"))
+        factory = MagicMock(return_value=fallback)
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+
+        with (
+            patch.object(
+                primary,
+                "chat",
+                new_callable=AsyncMock,
+                side_effect=[_error_response("rate limited"), _make_response("primary ok")],
+            ) as primary_chat,
+            patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock) as sleep,
+        ):
+            result = await fb.chat_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+            )
+
+        assert result.content == "primary ok"
+        assert primary_chat.await_count == 2
+        sleep.assert_awaited_once_with(1)
+        factory.assert_not_called()
+
+    @pytest.mark.asyncio
+    async def test_primary_exhausts_retries_before_fallback(self) -> None:
+        primary = _FakeProvider("primary")
+        fallback = _FakeProvider("fallback", _make_response("fallback ok"))
+        factory = MagicMock(return_value=fallback)
+        retry_events: list[str] = []
+
+        async def _on_retry_event(message: str) -> None:
+            retry_events.append(message)
+
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+
+        with (
+            patch.object(
+                primary,
+                "chat",
+                new_callable=AsyncMock,
+                side_effect=[
+                    _make_response(
+                        f"attempt {attempt}",
+                        finish_reason="error",
+                        error_status_code=503,
+                    )
+                    for attempt in range(4)
+                ],
+            ) as primary_chat,
+            patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock) as sleep,
+        ):
+            result = await fb.chat_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+                on_retry_wait=_on_retry_event,
+            )
+
+        assert result.content == "fallback ok"
+        assert primary_chat.await_count == 4
+        assert [call.args[0] for call in sleep.await_args_list] == [1, 2, 4]
+        assert not any("giving up" in event for event in retry_events)
+        factory.assert_called_once_with(_fallback("fallback-a"))
+
+    @pytest.mark.asyncio
+    async def test_all_models_exhaust_retries_emits_one_terminal_event(self) -> None:
+        primary = _FakeProvider(
+            "primary",
+            _make_response("primary unavailable", finish_reason="error", error_status_code=503),
+        )
+        fallback = _FakeProvider(
+            "fallback",
+            _make_response("fallback unavailable", finish_reason="error", error_status_code=503),
+        )
+        retry_events: list[str] = []
+
+        async def _on_retry_event(message: str) -> None:
+            retry_events.append(message)
+
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=MagicMock(return_value=fallback),
+        )
+
+        with patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock):
+            result = await fb.chat_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+                on_retry_wait=_on_retry_event,
+            )
+
+        terminal_events = [event for event in retry_events if "giving up" in event]
+        assert result.finish_reason == "error"
+        assert len(primary.chat_calls) == 4
+        assert len(fallback.chat_calls) == 4
+        assert terminal_events == ["Model request failed after 4 attempts, giving up."]
+
+    @pytest.mark.asyncio
+    async def test_persistent_mode_retries_each_candidate_before_repeating_chain(self) -> None:
+        primary = _FakeProvider("primary")
+        fallback = _FakeProvider("fallback", _make_response("fallback ok"))
+        factory = MagicMock(return_value=fallback)
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+
+        with (
+            patch.object(
+                primary,
+                "chat",
+                new_callable=AsyncMock,
+                side_effect=[
+                    _error_response(f"server_error request-{attempt}")
+                    for attempt in range(4)
+                ],
+            ) as primary_chat,
+            patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock) as sleep,
+        ):
+            result = await fb.chat_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+                retry_mode="persistent",
+            )
+
+        assert result.content == "fallback ok"
+        assert primary_chat.await_count == 4
+        assert [call.args[0] for call in sleep.await_args_list] == [1, 2, 4]
+        factory.assert_called_once_with(_fallback("fallback-a"))
+
+    @pytest.mark.asyncio
+    async def test_persistent_mode_repeats_the_whole_fallback_chain(self) -> None:
+        primary = _FakeProvider(
+            "primary",
+            _make_response("primary unavailable", finish_reason="error", error_status_code=503),
+        )
+        fallback = _FakeProvider(
+            "fallback",
+            _make_response("fallback unavailable", finish_reason="error", error_status_code=503),
+        )
+        factory = MagicMock(return_value=fallback)
+        retry_events: list[str] = []
+
+        async def _on_retry_event(message: str) -> None:
+            retry_events.append(message)
+
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+        fb._PERSISTENT_IDENTICAL_ERROR_LIMIT = 2
+
+        with patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock):
+            result = await fb.chat_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+                retry_mode="persistent",
+                on_retry_wait=_on_retry_event,
+            )
+
+        terminal_events = [
+            event
+            for event in retry_events
+            if "giving up" in event or "Persistent retry stopped" in event
+        ]
+        assert result.finish_reason == "error"
+        assert len(primary.chat_calls) == 8
+        assert len(fallback.chat_calls) == 8
+        assert factory.call_count == 2
+        assert terminal_events == ["Persistent retry stopped after 2 identical errors."]
+
+    @pytest.mark.asyncio
+    async def test_persistent_mode_repeats_when_fallback_construction_fails(self) -> None:
+        primary = _FakeProvider(
+            "primary",
+            _make_response(
+                "primary unavailable",
+                finish_reason="error",
+                error_status_code=503,
+            ),
+        )
+        factory = MagicMock(side_effect=ValueError("missing fallback credentials"))
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+        fb._PERSISTENT_IDENTICAL_ERROR_LIMIT = 2
+
+        with patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock):
+            result = await fb.chat_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+                retry_mode="persistent",
+            )
+
+        assert result.content == "primary unavailable"
+        assert result.error_status_code == 503
+        assert len(primary.chat_calls) == 8
+        assert factory.call_count == 2
+
+    @pytest.mark.asyncio
+    async def test_persistent_mode_waits_for_open_primary_when_fallback_construction_fails(
+        self,
+    ) -> None:
+        primary = _FakeProvider("primary")
+        factory = MagicMock(side_effect=ValueError("missing fallback credentials"))
+        retry_events: list[str] = []
+
+        async def _on_retry_event(message: str) -> None:
+            retry_events.append(message)
+
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+        fb._primary_tripped_at = 100.0
+
+        with (
+            patch(
+                "nanobot.providers.fallback_provider.time.monotonic",
+                return_value=100.0,
+            ),
+            patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock),
+        ):
+            result = await fb.chat_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+                retry_mode="persistent",
+                on_retry_wait=_on_retry_event,
+            )
+
+        terminal_events = [
+            event for event in retry_events if "Persistent retry stopped" in event
+        ]
+        assert result.finish_reason == "error"
+        assert result.error_should_retry is True
+        assert result.error_retry_after_s == 60
+        assert primary.chat_calls == []
+        assert factory.call_count == fb._PERSISTENT_IDENTICAL_ERROR_LIMIT
+        assert terminal_events == [
+            f"Persistent retry stopped after {fb._PERSISTENT_IDENTICAL_ERROR_LIMIT} "
+            "identical errors."
+        ]
+
+    @pytest.mark.asyncio
+    async def test_legacy_retry_override_does_not_receive_internal_callback(self) -> None:
+        class LegacyRetryProvider(_FakeProvider):
+            def __init__(self) -> None:
+                super().__init__("legacy", _make_response("legacy ok"))
+                self.retry_calls = 0
+
+            async def chat_with_retry(
+                self,
+                messages: list[dict[str, Any]],
+                tools: list[dict[str, Any]] | None = None,
+                model: str | None = None,
+                max_tokens: object = None,
+                temperature: object = None,
+                reasoning_effort: object = None,
+                tool_choice: str | dict[str, Any] | None = None,
+                retry_mode: str = "standard",
+                on_retry_wait: Any = None,
+                provider_context: ProviderCallContext | None = None,
+            ) -> LLMResponse:
+                self.retry_calls += 1
+                return await self.chat(messages=messages)
+
+        primary = LegacyRetryProvider()
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=MagicMock(),
+        )
+
+        result = await fb.chat_with_retry(messages=[{"role": "user", "content": "hi"}])
+
+        assert result.content == "legacy ok"
+        assert primary.retry_calls == 1
+
+    @pytest.mark.asyncio
+    async def test_fallback_retries_before_trying_next_model(self) -> None:
+        primary = _FakeProvider(
+            "primary",
+            _make_response(
+                "unauthorized",
+                finish_reason="error",
+                error_status_code=401,
+                error_kind="authentication",
+                error_should_retry=False,
+            ),
+        )
+        fallback_a = _FakeProvider("fallback-a")
+        fallback_b = _FakeProvider("fallback-b", _make_response("fallback b ok"))
+        factory = MagicMock(side_effect=[fallback_a, fallback_b])
+        fallback_a_preset = _fallback("fallback-a")
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[fallback_a_preset, _fallback("fallback-b")],
+            provider_factory=factory,
+        )
+
+        with (
+            patch.object(
+                fallback_a,
+                "chat",
+                new_callable=AsyncMock,
+                side_effect=[_error_response("rate limited"), _make_response("fallback a ok")],
+            ) as fallback_chat,
+            patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock),
+        ):
+            result = await fb.chat_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+            )
+
+        assert result.content == "fallback a ok"
+        assert fallback_chat.await_count == 2
+        factory.assert_called_once_with(fallback_a_preset)
+
+    @pytest.mark.asyncio
+    async def test_streaming_primary_retries_before_fallback(self) -> None:
+        primary = _FakeProvider("primary")
+        fallback = _FakeProvider("fallback", _make_response("fallback ok"))
+        factory = MagicMock(return_value=fallback)
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+        responses = iter([
+            _make_response("partial", finish_reason="error", error_kind="timeout"),
+            _make_response("primary ok"),
+        ])
+        streamed: list[str] = []
+        recoveries: list[str] = []
+
+        async def _stream(**kwargs: Any) -> LLMResponse:
+            response = next(responses)
+            if callback := kwargs.get("on_content_delta"):
+                await callback(response.content)
+            return response
+
+        async def _delta(text: str) -> None:
+            streamed.append(text)
+
+        async def _recover() -> None:
+            recoveries.append("recover")
+
+        with (
+            patch.object(
+                primary,
+                "chat_stream",
+                new_callable=AsyncMock,
+                side_effect=_stream,
+            ) as primary_stream,
+            patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock),
+        ):
+            result = await fb.chat_stream_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+                on_content_delta=_delta,
+                on_stream_recover=_recover,
+            )
+
+        assert result.content == "primary ok"
+        assert primary_stream.await_count == 2
+        assert streamed == ["partial", "primary ok"]
+        assert recoveries == ["recover"]
+        factory.assert_not_called()
+
+    @pytest.mark.asyncio
+    async def test_streaming_primary_exhausts_retries_before_fallback(self) -> None:
+        primary = _FakeProvider("primary")
+        fallback = _FakeProvider("fallback", _make_response("fallback ok"))
+        factory = MagicMock(return_value=fallback)
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+        streamed: list[str] = []
+
+        async def _delta(text: str) -> None:
+            streamed.append(text)
+
+        with (
+            patch.object(
+                primary,
+                "chat_stream",
+                new_callable=AsyncMock,
+                side_effect=[_error_response(f"server_error {attempt}") for attempt in range(4)],
+            ) as primary_stream,
+            patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock),
+        ):
+            result = await fb.chat_stream_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+                on_content_delta=_delta,
+            )
+
+        assert result.content == "fallback ok"
+        assert primary_stream.await_count == 4
+        assert streamed == ["fallback ok"]
+        factory.assert_called_once_with(_fallback("fallback-a"))
+
+    @pytest.mark.asyncio
+    async def test_streaming_without_delta_callback_still_retries_and_falls_back(self) -> None:
+        primary = _FakeProvider("primary")
+        fallback = _FakeProvider("fallback", _make_response("fallback ok"))
+        factory = MagicMock(return_value=fallback)
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+        responses = iter([_error_response(f"server_error {attempt}") for attempt in range(4)])
+
+        async def _stream(**kwargs: Any) -> LLMResponse:
+            response = next(responses)
+            if callback := kwargs.get("on_content_delta"):
+                await callback("provider-only delta")
+            return response
+
+        with (
+            patch.object(
+                primary,
+                "chat_stream",
+                new_callable=AsyncMock,
+                side_effect=_stream,
+            ) as primary_stream,
+            patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock),
+        ):
+            result = await fb.chat_stream_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+            )
+
+        assert result.content == "fallback ok"
+        assert primary_stream.await_count == 4
+        factory.assert_called_once_with(_fallback("fallback-a"))
+
+    @pytest.mark.asyncio
+    async def test_streaming_recovery_keeps_fallback_eligible(self) -> None:
+        primary = _FakeProvider("primary")
+        fallback = _FakeProvider("fallback", _make_response("fallback ok"))
+        factory = MagicMock(return_value=fallback)
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+        responses = iter([
+            _make_response("partial", finish_reason="error", error_kind="timeout"),
+            *[_error_response(f"server_error {attempt}") for attempt in range(3)],
+        ])
+        streamed: list[str] = []
+        recoveries: list[str] = []
+
+        async def _stream(**kwargs: Any) -> LLMResponse:
+            response = next(responses)
+            if response.content == "partial" and (callback := kwargs.get("on_content_delta")):
+                await callback(response.content)
+            return response
+
+        async def _delta(text: str) -> None:
+            streamed.append(text)
+
+        async def _recover() -> None:
+            recoveries.append("recover")
+
+        with (
+            patch.object(primary, "chat_stream", new_callable=AsyncMock, side_effect=_stream),
+            patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock),
+        ):
+            result = await fb.chat_stream_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+                on_content_delta=_delta,
+                on_stream_recover=_recover,
+            )
+
+        assert result.content == "fallback ok"
+        assert streamed == ["partial", "fallback ok"]
+        assert recoveries == ["recover"]
+        factory.assert_called_once_with(_fallback("fallback-a"))
+
+    @pytest.mark.asyncio
+    async def test_unrecovered_stream_keeps_non_timeout_fallback_blocked(self) -> None:
+        primary = _FakeProvider("primary")
+        factory = MagicMock()
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+        responses = iter([
+            _make_response("partial", finish_reason="error", error_kind="timeout"),
+            *[_error_response(f"server_error {attempt}") for attempt in range(3)],
+        ])
+        streamed: list[str] = []
+
+        async def _stream(**kwargs: Any) -> LLMResponse:
+            response = next(responses)
+            if response.content == "partial" and (callback := kwargs.get("on_content_delta")):
+                await callback(response.content)
+            return response
+
+        async def _delta(text: str) -> None:
+            streamed.append(text)
+
+        with (
+            patch.object(primary, "chat_stream", new_callable=AsyncMock, side_effect=_stream),
+            patch("nanobot.providers.base.asyncio.sleep", new_callable=AsyncMock),
+        ):
+            result = await fb.chat_stream_with_retry(
+                messages=[{"role": "user", "content": "hi"}],
+                on_content_delta=_delta,
+            )
+
+        assert result.content == "server_error 2"
+        assert streamed == ["partial"]
+        factory.assert_not_called()
+
+
 class TestFailoverOnArrearageError:
     @pytest.mark.asyncio
     async def test_non_retryable_quota_tries_configured_fallback(self) -> None:

tests/providers/test_provider_retry.py

@@ -129,7 +129,7 @@ async def test_chat_with_retry_emits_terminal_progress_when_standard_retries_exh
     )
 
     assert response.content == "503 final server error"
-    assert progress[-1] == "Model request failed after 4 retries, giving up."
+    assert progress[-1] == "Model request failed after 4 attempts, giving up."
 
 
 @pytest.mark.asyncio