fix: treat reasoning_effort="none" as thinking disabled and route gemma to Gemini provider

- Do not send reasoning_effort="none" to APIs (prevents 400 on gemma/Gemini)
- Treat "none" as thinking disabled in thinking_style, Kimi, and reasoning_content backfill paths
- Fix Anthropic extended thinking not respecting "none"
- Fix Azure OpenAI temperature suppression and reasoning body for "none"
- Fix Codex reasoning body for "none"
- Add "gemma" keyword to Gemini ProviderSpec for correct auto routing

nanobot/providers/anthropic_provider.py

@@ -434,7 +434,7 @@ class AnthropicProvider(LLMProvider):
             )
 
         max_tokens = max(1, max_tokens)
-        thinking_enabled = bool(reasoning_effort)
+        thinking_enabled = bool(reasoning_effort) and reasoning_effort.lower() != "none"
 
         # claude-opus-4-7 deprecated the `temperature` parameter entirely — the
         # API returns 400 if it is present, on any code path.

nanobot/providers/azure_openai_provider.py

@@ -71,7 +71,7 @@ class AzureOpenAIProvider(LLMProvider):
         reasoning_effort: str | None = None,
     ) -> bool:
         """Return True when temperature is likely supported for this deployment."""
-        if reasoning_effort:
+        if reasoning_effort and reasoning_effort.lower() != "none":
             return False
         name = deployment_name.lower()
         return not any(token in name for token in ("gpt-5", "o1", "o3", "o4"))
@@ -102,7 +102,7 @@ class AzureOpenAIProvider(LLMProvider):
         if self._supports_temperature(deployment, reasoning_effort):
             body["temperature"] = temperature
 
-        if reasoning_effort:
+        if reasoning_effort and reasoning_effort.lower() != "none":
             body["reasoning"] = {"effort": reasoning_effort}
             body["include"] = ["reasoning.encrypted_content"]
 

nanobot/providers/openai_codex_provider.py

@@ -60,7 +60,7 @@ class OpenAICodexProvider(LLMProvider):
             "tool_choice": tool_choice or "auto",
             "parallel_tool_calls": True,
         }
-        if reasoning_effort:
+        if reasoning_effort and reasoning_effort.lower() != "none":
             body["reasoning"] = {"effort": reasoning_effort}
         if tools:
             body["tools"] = convert_tools(tools)

nanobot/providers/openai_compat_provider.py

@@ -570,7 +570,7 @@ class OpenAICompatProvider(LLMProvider):
             # DashScope accepts none/minimum/low/medium/high/xhigh; "minimal" 400s.
             wire_effort = "minimum"
 
-        if wire_effort:
+        if wire_effort and semantic_effort != "none":
             kwargs["reasoning_effort"] = wire_effort
 
         # Provider-specific thinking parameters.
@@ -579,7 +579,7 @@ class OpenAICompatProvider(LLMProvider):
         # The mapping is driven by ProviderSpec.thinking_style so that adding
         # a new provider never requires touching this function.
         if spec and spec.thinking_style and reasoning_effort is not None:
-            thinking_enabled = semantic_effort != "minimal"
+            thinking_enabled = semantic_effort not in ("none", "minimal")
             extra = _THINKING_STYLE_MAP.get(spec.thinking_style, lambda _: None)(thinking_enabled)
             if extra:
                 kwargs.setdefault("extra_body", {}).update(extra)
@@ -589,7 +589,7 @@ class OpenAICompatProvider(LLMProvider):
         # so that OpenRouter-style names like "moonshotai/kimi-k2.5" are handled
         # identically to bare names like "kimi-k2.5".
         if reasoning_effort is not None and _is_kimi_thinking_model(model_name):
-            thinking_enabled = semantic_effort != "minimal"
+            thinking_enabled = semantic_effort not in ("none", "minimal")
             kwargs.setdefault("extra_body", {}).update(
                 {"thinking": {"type": "enabled" if thinking_enabled else "disabled"}}
             )
@@ -609,9 +609,9 @@ class OpenAICompatProvider(LLMProvider):
         # thinking happened on that turn").
         thinking_active = (
             (spec and spec.thinking_style and reasoning_effort is not None
-             and semantic_effort != "minimal")
+             and semantic_effort not in ("none", "minimal"))
             or (reasoning_effort is not None and _is_kimi_thinking_model(model_name)
-                and semantic_effort != "minimal")
+                and semantic_effort not in ("none", "minimal"))
         )
         if thinking_active:
             for msg in kwargs["messages"]:

nanobot/providers/registry.py

@@ -267,7 +267,7 @@ PROVIDERS: tuple[ProviderSpec, ...] = (
     # Gemini: Google's OpenAI-compatible endpoint
     ProviderSpec(
         name="gemini",
-        keywords=("gemini",),
+        keywords=("gemini", "gemma"),
         env_key="GEMINI_API_KEY",
         display_name="Gemini",
         backend="openai_compat",