fix(llm): strip non-generation keys from gen_conf for LiteLLM providers (#15427) (#15432)

### What problem does this PR solve?

Fixes #15427.

All LiteLLM-routed chats fail with:

- Anthropic: `litellm.BadRequestError: AnthropicException -
{"type":"invalid_request_error","message":"model_type: Extra inputs are
not permitted"}`
- OpenAI: `litellm.BadRequestError: OpenAIException - Unknown parameter:
'model_type'`

This is a regression from v0.25.4.

#### Root cause

A chat assistant's `llm_setting` is forwarded to the model as
`gen_conf`. `llm_setting` can legitimately carry RAGFlow-internal
metadata such as `model_type` (the chat REST APIs in
`api/apps/restful_apis/` read it back out of `llm_setting`), so that key
ends up inside `gen_conf`.

`Base._clean_conf` (OpenAI-compatible providers) already **whitelists**
the keys it forwards, so direct-OpenAI providers were unaffected.
`LiteLLMBase._clean_conf` only dropped `max_tokens` and passed
everything else straight through to `litellm.acompletion`, which
forwarded `model_type` to the upstream provider — and Anthropic / OpenAI
reject it. Because both Claude and GPT route through LiteLLM, every chat
broke.

#### Fix

- Extract the allowed-key set into a shared `ALLOWED_GEN_CONF_KEYS`
constant and reuse it in `Base._clean_conf`.
- Apply the same whitelist in `LiteLLMBase._clean_conf`, plus the
LiteLLM-specific reasoning params (`thinking`, `reasoning_effort`,
`extra_body`) that the model-family policies inject for reasoning
models.

This covers all four LiteLLM completion paths (`async_chat`,
`async_chat_streamly`, `async_chat_with_tools`,
`async_chat_streamly_with_tools`), since they all route through
`_clean_conf`.

#### Tests

Adds `test/unit_test/rag/llm/test_clean_conf_whitelist.py` covering both
backends: `model_type` (and other stray keys) are dropped, genuine
generation params and `thinking` survive, `max_tokens` is removed, and
the whitelist invariants hold.

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
- [x] Added test cases

rag/llm/chat_model.py

@@ -61,6 +61,47 @@ ERROR_PREFIX = "**ERROR**"
 LENGTH_NOTIFICATION_CN = "······\n由于大模型的上下文窗口大小限制，回答已经被大模型截断。"
 LENGTH_NOTIFICATION_EN = "...\nThe answer is truncated by your chosen LLM due to its limitation on context length."
 
+# Generation parameters that are safe to forward to the underlying completion
+# call. `gen_conf` originates from a chat assistant's `llm_setting`, which can
+# also carry RAGFlow-internal metadata (e.g. `model_type`). Anything outside
+# this set is dropped so providers don't reject the request with errors like
+# "Extra inputs are not permitted" / "Unknown parameter: 'model_type'" (#15427).
+ALLOWED_GEN_CONF_KEYS = frozenset(
+    {
+        "temperature",
+        "max_completion_tokens",
+        "top_p",
+        "stream",
+        "stream_options",
+        "stop",
+        "n",
+        "presence_penalty",
+        "frequency_penalty",
+        "functions",
+        "function_call",
+        "logit_bias",
+        "user",
+        "response_format",
+        "seed",
+        "tools",
+        "tool_choice",
+        "logprobs",
+        "top_logprobs",
+        "extra_headers",
+    }
+)
+
+# LiteLLM additionally understands reasoning-control parameters that the
+# model-family policies may inject into `gen_conf` (e.g. `thinking` for
+# Anthropic / Kimi reasoning models, `reasoning_effort` for OpenAI o-series).
+LITELLM_ALLOWED_GEN_CONF_KEYS = ALLOWED_GEN_CONF_KEYS | frozenset(
+    {
+        "thinking",
+        "reasoning_effort",
+        "extra_body",
+    }
+)
+
 
 def _apply_model_family_policies(
     model_name: str,
@@ -159,30 +200,7 @@ class Base(ABC):
         if "max_tokens" in gen_conf:
             del gen_conf["max_tokens"]
 
-        allowed_conf = {
-            "temperature",
-            "max_completion_tokens",
-            "top_p",
-            "stream",
-            "stream_options",
-            "stop",
-            "n",
-            "presence_penalty",
-            "frequency_penalty",
-            "functions",
-            "function_call",
-            "logit_bias",
-            "user",
-            "response_format",
-            "seed",
-            "tools",
-            "tool_choice",
-            "logprobs",
-            "top_logprobs",
-            "extra_headers",
-        }
-
-        gen_conf = {k: v for k, v in gen_conf.items() if k in allowed_conf}
+        gen_conf = {k: v for k, v in gen_conf.items() if k in ALLOWED_GEN_CONF_KEYS}
         return gen_conf
 
     async def _async_chat_streamly(self, history, gen_conf, **kwargs):
@@ -1395,6 +1413,7 @@ class LiteLLMBase(ABC):
         )
 
         gen_conf.pop("max_tokens", None)
+        gen_conf = {k: v for k, v in gen_conf.items() if k in LITELLM_ALLOWED_GEN_CONF_KEYS}
         return gen_conf
 
     def _need_reasoning_content_back(self) -> bool: