diff --git a/src/memu/app/service.py b/src/memu/app/service.py
index 4e2dea04..0002ed59 100644
--- a/src/memu/app/service.py
+++ b/src/memu/app/service.py
@@ -107,6 +107,7 @@ def _init_llm_client(self, config: LLMConfig | None = None) -> Any:
                 chat_model=cfg.chat_model,
                 embed_model=cfg.embed_model,
                 embed_batch_size=cfg.embed_batch_size,
+                proxy=cfg.proxy,
             )
         elif backend == "httpx":
             return HTTPLLMClient(
@@ -116,6 +117,7 @@ def _init_llm_client(self, config: LLMConfig | None = None) -> Any:
                 provider=cfg.provider,
                 endpoint_overrides=cfg.endpoint_overrides,
                 embed_model=cfg.embed_model,
+                proxy=cfg.proxy,
             )
         elif backend == "lazyllm_backend":
             from memu.llm.lazyllm_client import LazyLLMClient
@@ -129,6 +131,7 @@ def _init_llm_client(self, config: LLMConfig | None = None) -> Any:
                 embed_model=cfg.embed_model,
                 vlm_model=cfg.lazyllm_source.vlm_model,
                 stt_model=cfg.lazyllm_source.stt_model,
+                proxy=cfg.proxy,
             )
         else:
             msg = f"Unknown llm_client_backend '{cfg.client_backend}'"
diff --git a/src/memu/app/settings.py b/src/memu/app/settings.py
index adcb4f16..ff32d4f9 100644
--- a/src/memu/app/settings.py
+++ b/src/memu/app/settings.py
@@ -124,6 +124,10 @@ class LLMConfig(BaseModel):
         default=1,
         description="Maximum batch size for embedding API calls (used by SDK client backends).",
     )
+    proxy: str | None = Field(
+        default=None,
+        description="HTTP proxy URL for LLM requests (e.g., 'http://proxy.example.com:8080').",
+    )
 
     @model_validator(mode="after")
     def set_provider_defaults(self) -> "LLMConfig":
diff --git a/src/memu/embedding/backends/doubao.py b/src/memu/embedding/backends/doubao.py
index ae8a4ae7..c9405927 100644
--- a/src/memu/embedding/backends/doubao.py
+++ b/src/memu/embedding/backends/doubao.py
@@ -37,7 +37,7 @@ class DoubaoEmbeddingBackend(EmbeddingBackend):
 
     def build_embedding_payload(self, *, inputs: list[str], embed_model: str) -> dict[str, Any]:
         """Build payload for standard text embeddings."""
-        return {"model": embed_model, "input": inputs, "encoding_format": "float"}
+        return {"model": embed_model, "inputs": inputs, "encoding_format": "float"}
 
     def parse_embedding_response(self, data: dict[str, Any]) -> list[list[float]]:
         """Parse embedding response."""
@@ -64,7 +64,7 @@ def build_multimodal_embedding_payload(
         return {
             "model": embed_model,
             "encoding_format": encoding_format,
-            "input": [inp.to_dict() for inp in inputs],
+            "inputs": [inp.to_dict() for inp in inputs],
         }
 
     def parse_multimodal_embedding_response(self, data: dict[str, Any]) -> list[list[float]]:
diff --git a/src/memu/llm/http_client.py b/src/memu/llm/http_client.py
index ba84b05b..b601c158 100644
--- a/src/memu/llm/http_client.py
+++ b/src/memu/llm/http_client.py
@@ -90,6 +90,7 @@ def __init__(
         endpoint_overrides: dict[str, str] | None = None,
         timeout: int = 60,
         embed_model: str | None = None,
+        proxy: str | None = None,
     ):
         # Ensure base_url ends with "/" so httpx doesn't discard the path
         # component when joining with endpoint paths.
@@ -101,7 +102,9 @@ def __init__(
         self.backend = self._load_backend(self.provider)
         self.embedding_backend = self._load_embedding_backend(self.provider)
         overrides = endpoint_overrides or {}
-        raw_summary_ep = overrides.get("chat") or overrides.get("summary") or self.backend.summary_endpoint
+        raw_summary_ep = (
+            overrides.get("chat") or overrides.get("summary") or self.backend.summary_endpoint
+        )
         raw_embedding_ep = (
             overrides.get("embeddings")
             or overrides.get("embedding")
@@ -114,7 +117,7 @@ def __init__(
         self.embedding_endpoint = raw_embedding_ep.lstrip("/")
         self.timeout = timeout
         self.embed_model = embed_model or chat_model
-        self.proxy = _load_proxy()
+        self.proxy = proxy or _load_proxy()
 
     async def chat(
         self,
diff --git a/src/memu/llm/lazyllm_client.py b/src/memu/llm/lazyllm_client.py
index 8446b6a5..c94bf750 100644
--- a/src/memu/llm/lazyllm_client.py
+++ b/src/memu/llm/lazyllm_client.py
@@ -22,6 +22,7 @@ def __init__(
         vlm_model: str | None = None,
         embed_model: str | None = None,
         stt_model: str | None = None,
+        proxy: str | None = None,
     ):
         self.llm_source = llm_source or self.DEFAULT_SOURCE
         self.vlm_source = vlm_source or self.DEFAULT_SOURCE
@@ -31,6 +32,13 @@ def __init__(
         self.vlm_model = vlm_model
         self.embed_model = embed_model
         self.stt_model = stt_model
+        self.proxy = proxy
+        
+        # Set proxy for LazyLLM if provided
+        if proxy:
+            import os
+            os.environ["HTTP_PROXY"] = proxy
+            os.environ["HTTPS_PROXY"] = proxy
 
     async def _call_async(self, client: Any, *args: Any, **kwargs: Any) -> Any:
         """
diff --git a/src/memu/llm/openai_sdk.py b/src/memu/llm/openai_sdk.py
index 38c6c8bb..8ad08b83 100644
--- a/src/memu/llm/openai_sdk.py
+++ b/src/memu/llm/openai_sdk.py
@@ -3,6 +3,7 @@
 from pathlib import Path
 from typing import Any, Literal, cast
 
+import httpx
 from openai import AsyncOpenAI
 from openai.types import CreateEmbeddingResponse
 from openai.types.chat import (
@@ -28,13 +29,24 @@ def __init__(
         chat_model: str,
         embed_model: str,
         embed_batch_size: int = 1,
+        proxy: str | None = None,
     ):
         self.base_url = base_url.rstrip("/")
         self.api_key = api_key or ""
         self.chat_model = chat_model
         self.embed_model = embed_model
         self.embed_batch_size = embed_batch_size
-        self.client = AsyncOpenAI(api_key=self.api_key, base_url=self.base_url)
+        
+        # Create httpx client with proxy if provided
+        http_client = None
+        if proxy:
+            http_client = httpx.AsyncClient(proxy=proxy)
+        
+        self.client = AsyncOpenAI(
+            api_key=self.api_key, 
+            base_url=self.base_url,
+            http_client=http_client
+        )
 
     async def chat(
         self,