feat(http): add HTTP proxy support for LLM & embedding clients (#310)

Co-authored-by: ankaisen <51148505+ankaisen@users.noreply.github.com>

Author: Noel2521

src/memu/embedding/http_client.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import logging
+import os
 from collections.abc import Callable
 from typing import Literal
 
@@ -10,6 +11,11 @@
 from memu.embedding.backends.doubao import DoubaoEmbeddingBackend, DoubaoMultimodalEmbeddingInput
 from memu.embedding.backends.openai import OpenAIEmbeddingBackend
 
+
+def _load_proxy() -> str | None:
+    return os.getenv("MEMU_HTTP_PROXY") or os.getenv("HTTP_PROXY") or os.getenv("HTTPS_PROXY") or None
+
+
 logger = logging.getLogger(__name__)
 
 EMBEDDING_BACKENDS: dict[str, Callable[[], EmbeddingBackend]] = {
@@ -49,6 +55,7 @@ def __init__(
         # Strip leading "/" so httpx resolves relative to base_url
         self.embedding_endpoint = raw_embedding_ep.lstrip("/")
         self.timeout = timeout
+        self.proxy = _load_proxy()
 
     async def embed(self, inputs: list[str]) -> list[list[float]]:
         """
@@ -61,7 +68,7 @@ async def embed(self, inputs: list[str]) -> list[list[float]]:
             List of embedding vectors
         """
         payload = self.backend.build_embedding_payload(inputs=inputs, embed_model=self.embed_model)
-        async with httpx.AsyncClient(base_url=self.base_url, timeout=self.timeout) as client:
+        async with httpx.AsyncClient(base_url=self.base_url, timeout=self.timeout, proxy=self.proxy) as client:
             resp = await client.post(self.embedding_endpoint, json=payload, headers=self._headers())
             resp.raise_for_status()
             data = resp.json()
@@ -123,7 +130,7 @@ async def embed_multimodal(
         )
 
         endpoint = self.backend.multimodal_embedding_endpoint.lstrip("/")
-        async with httpx.AsyncClient(base_url=self.base_url, timeout=self.timeout) as client:
+        async with httpx.AsyncClient(base_url=self.base_url, timeout=self.timeout, proxy=self.proxy) as client:
             resp = await client.post(endpoint, json=payload, headers=self._headers())
             resp.raise_for_status()
             data = resp.json()

src/memu/llm/http_client.py

@@ -2,6 +2,7 @@
 
 import base64
 import logging
+import os
 from collections.abc import Callable
 from pathlib import Path
 from typing import Any, cast
@@ -15,6 +16,10 @@
 from memu.llm.backends.openrouter import OpenRouterLLMBackend
 
 
+def _load_proxy() -> str | None:
+    return os.getenv("MEMU_HTTP_PROXY") or os.getenv("HTTP_PROXY") or os.getenv("HTTPS_PROXY") or None
+
+
 # Minimal embedding backend support (moved from embedding module)
 class _EmbeddingBackend:
     name: str
@@ -109,6 +114,7 @@ def __init__(
         self.embedding_endpoint = raw_embedding_ep.lstrip("/")
         self.timeout = timeout
         self.embed_model = embed_model or chat_model
+        self.proxy = _load_proxy()
 
     async def chat(
         self,
@@ -145,7 +151,7 @@ async def summarize(
         payload = self.backend.build_summary_payload(
             text=text, system_prompt=system_prompt, chat_model=self.chat_model, max_tokens=max_tokens
         )
-        async with httpx.AsyncClient(base_url=self.base_url, timeout=self.timeout) as client:
+        async with httpx.AsyncClient(base_url=self.base_url, timeout=self.timeout, proxy=self.proxy) as client:
             resp = await client.post(self.summary_endpoint, json=payload, headers=self._headers())
             resp.raise_for_status()
             data = resp.json()
@@ -195,7 +201,7 @@ async def vision(
             max_tokens=max_tokens,
         )
 
-        async with httpx.AsyncClient(base_url=self.base_url, timeout=self.timeout) as client:
+        async with httpx.AsyncClient(base_url=self.base_url, timeout=self.timeout, proxy=self.proxy) as client:
             resp = await client.post(self.summary_endpoint, json=payload, headers=self._headers())
             resp.raise_for_status()
             data = resp.json()
@@ -205,7 +211,7 @@ async def vision(
     async def embed(self, inputs: list[str]) -> tuple[list[list[float]], dict[str, Any]]:
         """Create text embeddings using the provider-specific embedding API."""
         payload = self.embedding_backend.build_embedding_payload(inputs=inputs, embed_model=self.embed_model)
-        async with httpx.AsyncClient(base_url=self.base_url, timeout=self.timeout) as client:
+        async with httpx.AsyncClient(base_url=self.base_url, timeout=self.timeout, proxy=self.proxy) as client:
             resp = await client.post(self.embedding_endpoint, json=payload, headers=self._headers())
             resp.raise_for_status()
             data = resp.json()
@@ -246,7 +252,9 @@ async def transcribe(
                 if language:
                     data["language"] = language
 
-                async with httpx.AsyncClient(base_url=self.base_url, timeout=self.timeout * 3) as client:
+                async with httpx.AsyncClient(
+                    base_url=self.base_url, timeout=self.timeout * 3, proxy=self.proxy
+                ) as client:
                     resp = await client.post(
                         "/v1/audio/transcriptions",
                         files=files,