pygpt-net 2.6.45__py3-none-any.whl → 2.6.47__py3-none-any.whl

pygpt_net/provider/api/anthropic/__init__.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.09.05 01:00:00                  #
+# Updated Date: 2025.09.15 01:00:00                  #
 # ================================================== #
 
 from typing import Optional, Dict, Any
@@ -64,6 +64,8 @@ class ApiAnthropic:
         filtered = {}
         if args.get("api_key"):
             filtered["api_key"] = args["api_key"]
+        if args.get("api_proxy"):
+            filtered["api_proxy"] = args["api_proxy"]
 
         # Optionally honor custom base_url if present in config (advanced)
         # base_url = self.window.core.config.get("api_native_anthropic.base_url", "").strip()

pygpt_net/provider/api/anthropic/tools.py

@@ -174,7 +174,7 @@ class Tools:
         tools: List[dict] = []
 
         # sonnet-3.5 is not supported
-        if model and model.id and model.id.startswith("claude-3.5"):
+        if model and model.id and model.id.startswith("claude-3-5"):
             return tools
 
         # Web Search tool

pygpt_net/provider/api/google/__init__.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.09.01 23:00:00                  #
+# Updated Date: 2025.09.15 01:00:00                  #
 # ================================================== #
 
 import os
@@ -72,6 +72,12 @@ class ApiGoogle:
         filtered = {}
         if args.get("api_key"):
             filtered["api_key"] = args["api_key"]
+        if args.get("api_proxy"):
+            http_options = gtypes.HttpOptions(
+                client_args={"proxy": args["api_proxy"]},
+                async_client_args={"proxy": args["api_proxy"]},
+            )
+            filtered["http_options"] = http_options
 
         # setup VertexAI if enabled
         use_vertex = self.setup_env()

pygpt_net/provider/api/x_ai/__init__.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.09.05 01:00:00                  #
+# Updated Date: 2025.09.15 01:00:00                  #
 # ================================================== #
 
 from typing import Optional, Dict, Any
@@ -74,6 +74,7 @@ class ApiXAI:
         cfg = self.window.core.config
         api_key = cfg.get("api_key_xai") or os.environ.get("XAI_API_KEY") or ""
         timeout = cfg.get("api_native_xai.timeout")  # optional
+        proxy = cfg.get("api_proxy") or ""
 
         kwargs: Dict[str, Any] = {}
         if api_key:
@@ -81,6 +82,9 @@ class ApiXAI:
         if timeout is not None:
             # Official SDK supports setting a global timeout on client init.
             kwargs["timeout"] = timeout
+        if proxy:
+            kwargs["channel_options"] = []
+            kwargs["channel_options"].append(("grpc.http_proxy", proxy))
 
         self.client = xai_sdk.Client(**kwargs)
         return self.client

pygpt_net/provider/core/config/patch.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.09.12 23:00:00                  #
+# Updated Date: 2025.09.15 01:00:00                  #
 # ================================================== #
 
 import copy
@@ -75,6 +75,19 @@ class Patch:
                     data["render.msg.user.collapse.px"] = 1500
                 updated = True
 
+            # < 2.6.46
+            if old < parse_version("2.6.46"):
+                print("Migrating config from < 2.6.46...")
+                # output stream margin-top: 0
+                patch_css('web-chatgpt.css', True)
+                patch_css('web-chatgpt_wide.css', True)
+                patch_css('web-blocks.css', True)
+                patch_css('style.dark.css', True)
+                patch_css('web-blocks.light.css', True)
+                patch_css('web-chatgpt.light.css', True)
+                patch_css('web-chatgpt_wide.light.css', True)
+                updated = True
+
         # update file
         migrated = False
         if updated:

pygpt_net/provider/llms/anthropic.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.28 09:00:00                  #
+# Updated Date: 2025.09.15 01:00:00                  #
 # ================================================== #
 
 from typing import List, Dict, Optional
@@ -50,12 +50,35 @@ class AnthropicLLM(BaseLLM):
         :return: LLM provider instance
         """
         from llama_index.llms.anthropic import Anthropic
+        class AnthropicWithProxy(Anthropic):
+            def __init__(self, *args, proxy: str = None, **kwargs):
+                super().__init__(*args, **kwargs)
+                if not proxy:
+                    return
+
+                # sync
+                from anthropic import DefaultHttpxClient
+                self._client = self._client.with_options(
+                    http_client=DefaultHttpxClient(proxy=proxy)
+                )
+
+                # async
+                import httpx
+                try:
+                    async_http = httpx.AsyncClient(proxy=proxy)  # httpx >= 0.28
+                except TypeError:
+                    async_http = httpx.AsyncClient(proxies=proxy)  # httpx <= 0.27
+
+                self._aclient = self._aclient.with_options(http_client=async_http)
+
         args = self.parse_args(model.llama_index, window)
+        proxy = window.core.config.get("api_proxy", None)
         if "model" not in args:
             args["model"] = model.id
         if "api_key" not in args or args["api_key"] == "":
             args["api_key"] = window.core.config.get("api_key_anthropic", "")
-        return Anthropic(**args)
+
+        return AnthropicWithProxy(**args, proxy=proxy)
 
     def get_embeddings_model(
             self,
@@ -69,7 +92,7 @@ class AnthropicLLM(BaseLLM):
         :param config: config keyword arguments list
         :return: Embedding provider instance
         """
-        from llama_index.embeddings.voyageai import VoyageEmbedding
+        from .voyage import VoyageEmbeddingWithProxy
         args = {}
         if config is not None:
             args = self.parse_args({
@@ -81,7 +104,16 @@ class AnthropicLLM(BaseLLM):
             args["voyage_api_key"] = window.core.config.get("api_key_voyage", "")
         if "model" in args and "model_name" not in args:
             args["model_name"] = args.pop("model")
-        return VoyageEmbedding(**args)
+
+        timeout = window.core.config.get("api_native_voyage.timeout")
+        max_retries = window.core.config.get("api_native_voyage.max_retries")
+        proxy = window.core.config.get("api_proxy")
+        return VoyageEmbeddingWithProxy(
+            **args,
+            proxy=proxy,
+            timeout=timeout,
+            max_retries=max_retries,
+        )
 
     def get_models(
             self,
@@ -104,4 +136,4 @@ class AnthropicLLM(BaseLLM):
                     "id": item.id,
                     "name": item.id,
                 })
-        return items
+        return items

pygpt_net/provider/llms/azure_openai.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.26 19:00:00                  #
+# Updated Date: 2025.09.15 01:00:00                  #
 # ================================================== #
 
 from typing import Optional, List, Dict
@@ -94,6 +94,7 @@ class AzureOpenAILLM(BaseLLM):
             args["api_key"] = window.core.config.get("api_key", "")
         if "model" not in args:
             args["model"] = model.id
+        args = self.inject_llamaindex_http_clients(args, window.core.config)
         return LlamaAzureOpenAI(**args)
 
     def get_embeddings_model(
@@ -118,4 +119,5 @@ class AzureOpenAILLM(BaseLLM):
             args["api_key"] = window.core.config.get("api_key", "")
         if "model" in args and "model_name" not in args:
             args["model_name"] = args.pop("model")
+        args = self.inject_llamaindex_http_clients(args, window.core.config)
         return AzureOpenAIEmbedding(**args)

pygpt_net/provider/llms/base.py

@@ -6,12 +6,13 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.28 09:00:00                  #
+# Updated Date: 2025.09.15 01:00:00                  #
 # ================================================== #
 
 import os
 from typing import Optional, List, Dict
 
+import httpx
 from llama_index.core.base.embeddings.base import BaseEmbedding
 from llama_index.core.llms.llm import BaseLLM as LlamaBaseLLM
 from llama_index.core.multi_modal_llms import MultiModalLLM as LlamaMultiModalLLM
@@ -226,3 +227,14 @@ class BaseLLM:
             mode=MODE_CHAT,
             model=model,
         )
+
+    def inject_llamaindex_http_clients(self, args: dict, cfg) -> dict:
+        import httpx
+        proxy = (cfg.get("api_proxy") or "").strip()  # e.g. "http://user:pass@host:3128"
+        common_kwargs = dict(timeout=60.0, follow_redirects=True)
+        if proxy:
+            common_kwargs["proxy"] = proxy  # httpx>=0.28
+
+        args["http_client"] = httpx.Client(**common_kwargs)
+        args["async_http_client"] = httpx.AsyncClient(**common_kwargs)
+        return args

pygpt_net/provider/llms/deepseek_api.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.26 19:00:00                  #
+# Updated Date: 2025.09.15 01:00:00                  #
 # ================================================== #
 
 from typing import List, Dict, Optional
@@ -49,6 +49,7 @@ class DeepseekApiLLM(BaseLLM):
             args["model"] = model.id
         if "api_key" not in args or args["api_key"] == "":
             args["api_key"] = window.core.config.get("api_key_deepseek", "")
+        args = self.inject_llamaindex_http_clients(args, window.core.config)
         return DeepSeek(**args)
 
     def get_embeddings_model(
@@ -63,7 +64,7 @@ class DeepseekApiLLM(BaseLLM):
         :param config: config keyword arguments list
         :return: Embedding provider instance
         """
-        from llama_index.embeddings.voyageai import VoyageEmbedding
+        from .voyage import VoyageEmbeddingWithProxy
         args = {}
         if config is not None:
             args = self.parse_args({
@@ -75,7 +76,16 @@ class DeepseekApiLLM(BaseLLM):
             args["voyage_api_key"] = window.core.config.get("api_key_voyage", "")
         if "model" in args and "model_name" not in args:
             args["model_name"] = args.pop("model")
-        return VoyageEmbedding(**args)
+
+        timeout = window.core.config.get("api_native_voyage.timeout")
+        max_retries = window.core.config.get("api_native_voyage.max_retries")
+        proxy = window.core.config.get("api_proxy")
+        return VoyageEmbeddingWithProxy(
+            **args,
+            proxy=proxy,
+            timeout=timeout,
+            max_retries=max_retries,
+        )
 
     def get_models(
             self,

pygpt_net/provider/llms/google.py

@@ -6,11 +6,12 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.30 06:00:00                  #
+# Updated Date: 2025.09.15 01:00:00                  #
 # ================================================== #
 
 from typing import Optional, List, Dict
 
+from google.genai import types as gtypes
 from llama_index.core.llms.llm import BaseLLM as LlamaBaseLLM
 from llama_index.core.base.embeddings.base import BaseEmbedding
 
@@ -57,6 +58,7 @@ class GoogleLLM(BaseLLM):
             args["api_key"] = window.core.config.get("api_key_google", "")
 
         window.core.api.google.setup_env()  # setup VertexAI if configured
+        args = self.inject_llamaindex_http_clients(args, window.core.config)
         return GoogleGenAI(**args)
 
     def get_embeddings_model(
@@ -83,6 +85,7 @@ class GoogleLLM(BaseLLM):
             args["model_name"] = args.pop("model")
 
         window.core.api.google.setup_env()  # setup VertexAI if configured
+        args = self.inject_llamaindex_http_clients(args, window.core.config)
         return GoogleGenAIEmbedding(**args)
 
     def get_models(
@@ -105,3 +108,13 @@ class GoogleLLM(BaseLLM):
                 "name": id,  # TODO: token limit get from API
             })
         return items
+
+    def inject_llamaindex_http_clients(self, args: dict, cfg) -> dict:
+        proxy = cfg.get("api_proxy")
+        if proxy:
+            http_options = gtypes.HttpOptions(
+                client_args={"proxy": proxy},
+                async_client_args={"proxy": proxy},
+            )
+            args["http_options"] = http_options
+        return args

pygpt_net/provider/llms/hugging_face_api.py

@@ -36,44 +36,125 @@ class HuggingFaceApiLLM(BaseLLM):
             stream: bool = False
     ) -> LlamaBaseLLM:
         """
-        Return LLM provider instance for llama
-
-        :param window: window instance
-        :param model: model instance
-        :param stream: stream mode
-        :return: LLM provider instance
+        Return LLM provider instance for llama (Hugging Face Inference API) z obsługą proxy.
         """
-        from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
+        from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI as LIHF
+
+        class HuggingFaceInferenceAPIWithProxy(LIHF):
+            def __init__(self, *, proxy=None, trust_env=None, **kwargs):
+                """
+                proxy: str or dict compatible with huggingface_hub (e.g. "http://user:pass@host:3128"
+                       lub {"http": "...", "https": "..."})
+                trust_env: forAsyncInferenceClient (default False) - whether to trust environment variables
+                """
+                # alias: api_key -> token
+                if "api_key" in kwargs and "token" not in kwargs:
+                    kwargs["token"] = kwargs.pop("api_key")
+
+                super().__init__(**kwargs)
+
+                if proxy is None and trust_env is None:
+                    return
+
+                if isinstance(proxy, str):
+                    hf_proxies = {"http": proxy, "https": proxy}
+                else:
+                    hf_proxies = proxy
+
+                base_kwargs = {}
+                if hasattr(self, "_get_inference_client_kwargs"):
+                    try:
+                        base_kwargs = self._get_inference_client_kwargs()  # type: ignore
+                    except Exception:
+                        base_kwargs = {}
+
+                # fallback
+                if not base_kwargs:
+                    for src in ["model", "token", "timeout", "headers", "cookies", "base_url", "task"]:
+                        if hasattr(self, src):
+                            val = getattr(self, src)
+                            if val not in (None, ""):
+                                base_kwargs[src] = val
+
+                from huggingface_hub import InferenceClient, AsyncInferenceClient
+
+                sync_kwargs = dict(base_kwargs)
+                async_kwargs = dict(base_kwargs)
+                if hf_proxies is not None:
+                    sync_kwargs["proxies"] = hf_proxies
+                    async_kwargs["proxies"] = hf_proxies
+                if trust_env is not None:
+                    async_kwargs["trust_env"] = trust_env  # default False
+
+                sync_client = InferenceClient(**sync_kwargs)
+                async_client = AsyncInferenceClient(**async_kwargs)
+
+                for name, client in (("_client", sync_client), ("_sync_client", sync_client), ("client", sync_client)):
+                    if hasattr(self, name):
+                        setattr(self, name, client)
+                for name, client in (("_aclient", async_client), ("_async_client", async_client)):
+                    if hasattr(self, name):
+                        setattr(self, name, client)
+
+        cfg = window.core.config
         args = self.parse_args(model.llama_index, window)
+
+        # model
         if "model" not in args:
             args["model"] = model.id
-        if "api_key" not in args or args["api_key"] == "":
-            args["api_key"] = window.core.config.get("api_key_hugging_face", "")
-        return HuggingFaceInferenceAPI(**args)
+
+        # token / api_key
+        if "token" not in args:
+            if "api_key" in args and args["api_key"]:
+                args["token"] = args.pop("api_key")
+            else:
+                args["token"] = cfg.get("api_key_hugging_face", "")
+
+        # Inference Endpoint / router
+        base_url = cfg.get("api_endpoint_hugging_face", "").strip()
+        if base_url and "base_url" not in args:
+            args["base_url"] = base_url
+
+        # proxy + trust_env (async)
+        proxy = cfg.get("api_proxy") or cfg.get("api_native_hf.proxy")
+        trust_env = cfg.get("api_native_hf.trust_env", False)
+
+        return HuggingFaceInferenceAPIWithProxy(proxy=proxy, trust_env=trust_env, **args)
 
     def get_embeddings_model(
             self,
             window,
             config: Optional[List[Dict]] = None
     ) -> BaseEmbedding:
-        """
-        Return provider instance for embeddings
+        from .hugging_face_embedding import (
+            HuggingFaceInferenceAPIEmbeddingWithProxy as HFEmbed,
+        )
 
-        :param window: window instance
-        :param config: config keyword arguments list
-        :return: Embedding provider instance
-        """
-        from llama_index.embeddings.huggingface_api import HuggingFaceInferenceAPIEmbedding
-        args = {}
+        args: Dict = {}
         if config is not None:
-            args = self.parse_args({
-                "args": config,
-            }, window)
-        if "api_key" not in args or args["api_key"] == "":
-            args["api_key"] = window.core.config.get("api_key_hugging_face", "")
+            args = self.parse_args({"args": config}, window)
+
+        # token / api_key
+        if "token" not in args:
+            if "api_key" in args:
+                args["token"] = args.pop("api_key")
+            else:
+                args["token"] = window.core.config.get("api_key_hugging_face", "")
+
+        # model_name alias
         if "model" in args and "model_name" not in args:
             args["model_name"] = args.pop("model")
-        return HuggingFaceInferenceAPIEmbedding(**args)
+
+        # Inference Endpoint / router
+        base_url = window.core.config.get("api_endpoint_hugging_face", "").strip()
+        if base_url and "base_url" not in args:
+            args["base_url"] = base_url
+
+        # proxy + trust_env (async)
+        proxy = window.core.config.get("api_proxy") or window.core.config.get("api_native_hf.proxy")
+        trust_env = window.core.config.get("api_native_hf.trust_env", False)
+
+        return HFEmbed(proxy=proxy, trust_env=trust_env, **args)
 
     def init_embeddings(
             self,

pygpt_net/provider/llms/hugging_face_embedding.py

@@ -0,0 +1,88 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+# ================================================== #
+# This file is a part of PYGPT package               #
+# Website: https://pygpt.net                         #
+# GitHub:  https://github.com/szczyglis-dev/py-gpt   #
+# MIT License                                        #
+# Created By  : Marcin Szczygliński                  #
+# Updated Date: 2025.09.15 01:00:00                  #
+# ================================================== #
+
+from typing import Optional, Dict, Any, Union
+
+from llama_index.embeddings.huggingface_api import (
+    HuggingFaceInferenceAPIEmbedding as _HFEmbed,
+)
+
+class HuggingFaceInferenceAPIEmbeddingWithProxy(_HFEmbed):
+    def __init__(
+        self,
+        *,
+        proxy: Optional[Union[str, Dict[str, str]]] = None,
+        trust_env: Optional[bool] = None,
+        **kwargs: Any,
+    ) -> None:
+        """
+        proxy: e.g. "http://user:pass@host:3128" albo {"http": "...", "https": "..."}
+        trust_env: AsyncInferenceClient; if True, then use the system's proxy settings
+        """
+
+        if "api_key" in kwargs and "token" not in kwargs:
+            kwargs["token"] = kwargs.pop("api_key")
+        if "model" in kwargs and "model_name" not in kwargs:
+            kwargs["model_name"] = kwargs.pop("model")
+
+        super().__init__(**kwargs)
+
+        # if no proxy and no trust_env, then do nothing
+        if proxy is None and trust_env is None:
+            return
+
+        # dict proxies for huggingface_hub.*
+        hf_proxies: Optional[Union[str, Dict[str, str]]] = None
+        if proxy is not None:
+            hf_proxies = {"http": proxy, "https": proxy} if isinstance(proxy, str) else proxy
+
+        base_kwargs: Dict[str, Any] = {}
+
+        for src, dst in [
+            ("model_name", "model"),
+            ("token", "token"),
+            ("timeout", "timeout"),
+            ("headers", "headers"),
+            ("cookies", "cookies"),
+            ("base_url", "base_url"),
+        ]:
+            if hasattr(self, src):
+                val = getattr(self, src)
+                if val is not None and val != "":
+                    base_kwargs[dst] = val
+
+        from huggingface_hub import InferenceClient, AsyncInferenceClient
+
+        sync_kwargs = dict(base_kwargs)
+        async_kwargs = dict(base_kwargs)
+
+        if hf_proxies is not None:
+            sync_kwargs["proxies"] = hf_proxies
+            async_kwargs["proxies"] = hf_proxies
+        if trust_env is not None:
+            async_kwargs["trust_env"] = trust_env  # default False w AsyncInferenceClient
+
+        sync_client = InferenceClient(**sync_kwargs)
+        async_client = AsyncInferenceClient(**async_kwargs)
+
+        for name, client in (
+            ("_sync_client", sync_client),
+            ("_client", sync_client),
+        ):
+            if hasattr(self, name):
+                setattr(self, name, client)
+
+        for name, client in (
+            ("_async_client", async_client),
+            ("_aclient", async_client),
+        ):
+            if hasattr(self, name):
+                setattr(self, name, client)

pygpt_net/provider/llms/hugging_face_router.py

@@ -6,10 +6,10 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.26 19:00:00                  #
+# Updated Date: 2025.09.15 01:00:00                  #
 # ================================================== #
 
-from typing import Optional, List, Dict
+from typing import Optional, List, Dict, Union
 
 from llama_index.core.llms.llm import BaseLLM as LlamaBaseLLM
 from llama_index.core.multi_modal_llms import MultiModalLLM as LlamaMultiModalLLM
@@ -82,6 +82,7 @@ class HuggingFaceRouterLLM(BaseLLM):
             args["model"] = model.id
         if "api_key" not in args or args["api_key"] == "":
             args["api_key"] = window.core.config.get("api_key_hugging_face", "")
+        args = self.inject_llamaindex_http_clients(args, window.core.config)
         return OpenAILike(**args)
 
     def llama_multimodal(
@@ -105,24 +106,35 @@ class HuggingFaceRouterLLM(BaseLLM):
             window,
             config: Optional[List[Dict]] = None
     ) -> BaseEmbedding:
-        """
-        Return provider instance for embeddings
+        from .hugging_face_embedding import (
+            HuggingFaceInferenceAPIEmbeddingWithProxy as HFEmbed,
+        )
 
-        :param window: window instance
-        :param config: config keyword arguments list
-        :return: Embedding provider instance
-        """
-        from llama_index.embeddings.huggingface_api import HuggingFaceInferenceAPIEmbedding
-        args = {}
+        args: Dict = {}
         if config is not None:
-            args = self.parse_args({
-                "args": config,
-            }, window)
-        if "api_key" not in args or args["api_key"] == "":
-            args["api_key"] = window.core.config.get("api_key_hugging_face", "")
+            args = self.parse_args({"args": config}, window)
+
+        # token / api_key
+        if "token" not in args:
+            if "api_key" in args:
+                args["token"] = args.pop("api_key")
+            else:
+                args["token"] = window.core.config.get("api_key_hugging_face", "")
+
+        # model_name alias
         if "model" in args and "model_name" not in args:
             args["model_name"] = args.pop("model")
-        return HuggingFaceInferenceAPIEmbedding(**args)
+
+        # Inference Endpoint / router
+        base_url = window.core.config.get("api_endpoint_hugging_face", "").strip()
+        if base_url and "base_url" not in args:
+            args["base_url"] = base_url
+
+        # proxy + trust_env (async)
+        proxy = window.core.config.get("api_proxy") or window.core.config.get("api_native_hf.proxy")
+        trust_env = window.core.config.get("api_native_hf.trust_env", False)
+
+        return HFEmbed(proxy=proxy, trust_env=trust_env, **args)
 
     def get_models(
             self,

pygpt_net/provider/llms/local.py

@@ -47,6 +47,7 @@ class LocalLLM(BaseLLM):
             }, window)
         if "model" in args and "model_name" not in args:
             args["model_name"] = args.pop("model")
+        args = self.inject_llamaindex_http_clients(args, window.core.config)
         return OpenAILikeEmbedding(**args)
 
     def llama(
@@ -71,4 +72,5 @@ class LocalLLM(BaseLLM):
             args["is_chat_model"] = True
         if "is_function_calling_model" not in args:
             args["is_function_calling_model"] = model.tool_calls
+        args = self.inject_llamaindex_http_clients(args, window.core.config)
         return OpenAILike(**args)