feat: add LM Arena provider, async‑ify Copilot & surface follow‑up suggestions

* **Provider/Blackbox.py**
  * Raise `RateLimitError` when `"You have reached your request limit for the hour"` substring is detected
* **Provider/Copilot.py**
  * Convert class to `AsyncGeneratorProvider`; rename `create_completion` → `create_async_generator`
  * Swap `curl_cffi.requests.Session` for `AsyncSession`; reduce default timeout to **30 s**
  * Fully async websocket flow (`await session.ws_connect`, `await wss.send/recv/close`)
  * Emit new response types: `TitleGeneration`, `SourceLink`, aggregated `Sources`
  * Track request completion with `done` flag; collect citations in `sources` dict
* **Provider/DuckDuckGo.py**
  * Replace `duckduckgo_search.DDGS` with `duckai.DuckAI`
  * Change base class to `AbstractProvider`; drop nodriver‑based auth
* **Provider/PollinationsAI.py**
  * Re‑build text/audio model lists ensuring uniqueness; remove unused `extra_text_models`
  * Fix image seed logic (`i==1` for first retry); propagate streaming `error` field via `ResponseError`
* **Provider/hf_space**
  * **New file** `LMArenaProvider.py` implementing async queue/stream client
  * Register `LMArenaProvider` in `hf_space/__init__.py`; delete `G4F` import
* **Provider/needs_auth/CopilotAccount.py**
  * Inherit order changed to `Copilot, AsyncAuthedProvider`
  * Refactor token & cookie propagation; add `cookies_to_dict` helper
* **Provider/needs_auth/OpenaiChat.py**
  * Parse reasoning thoughts/summary; yield `Reasoning` responses
  * Tighten access‑token validation and nodriver JS evaluations (`return_by_value`)
  * Extend `Conversation` with `p` and `thoughts_summary`
* **providers/response.py**
  * Add `SourceLink` response class returning single formatted citation link
* **providers/base_provider.py**
  * Serialize `AuthResult` with custom `json.dump` to handle non‑serializable fields
  * Gracefully skip empty cache files when loading auth data
* **image/copy_images.py**
  * Ignore file extensions longer than 4 chars when inferring type
* **requests/__init__.py**
  * Use `return_by_value=True` for `navigator.userAgent` extraction
* **models.py**
  * Remove `G4F` from model provider lists; update `janus_pro_7b` best providers
* **GUI server/api.py**
  * Stream `SuggestedFollowups` to client (`"suggestions"` event)
* **GUI static assets**
  * **style.css**: bold chat title, add `.suggestions` styles, remove padding from `.chat-body`
  * **chat.v1.js**
    * Capture `suggestions` packets, render buttons, and send as quick replies
    * Re‑order finish‑reason logic; adjust token count placement and system‑prompt toggling
  * **chat-top-panel / footer** interactions updated accordingly
* **gui/client/static/js/chat.v1.js** & **css** further UI refinements (scroll handling, token counting, hide prompt toggle)
* Minor updates across multiple files to match new async interfaces and headers (`userAgent`, `raise_for_status`)

g4f/Provider/openai/har_file.py

@@ -86,8 +86,6 @@ def readHAR(request_config: RequestConfig):
                         request_config.cookies = {c['name']: c['value'] for c in v['request']['cookies']}
                     except Exception as e:
                         debug.log(f"Error on read headers: {e}")
-    if request_config.proof_token is None:
-        raise NoValidHarFileError("No proof_token found in .har files")
 
 def get_headers(entry) -> dict:
     return {h['name'].lower(): h['value'] for h in entry['request']['headers'] if h['name'].lower() not in ['content-length', 'cookie'] and not h['name'].startswith(':')}
@@ -152,8 +150,9 @@ def getN() -> str:
     return base64.b64encode(timestamp.encode()).decode()
 
 async def get_request_config(request_config: RequestConfig, proxy: str) -> RequestConfig:
-    if request_config.proof_token is None:
-        readHAR(request_config)
+    readHAR(request_config)
     if request_config.arkose_request is not None:
         request_config.arkose_token = await sendRequest(genArkReq(request_config.arkose_request), proxy)
+    if request_config.proof_token is None:
+        raise NoValidHarFileError("No proof_token found in .har files")
     return request_config