""" title: OpenRouter Reasoning Tokens author: rmarfil3 repo: https://github.com/rmarfil3/openwebui-openrouter-reasoning-tokens date: 2025-01-30 version: 0.2 license: MIT description: Enables reasoning tokens and displays them visually in OWUI NOTE: After installing and enabling, new models will be added which will have the "Thinking..." feature. These models will be prefixed with "reasoning/". You will probably have an issue with the generated titles. You can go to Admin Panel -> Settings -> Interface, then change the Task Model to something else instead of "Current Model". UPDATES: v0.2: - Asynchronous API call to avoid blocking the UI - Fixed reasoning responses showing up as actual response """ from pydantic import BaseModel, Field import httpx import json import time class Pipe: class Valves(BaseModel): OPENROUTER_API_BASE_URL: str = Field(default="https://openrouter.ai/api/v1") OPENROUTER_API_KEY: str = Field( default="", description="Your OpenRouter API key for authentication", ) def __init__(self): self.valves = self.Valves() self.timeout = 60 def pipes(self): models = [ "qwen/qwq-32b:free", "moonshotai/moonlight-16b-a3b-instruct:free", "nousresearch/deephermes-3-llama-3-8b-preview:free", "cognitivecomputations/dolphin3.0-r1-mistral-24b:free", "cognitivecomputations/dolphin3.0-mistral-24b:free", "google/gemini-2.0-flash-lite-preview-02-05:free", "google/gemini-2.0-pro-exp-02-05:free", "qwen/qwen-vl-plus:free", "qwen/qwen2.5-vl-72b-instruct:free", "mistralai/mistral-small-24b-instruct-2501:free", "deepseek/deepseek-r1-distill-llama-70b:free", "google/gemini-2.0-flash-thinking-exp:free", "deepseek/deepseek-r1:free", "sophosympatheia/rogue-rose-103b-v0.2:free", "deepseek/deepseek-chat:free", "google/gemini-2.0-flash-thinking-exp-1219:free", "google/gemini-2.0-flash-exp:free", "google/gemini-exp-1206:free", "meta-llama/llama-3.3-70b-instruct:free", "google/learnlm-1.5-pro-experimental:free", "qwen/qwen-2.5-coder-32b-instruct:free", "nvidia/llama-3.1-nemotron-70b-instruct:free", "meta-llama/llama-3.2-1b-instruct:free", "meta-llama/llama-3.2-11b-vision-instruct:free", "meta-llama/llama-3.1-8b-instruct:free", "mistralai/mistral-nemo:free", "qwen/qwen-2-7b-instruct:free", "google/gemma-2-9b-it:free", "mistralai/mistral-7b-instruct:free", "microsoft/phi-3-mini-128k-instruct:free", "microsoft/phi-3-medium-128k-instruct:free", "meta-llama/llama-3-8b-instruct:free", "openchat/openchat-7b:free", "undi95/toppy-m-7b:free", "huggingfaceh4/zephyr-7b-beta:free", "gryphe/mythomax-l2-13b:free", ] return [ { "id": f"reasoning/{model}", "name": f"reasoning/{model}", } for model in models ] async def pipe(self, body: dict): modified_body = {**body} if "model" in modified_body: # Remove "reasoning/" prefix from model ID modified_body["model"] = ( modified_body["model"].split(".", 1)[-1].replace("reasoning/", "", 1) ) modified_body["include_reasoning"] = True headers = { "Authorization": f"Bearer {self.valves.OPENROUTER_API_KEY}", "Content-Type": "application/json", "HTTP-Referer": "https://openwebui.com/", "X-Title": "Open WebUI", } try: if body.get("stream", False): return self._handle_streaming_request(modified_body, headers) else: return self._handle_normal_request(modified_body, headers) except Exception as e: print(e) return json.dumps({"error": str(e)}) async def _handle_normal_request(self, body: dict, headers: dict): async with httpx.AsyncClient(timeout=self.timeout) as client: response = await client.post( f"{self.valves.OPENROUTER_API_BASE_URL}/chat/completions", json=body, headers=headers, ) response.raise_for_status() data = response.json() if "choices" in data: for choice in data["choices"]: if "message" in choice and "reasoning" in choice["message"]: reasoning = choice["message"]["reasoning"] choice["message"][ "content" ] = f"<think>{reasoning}</think>\n{choice['message']['content']}" return data async def _handle_streaming_request(self, body: dict, headers: dict): def construct_chunk(content: str): return f"""data: {json.dumps({ 'id': data.get('id'), 'object': 'chat.completion.chunk', 'created': int(time.time()), 'model': body.get("model"), 'choices': [{ 'index': 0, 'delta': {'content': content, 'role': 'assistant'}, 'finish_reason': None }] })}\n\n""" async with httpx.AsyncClient(timeout=self.timeout) as client: async with client.stream( "POST", f"{self.valves.OPENROUTER_API_BASE_URL}/chat/completions", json=body, headers=headers, ) as response: response.raise_for_status() # Fail fast if needed thinking_state = -1 # -1: not started, 0: thinking, 1: answered async for line in response.aiter_lines(): if not line.startswith("data: "): continue data = json.loads(line[6:]) choice = data.get("choices", [{}])[0] delta = choice.get("delta", {}) # State transitions if thinking_state == -1 and delta.get("reasoning"): thinking_state = 0 yield construct_chunk("<think>") elif ( thinking_state == 0 and not delta.get("reasoning") and delta.get("content") ): thinking_state = 1 yield construct_chunk("</think>\n\n") # Handle content output content = delta.get("reasoning") or delta.get("content", "") if content: yield construct_chunk(content) if choice.get("finish_reason"): yield "data: [DONE]\n\n" return