Open WebUI

Open WebUI is an extensible, feature-rich, and user-friendly self-hosted AI platform designed to operate entirely offline.

Functions are like plugins for Open WebUI. We can create a custom Pipe Function that process inputs and generate responses by invoking Flowise Prediction API before returning results to the user. Through this, Flowise can be used in Open WebUI.

Setup

1. First, have Open WebUI up and running, you can refer to the Quickstart guide. From the left bottom, click your profile and Admin Panel

1. Open Functions tab, and add a new Function.

1. Name the Function, and add the following code:

"""
title: Flowise Integration for OpenWebUI
Requirements:
  - Flowise API URL (set via FLOWISE_API_URL)
  - Flowise API Key (set via FLOWISE_API_KEY)
"""

from pydantic import BaseModel, Field
from typing import Optional, Dict, Any, List, Union, Generator, Iterator
import requests
import json
import os


class Pipe:
    class Valves(BaseModel):
        flowise_url: str = Field(
            default=os.getenv("FLOWISE_API_URL", ""),
            description="Flowise URL",
        )
        flowise_api_key: str = Field(
            default=os.getenv("FLOWISE_API_KEY", ""),
            description="Flowise API key for authentication",
        )

    def __init__(self):
        self.type = "manifold"
        self.id = "flowise_chat"
        self.valves = self.Valves()

        # Validate required settings
        if not self.valves.flowise_url:
            print(
                "⚠️ Please set your Flowise URL using the FLOWISE_API_URL environment variable"
            )
        if not self.valves.flowise_api_key:
            print(
                "⚠️ Please set your Flowise API key using the FLOWISE_API_KEY environment variable"
            )

    def pipes(self):
        if self.valves.flowise_api_key and self.valves.flowise_url:
            try:
                headers = {
                    "Authorization": f"Bearer {self.valves.flowise_api_key}",
                    "Content-Type": "application/json",
                }

                r = requests.get(
                    f"{self.valves.flowise_url}/api/v1/chatflows?type=AGENTFLOW",
                    headers=headers,
                )
                models = r.json()
                return [
                    {
                        "id": model["id"],
                        "name": model["name"],
                    }
                    for model in models
                ]

            except Exception as e:
                return [
                    {
                        "id": "error",
                        "name": str(e),
                    },
                ]
        else:
            return [
                {
                    "id": "error",
                    "name": "API Key not provided.",
                },
            ]

    def _process_message_content(self, message: dict) -> str:
        """Process message content, handling text for now"""
        if isinstance(message.get("content"), list):
            processed_content = []
            for item in message["content"]:
                if item["type"] == "text":
                    processed_content.append(item["text"])
            return " ".join(processed_content)
        return message.get("content", "")

    def pipe(
        self, body: dict, __user__: Optional[dict] = None, __metadata__: dict = None
    ):
        try:
            stream_enabled = body.get("stream", True)
            session_id = (__metadata__ or {}).get("chat_id") or "owui-session"
            # model can be "flowise.<id>" or just "<id>"
            model_name = body.get("model", "")
            dot = model_name.find(".")
            model_id = model_name[dot + 1 :] if dot != -1 else model_name

            messages = body.get("messages") or []
            if not messages:
                raise Exception("No messages found in request body")
            question = self._process_message_content(messages[-1])

            data = {
                "question": question,
                "overrideConfig": {"sessionId": session_id},
                "streaming": stream_enabled,
            }

            headers = {
                "Authorization": f"Bearer {self.valves.flowise_api_key}",
                "Content-Type": "application/json",
                "Accept": "text/event-stream" if stream_enabled else "application/json",
            }

            url = f"{self.valves.flowise_url}/api/v1/prediction/{model_id}"
            with requests.post(
                url, json=data, headers=headers, stream=stream_enabled, timeout=60
            ) as r:
                r.raise_for_status()

                if stream_enabled:
                    # Ensure correct decoding for SSE (prevents ’ etc.)
                    r.encoding = "utf-8"

                    for raw_line in r.iter_lines(decode_unicode=True):
                        if not raw_line:
                            continue
                        line = raw_line.strip()

                        # Skip keep-alives or non-data fields
                        if not line.startswith("data:"):
                            continue

                        payload = line[5:].strip()
                        if payload in ("[DONE]", '"[DONE]"'):
                            break

                        # Flowise usually sends {"event":"token","data":"..."}
                        try:
                            obj = json.loads(payload)
                        except json.JSONDecodeError:
                            # Occasionally plain text arrives—stream it anyway
                            if payload:
                                yield payload
                            continue

                        if isinstance(obj, dict):
                            if obj.get("event") == "token":
                                token = obj.get("data") or ""
                                if token:
                                    yield token
                            else:
                                # Some versions send {"data":{"text":"..."}}
                                data_field = obj.get("data")
                                if isinstance(data_field, dict):
                                    text = data_field.get("text")
                                    if text:
                                        yield text
                    return  # end streaming

                # Non-streaming fallback
                resp = r.json()
                return (
                    resp.get("text") or (resp.get("data") or {}).get("text", "") or ""
                )

        except requests.HTTPError as http_err:
            try:
                detail = http_err.response.text[:500]
            except Exception:
                detail = ""
            return f"HTTP error from Flowise: {http_err.response.status_code} {detail}"
        except Exception as e:
            return f"Error in Flowise pipe: {e}"

1. After Function has been saved, enable it, and click the settings button to put in your Flowise URL and Flowise API Key:

1. Now when you refresh and click New Chat, you will be able to see the list of flows. You can modify the code to show:

• Only Agentflows V2: f"{self.valves.flowise_url}/api/v1/chatflows?type=AGENTFLOW"

• Only Chatflows: f"{self.valves.flowise_url}/api/v1/chatflows?type=CHATFLOW"

• Only Assistants: f"{self.valves.flowise_url}/api/v1/chatflows?type=ASSISTANT"

1. Test: