EmergentFlow How-To Guide

🖱️ Basic Navigation

• Pan: Click and drag on empty canvas space, or use middle mouse button
• Zoom: Scroll wheel or pinch gesture on trackpad
• Select Node: Click on any node
• Multi-Select: Hold Shift and click nodes, or drag a selection box
• Delete: Select node(s) and press Delete or Backspace

Every node has these components:

• Title Bar: Shows node type and name (double-click to rename)
• Input Ports (left): Receive data from other nodes - shown as colored circles
• Output Ports (right): Send data to other nodes
• Body: Contains configuration options, text fields, dropdowns
• Inspector Button (🔍): Toggle debug panel showing last inputs/outputs

When you click Run:

1. Flow starts from nodes with no inputs (Start, Text Input, etc.)
2. Each node processes its inputs and produces outputs
3. Connected downstream nodes receive data and execute
4. This continues until all paths complete

Inputs

• Prompt: The user message/question
• System: System prompt defining AI behavior (optional)
• Image: Image for vision models (optional)
• Trigger: Manual trigger signal

Options

• Provider: OpenAI, Anthropic, Google, Groq, Ollama, etc.
• Model: Specific model (GPT-4, Claude, Gemini, etc.)
• Stream: Show response as it generates
• Remember History: Maintain conversation context

Features

• Built-in chat interface with message history
• System prompt input for personality/instructions
• Automatic conversation memory
• Export chat log functionality

Inputs

• Goal: What the agent should accomplish
• Tools: JSON array of tool definitions
• Trigger: Start execution

Outputs

• Result: Final answer/output
• Trace: Step-by-step reasoning log

Capabilities

• Navigate to URLs
• Click buttons and links
• Fill out forms
• Extract text and images
• Take screenshots
• Answer questions about page content

Outputs

• Current File: Content of current file in batch
• Filename: Name of current file
• All Files: Array of all filenames
• Done: Fires when batch completes

Options

• Chunk Size: Characters per chunk (default 1000)
• Overlap: Character overlap between chunks

Conditions

• Contains: Text includes value
• Equals: Exact match
• Starts with: Text begins with value
• Ends with: Text ends with value
• Regex: Regular expression match

Outputs

• Match: Fires if condition is true
• No Match: Fires if condition is false

Example

# Available: in1, in2 (inputs)
# Return value becomes output

result = in1.upper() + " - " + str(len(in1))
return result

STT Engines

• Fast Local (Moonshine): Local processing, ~40-75MB model, English optimized
• Web Speech API: Browser-native, instant but sends to Google
• Whisper: Most accurate, multilingual, slower

Modes

• Conversation: Listen → Transcribe → Send → Repeat
• Continuous: Always listening, building transcript
• Push-to-Talk: Hold button to record

Engines

• Native: Browser's built-in TTS (fast, free)
• Kokoro AI: High-quality neural TTS (requires loading)

Options

• Voice selection
• Speed adjustment
• Pitch control
• Auto-play toggle

Personas

• Assistant, Friend, Teacher, Expert
• Coach, Therapist, Comedian, Storyteller
• Custom character with name and background

Traits

• Friendly, Professional, Humorous, Empathetic
• Analytical, Creative, Patient, Enthusiastic
• And more...

Inputs

• Ingest: Text to add to the knowledge base
• Query: Search query to find relevant content
• Trigger: Manually trigger search

Output

Returns relevant text chunks that match the query, ready to be used as context for an LLM.

Server Credits (Default)

EmergentFlow provides credits for using AI services without your own API keys. Check your balance in Settings.

Bring Your Own Key (BYOK)

For unlimited usage or specific providers, add your own API keys:

• OpenAI: For GPT-4, GPT-3.5
• Anthropic: For Claude models
• Google: For Gemini models
• Groq: For fast inference
• xAI: For Grok models
• DeepSeek: For DeepSeek models

Run AI models locally with Ollama:

1. Install Ollama on your computer
2. Run ollama serve
3. Pull models: ollama pull llama2
4. In Settings → Ollama, click "Fetch Models"
5. Select provider "Ollama" on LLM nodes

Fast STT (Speech-to-Text)

• Load the model before using Microphone nodes
• Choose tiny (~40MB) or base (~75MB) model
• Enable "Auto-load on startup" for convenience

Whisper (Alternative STT)

• Higher accuracy, slower processing
• Multilingual support

Kokoro TTS

• High-quality AI voice synthesis
• Requires loading model first

• Daily Credits: Free credits that refresh daily
• Wallet Credits: Purchased credits that don't expire
• Upgrade: Get more credits or Pro features

Basic question-answering flow:

Answer questions about your documents:

Talk to AI with your voice:

Analyze images with AI:

Automatically summarize incoming emails:

Research topics from the web:

Process multiple images:

Route based on content:

Match → Handle Error | No Match → Continue