Smart Chat

GPT AI Flow's chat features cover two levels: a lightweight standard chat for quick starts, and AI Smart Conversations that can directly operate on your files. You can switch between them at any time depending on your task.

Standard Chat Mode

Standard chat is the fastest way to get started. Open it and start talking — no extra tools in the way.

It comes in two behaviors. The only difference: whether the AI remembers your previous turns.

History Chat

The default behavior. Every turn carries forward the full conversation context. The AI knows what came before.

Use for: deep discussions, multi-step reasoning, writing tasks that need continuity.

Single Chat

Every message starts fresh. The AI has no memory of the previous exchange.

Use for: one-off queries, quick experiments, when you don't want old context polluting your prompt.

Default Tools

Standard chat mode ships with no extra tools enabled by default. It runs the lightest agents-team (planner + implementer + reviewer), keeping only basic file operations and search.

If you need the AI to read PDFs, recognize text in images, or search the web, you can enable those in settings.

When to Use Standard Chat

Everyday Q&A, translation, summarization
Testing a prompt idea quickly
When you're not sure which mode to use — start here

AI Smart Conversations — More Than Chat, It Takes Action

Ordinary AI chat tools can only give you text replies. GPT AI Flow is different — the AI doesn't just answer your questions, it can directly operate on the files on your computer.

This corresponds to the execution layer of LLM-Wiki: you give the goal, and the AI handles retrieval, organization, and writing to disk. If you haven't read the core concepts yet, we recommend starting with the LLM-Wiki concept page.

What You Can Actually Do

These aren't hypotheticals — these are things you can do every day:

"Summarize the PDF on my desktop" → AI reads the PDF directly and extracts the key points
"Organize this article's outline into outline.md" → AI creates the file and writes the content automatically
"Search for the latest AI industry trends" → AI runs a web search and compiles the results
"What does this screenshot say?" → AI automatically extracts text from the image (OCR)
"Find my old notes about procrastination" → AI performs semantic search across your note vault
"How many articles are in this folder? Give me a word count." → AI traverses the directory and generates stats

Six Built-in Tool Categories — AI Calls Them Automatically

You don't need to pick tools manually — the AI decides on its own when to use what. Just describe what you need in plain language.

📁 File Operations

AI can read, create, edit, delete, and search files in your workspace. In the Obsidian context, this is typically your Vault.

Operation	Description
Read file	Read the contents of any text file
Create file	Create a new file at a specified location
Edit file	Overwrite or partially modify file content
Delete/Rename	Clean up or reorganize files
Search content	Search by keyword across the workspace
List directory	View folder structure
Git versioning	Every edit auto-saves a history, so you can review or roll back anytime

Safety note: All file operations are restricted to your chosen workspace. For Obsidian users, this is the Vault you specify — the AI cannot access files outside it.

🔍 Semantic Note Search

AI can search your note vault semantically, finding notes related to your current topic — even if you used completely different wording.

Search vault — Search your Vault using natural language
Discover related notes — Recommend "second-circle" notes based on current content
Surface forgotten notes — Find older, high-quality notes you haven't touched in a while

tip

Want more details on semantic search? See the Semantic Note Search deep dive.

🌐 Web Search

AI can search the internet for up-to-date information, helping you stay current, look up references, or verify facts.

📄 Document Parsing

Beyond text files, AI can understand richer document formats:

Format	Capability
PDF files	Extract all text content from PDFs
Image files	OCR text extraction (supports English and Chinese)

🖥️ Data Analysis

AI has access to a set of read-only analysis commands:

Text search — Search for specific patterns within files
High-speed search — Use ripgrep to locate text across large file sets
File finder — Find files by name or attributes
Word count — Count lines and words in files

All analysis commands are read-only and will never modify your files.

📎 Skill Reference Files

When you activate a Skill, if that Skill includes reference materials (guides, templates, examples), the AI loads them on demand — not all at once, but only when it needs a specific piece of information.

Switch Between Multiple Models

Different AI models have different strengths. GPT AI Flow lets you switch to the best model for each task:

Model	Strength	Best For
OpenAI GPT-4o	Strong all-rounder, fast responses	Daily conversations, file operations, quick Q&A
Anthropic Claude	Excellent at long-form, deep analysis	Long-form writing, in-depth analysis, complex reasoning
DeepSeek	Strong Chinese comprehension, cost-effective	Chinese writing, everyday tasks, budget-conscious use

Switching models is as simple as selecting from the dropdown at the top of the chat — conversation history is preserved, so you can compare results across models.

Full Transparency on Tool Calls

Unlike many AI tools, GPT AI Flow is not a black box. Every time the AI decides to call a tool, you see it in real time:

🔄 AI is thinking — Deciding whether a tool is needed
🛠️ Which tool was called — e.g., "Search vault: writing methods"
📊 Execution progress — How many results were found, which file was read
💬 Final answer — Complete response based on tool results, streamed in real time

Why Transparency Matters

When you can see what the AI did — not just what it said — your trust in the results increases dramatically. You can judge whether the AI found the right materials, whether it missed something important, and decide whether to provide additional instructions.

Three Conversation Modes

Choose the mode that fits your scenario:

Mode	How It Works	Best For
Multi-turn Chat (CHAT)	Maintains full conversation history with continuous context	Deep-diving a topic, co-writing an article
Single Q&A (REPEAT)	Each answer is independent, no history retained	Quick translations, format conversions, one-off tasks
Image Generation (IMAGE)	Calls DALL-E to generate images	Cover images, illustrations, visual content

Wiki Mode — Let AI Dive into Your Obsidian Note Network

Switch to Wiki Mode, and the AI doesn't just read one note — it treats your entire Obsidian Vault as an interconnected knowledge network:

Understands note links: Mention [[A Note]], and the AI opens and references it directly
Follows links forward: Ask "Which notes mention this topic?" and the AI finds them for you
Saves conversations as notes: A valuable AI discussion can be saved directly as a note in your Obsidian Vault, permanently preserved
Works with the llm-wiki Skill: Around a topic, the AI automatically gathers scattered notes and pulls out the connecting thread

Wiki Mode means AI conversations are no longer "talk and forget" — good discussions grow directly into your knowledge base, ready to pick up next time.

A Concrete Workflow You Can Reuse

Type this in the chat:

@note-enricher Help me organize this meeting retrospective — add tags and linked notes
Write the organized conclusions to docs/meeting/2026-04-weekly-review.md
Then list 3 follow-up questions for my next session

The value here: discussion, organization, and output happen in a single chain — no switching between multiple tools.

Conversation History Management

Every conversation is properly saved:

Multi-turn conversations — Full context preserved for review anytime
Smart compression — When conversations get long, the system offers summary compression while retaining core content
Multiple threads — Different topics can have separate conversation threads
Export & share — Export valuable conversations as Markdown or copy to clipboard

File Attachments and Knowledge Base Search

Beyond typing text, you can also:

Drag and drop files — Drop files into the chat window; AI reads their content automatically as conversation context
Knowledge base search — If you've set up the local knowledge base, AI automatically searches it for relevant content, making its answers grounded in what you've already written

This means the AI's answers aren't based solely on general knowledge — they also reference content you've written yourself.

Next Steps

Experience for free immediately:
- Try the web version now
- Download the software to experience all features
Contact Us

Contact Email: hello@gptaiflow.com

Product Feedback:
- Tencent Questionnaire: Click here
- Google Questionnaire: Click here
💬 Have a question? Check out the FAQ for quick solutions: Click here

Thank you for choosing GPT AI Flow, together building the essential tools for the super individuals of the future!

Smart Chat

Standard Chat Mode​

History Chat​

Single Chat​

Default Tools​

When to Use Standard Chat​

AI Smart Conversations — More Than Chat, It Takes Action​

What You Can Actually Do​

Six Built-in Tool Categories — AI Calls Them Automatically​

📁 File Operations​

🔍 Semantic Note Search​

🌐 Web Search​

📄 Document Parsing​

🖥️ Data Analysis​

📎 Skill Reference Files​

Switch Between Multiple Models​

Full Transparency on Tool Calls​

Three Conversation Modes​

Wiki Mode — Let AI Dive into Your Obsidian Note Network​

A Concrete Workflow You Can Reuse​

Conversation History Management​

File Attachments and Knowledge Base Search​

Next Steps​