Migration guide
Move from other AI platforms to Xiajiao while keeping agent definitions and knowledge bases.
From ChatGPT
Custom Instructions / GPTs → Xiajiao SOUL.md + Agents.
Steps
- Install Xiajiao (~30 seconds)
git clone https://github.com/moziio/xiajiao.git
cd xiajiao && npm install && npm start- Configure OpenAI API key
Settings → Model management → Add OpenAI → paste key → pick a model (e.g. gpt-4o).
- Migrate Custom Instructions
Copy ChatGPT Custom Instructions into SOUL.md:
| ChatGPT | Xiajiao SOUL.md |
|---|---|
| "What would you like ChatGPT to know about you?" | ## User context |
| "How would you like ChatGPT to respond?" | ## Output rules |
Example Custom Instructions:
I'm a Python developer. I use FastAPI and PostgreSQL.
Keep responses concise. Use code blocks for code.As SOUL.md:
# Code assistant
You are a senior Python backend engineer.
## User context
- Python developer
- Stack: FastAPI + PostgreSQL
## Output rules
- Keep replies concise
- Use fenced code blocks
- Prefer Python examples- Migrate GPTs
Each GPT → one Xiajiao agent. GPT Instructions → SOUL.md; GPT Knowledge → RAG uploads.
What you gain
| Capability | ChatGPT | Xiajiao |
|---|---|---|
| Multi-agent teamwork | No | Yes (@mention + chains) |
| Agent-to-agent chat | No | Yes (call_agent) |
| Durable memory | Limited | Three-type + embedding dedup |
| Data locality | Cloud | Fully local |
| Cost | ~$20/mo | MIT + your API usage |
| Model choice | Fixed | Any compatible model |
From Dify
Dify Apps / Workflows → Xiajiao agents + collaboration chains.
Steps
Export the system prompt from Dify app settings.
Turn it into SOUL.md
Structure the prompt in Markdown:
# [Agent name]
[First paragraph as role description]
## Working principles
[Rules from the Dify prompt]
## Output format
[Format requirements from the prompt]- Migrate the knowledge base
Export documents (PDF/TXT) from Dify → upload to the agent’s RAG workspace. Xiajiao re-chunks and indexes automatically.
- Migrate workflows
| Dify | Xiajiao |
|---|---|
| Linear workflow | Collaboration chain |
| Branches | Leader agent routing |
| LLM node | Agent |
| Retrieval node | rag_query tool |
| HTTP node | Custom tool |
Note
Dify’s visual workflow canvas has no 1:1 equivalent. Complex branching is expressed via the leader agent’s SOUL.md routing logic.
Trade-offs
| You lose | You gain |
|---|---|
| Drag-and-drop canvas | IM-style multi-agent chat |
| Hosted API product | Agent collaboration chains |
| Multi-tenant SaaS | Three-type durable memory |
| — | No external infra |
| — | Fully offline capable (+ Ollama) |
From Coze
Coze Bot → Xiajiao agent.
Steps
- Export bot settings
Copy persona → SOUL.md; download knowledge docs → upload to Xiajiao RAG.
- Replace plugins
| Coze plugin | Xiajiao |
|---|---|
| Search | web_search |
| Knowledge | rag_query |
| Scheduled triggers | manage_schedule |
| Others | Custom tools |
Coze has 100+ plugins; Xiajiao ships seven built-ins. Heavy plugin reliance may be a poor fit.
- Models
Coze often uses ByteDance models; on Xiajiao you can use any supported provider.
From FastGPT
Knowledge apps → Xiajiao agent + RAG.
Steps
- Export KB documents → upload to Xiajiao
- Copy system prompt → SOUL.md
- Workflows → chains (linear) + leader routing (branches)
Universal migration checklist
- Install and run Xiajiao
- Configure at least one LLM provider
- Create agents with SOUL.md
- Upload knowledge (if any)
- Smoke-test chat
- Configure collaboration chains if needed
- Verify memory over several turns
- Set up schedules if needed
- Configure backups