Agently Langchain Langgraph Translation

LangChain/LangGraph -> Agently Translation Skill

Use this skill to translate LangChain/LangGraph code into Agently patterns for production services.

Core mapping (high level)

• PromptTemplate -> agent.set_agent_prompt / agent.set_request_prompt with mappings
• LLMChain -> agent.input(...).output(...).start()
• OutputParser -> Output Format + get_data() / ensure_keys
• Tools -> @agent.tool_func + agent.use_tool(...)
• Memory -> keep key memory (runtime_data or custom memory); avoid full history
• LangGraph node -> TriggerFlow chunk
• LangGraph edges/conditions -> TriggerFlow when()/to()/if_condition()/match()
• Streaming -> get_generator / get_async_generator or runtime_stream

Translation steps

1. Identify data contracts (inputs/outputs) and enforce with Output Format.
2. Convert graph nodes into TriggerFlow chunks.
3. Replace edges with when/to/condition/collect.
4. Convert tools to @agent.tool_func and call via use_tool or ReAct loop.
5. Use runtime_data for per-run state; keep only key memory.

References

• references/overview.md
• examples/langchain_to_agently_map.py
• examples/langgraph_to_agently_triggerflow.py

Examples

See examples/run.sh for runnable commands.