wehub-resource-sync
24 lines
1.2 KiB
JSON
24 lines
1.2 KiB
JSON
{
|
|
"_comment": "D5 multi-turn fixture for /demos/agentic-chat against LangGraph Python (LGP). The agent has no backend tools (see src/agents/agentic_chat.py) — it is a plain conversational agent. Three turns: greeting → follow-up that depends on prior context → memory check that depends on the whole conversation. userMessage matches are unique substrings of each turn so first-match-wins ordering is unambiguous. Turn 1 must NOT match a substring that also appears in turns 2/3.",
|
|
"fixtures": [
|
|
{
|
|
"match": { "userMessage": "good name for a goldfish" },
|
|
"response": {
|
|
"content": "How about Bubbles? It is friendly, classic, and easy to call out at the tank. If you want alternatives: Goldie, Finley, or Mango."
|
|
}
|
|
},
|
|
{
|
|
"match": { "userMessage": "name for its tank" },
|
|
"response": {
|
|
"content": "Following the Bubbles theme, you could call the tank The Bubble Bowl. It pairs naturally with the goldfish's name and keeps the playful tone."
|
|
}
|
|
},
|
|
{
|
|
"match": { "userMessage": "what we named the goldfish" },
|
|
"response": {
|
|
"content": "We named the goldfish Bubbles, and the tank The Bubble Bowl."
|
|
}
|
|
}
|
|
]
|
|
}
|