Open the main FastAPI app file (the one that defines require_api_key, require_auth, /api/lia/inbound, etc. – currently the dashboard server file).

1) Update imports from app.lia_rules

Find the import from app.lia_rules (it currently imports things like LIA_TIER_BEHAVIOUR, LIA_TONE_GUIDELINES, classify_inbound_tier, select_eot_intro) and make sure it ALSO imports LIA_NUDGE_TEMPLATES, for example:

from app.lia_rules import (
    LIA_TIER_BEHAVIOUR,
    LIA_TONE_GUIDELINES,
    classify_inbound_tier,
    select_eot_intro,
    LIA_NUDGE_TEMPLATES,
)

If the import is currently on one line, expand it to a multi-line import as above.

2) Add a Pydantic model for the nudge request

Make sure this file imports BaseModel from pydantic (it already does for WhatsAppMessage). Then, somewhere near the other models (e.g. near WhatsAppMessage), add:

class LiaNudgeRequest(BaseModel):
    conversation_id: str
    nudge_type: str = "nudge_1"   # e.g. "nudge_1" or "nudge_2"
    payments_remaining: int | None = None  # reserved for future use

3) Add the /api/lia/nudge endpoint

Below the existing /api/lia/inbound endpoint (or near the other Lia-related APIs), add:

@app.post("/api/lia/nudge", dependencies=[Depends(require_api_key)])
async def lia_nudge(payload: LiaNudgeRequest):
    """
    Generate and log a nudge message for a conversation, without needing an inbound customer message.
    This is designed to be called from Make.com using the x-api-key.
    It returns the text Lia should send, and logs it into dp_messages.
    """

    # Map simple nudge_type to the internal template key
    mapping = {
        "nudge_1": "NUDGE_1_EOT",
        "nudge_2": "NUDGE_2_EOT",
        "NUDGE_1_EOT": "NUDGE_1_EOT",
        "NUDGE_2_EOT": "NUDGE_2_EOT",
    }

    template_key = mapping.get(payload.nudge_type)
    if not template_key:
        raise HTTPException(status_code=400, detail="Invalid nudge_type")

    # Fetch the template text
    try:
        reply_text = LIA_NUDGE_TEMPLATES[template_key]
    except KeyError:
        raise HTTPException(status_code=500, detail="Nudge template not configured")

    # Log the nudge as an outbound Lia message
    try:
        await insert_message(
            conversation_id=payload.conversation_id,
            direction="outbound",
            channel="whatsapp",
            sender="lia",
            message_body=reply_text,
            template_key=template_key,
            mode=payload.nudge_type,
            ab_variant=None,
        )
    except Exception as e:
        print(f"Failed to insert nudge message: {e}")
        raise HTTPException(status_code=500, detail="Failed to log nudge message")

    return {
        "conversation_id": payload.conversation_id,
        "nudge_type": payload.nudge_type,
        "template_key": template_key,
        "reply": reply_text,
    }

4) Do not change any existing endpoints or business logic. Only add the model + new endpoint + updated import.

5) Save the file and restart the server. Confirm there are no errors.