Already have context in another AI? Bring it in one paste.

Convert my memory into Conxt import JSON.

Use, in this order: (1) any memories I paste below,
(2) otherwise everything from this conversation.
If there's nothing to draw on, return an empty array. Never invent anything.

Output shape:
{
  "memories": [
    { "type": "<type>", "content": { ...fields for that type... } }
  ]
}

"type" must be EXACTLY one of (lowercase, with underscores):
decision, preference, entity, open_question, coding_rule, tool_choice, workflow

Each type's "content" (the FIRST field is required, the rest optional):
  decision       -> description (required), rationale, context
  preference     -> preference  (required), category, strength
  entity         -> description (required), name, entity_type, attributes
  open_question  -> question    (required), context
  coding_rule    -> rule        (required), rationale, context
  tool_choice    -> tool        (required), purpose, rationale
  workflow       -> workflow    (required, the workflow's NAME), steps, context

Rules that matter:
- An entity needs a description of what it IS, not just a name.
- A workflow needs a name in "workflow", not only steps.
- "steps" is an array of plain strings; "attributes" is a flat object of string values.
- Drop any record missing its required field rather than guessing.
- Do not add author, source, or timestamp fields; those are set on import.

Return only the JSON. No explanation, no markdown fences.

Example:
{
  "memories": [
    { "type": "decision", "content": { "description": "Use Railway for backend hosting", "rationale": "simple deploys, no k8s overhead", "context": "infra" } },
    { "type": "open_question", "content": { "question": "Team pricing at $20 or $25 per seat?", "context": "pricing" } },
    { "type": "entity", "content": { "description": "Payments provider for subscription billing", "name": "Stripe", "entity_type": "vendor" } }
  ]
}

MY MEMORIES:
(paste here, or leave blank to use this conversation)