apply_training_plan
ChatGPTActivate the exact accepted draft and its first action with stale-state and scope guards.
attach_training_photo
ChatGPTAttach an owned finalized image to this dog with purpose and actual retention. Transient images expire after 30 days.
begin_role_setup
ChatGPTCall when starting role setup, before any form. This has no UI. For welcome status call show_role_welcome, say the introduction, and continue to the native setup form in the same response. For returning status skip the banner and introduction entirely. Retry transport loss with identical input and key.
cancel_upload
ChatGPTCancel your pending upload.
collect_training_intake_form
ChatGPTBefore proposing dates, times or a default weekend, inspect available calendar integrations and their read capability. If accessible, check availability over a bounded relevant period using the task duration, user's timezone and preferences; offer a few suitable windows without exposing private event details. Free time is not willingness. Do not preselect a time. If access is absent, denied, read-only without availability, or the user declines, explain that and ask about availability or an event-based cue. Reuse an explicitly requested time; do not invent calendar findings. Create or change an event only after a specific request or accepted proposal. For requested conversational follow-through use ChatGPT scheduled tasks when available; do not create both by default. Reconcile uncertain external writes before retrying. Present the current eligible native form when supported; return a reviewed draft without saving. Submit the exact draft with submit_training_answers.
collect_training_setup_form
ChatGPTStart here for setup before asking basic questions in prose. Collect only missing dog name, age and training goal in a native form. Pass only facts already supplied by the person or saved dog context. This is an unsaved opening draft; use create_dog/update_dog_context and start_training_intake afterward, submit_training_answers is only for saved setup IDs. On chat_fallback prefer the available host Visualize skill to present the returned questions as a form with reviewed answer return, unless the person prefers plain chat or asks to stop. If unavailable use chat. If collection aborts, call with preferChat=true to read its questions without reopening UI.
create_dog
ChatGPTSave one dog's volunteered profile facts without intake or a plan. First use list_dogs to avoid duplicates; for an existing dog use get_dog_context and update_dog_context. Include I01 for the reported training goal, I02 for name, and I03 for age in facts, with questionRevision 1 and reported provenance. Read get_dog_context afterward to verify all requested facts were saved.
delete_file
ChatGPTDelete your file if the application has no references to it.
draft_training_plan
ChatGPTUse this when the user wants a practical dog training plan for a supported goal after intake prerequisites are satisfied. Save a draft of the authored exercise at the current dog revision and present that exact draft for acceptance. Does not activate the plan or schedule reminders.
file_url
ChatGPTCreate a short-lived bearer URL to display or download your file.
finalize_upload
ChatGPTVerify uploaded bytes and publish one immutable file. Retry with the same upload ID.
get_dog_context
ChatGPTUse this when the user wants to recover a selected dog’s saved age, goals, household context, or other reported facts. Return provenance and the current revision. Use get_training_state for plans and practice history.
get_file
ChatGPTRead your published file metadata.
get_saved_intake
ChatGPTRead answers saved in Lilypad's browser survey for this account and role. Call after begin_role_setup when the person completed browser intake. The value contains choice identifiers; selectedLabels and otherText contain the person's visible selections and custom answer. Reuse known answers and save confirmed facts with the role's canonical write tools. Do not repeat answered questions; unknown answers remain open. Read-only and scoped to this connected instance.
get_training_briefing
ChatGPTRead a current bounded briefing of accepted actions, reports, gaps and holds. Useful for an on-demand or scheduled check-in.
get_training_intake
ChatGPTRead current missing questions and readiness. Reuse supplied facts; this read does not create setup.
get_training_state
ChatGPTUse this when the user wants to resume a saved dog training plan or see prior practice and observations. Read current plans, actions, reports, household arrangements, and attributed guidance for one selected dog without changing records.
get_upload
ChatGPTRead the state of your upload.
list_dogs
ChatGPTUse this when the user wants to find a saved dog profile or resume dog training across conversations. List only the connected account’s dogs; follow nextCursor and resolve ambiguous labels before selecting a dog.
list_training_definitions
ChatGPTUse this when the user asks which dog training exercises are supported or needs the authored steps before planning. Read all twelve reward-based experiences, including new-puppy routines, potty training and accidents, attention foundations through hand targeting, loose-leash walking, contained recall, settling, and enrichment, with prerequisites, bounds, and AI source review. Does not create a dog or plan; not veterinary care or clinical behavior treatment.
prepare_training_handoff
ChatGPTUse this when the user wants to prepare for a veterinarian or qualified behavior professional consultation using the selected dog’s saved reports, including reports of fear, panic, or intense reactions. Prepare a private, source-attributed summary and questions. Does not diagnose, book a visit, send a message, or export a dataset.
prepare_training_practice
ChatGPTUse this when the user wants to practice a saved, accepted dog training plan now. Read current instructions, completion evidence, and stop conditions without recording an attempt. Do not use to invent an unaccepted exercise or to mark practice complete.
prepare_training_schedule
ChatGPTBefore proposing dates, times or a default weekend, inspect available calendar integrations and their read capability. If accessible, check availability over a bounded relevant period using the task duration, user's timezone and preferences; offer a few suitable windows without exposing private event details. Free time is not willingness. Do not preselect a time. If access is absent, denied, read-only without availability, or the user declines, explain that and ask about availability or an event-based cue. Reuse an explicitly requested time; do not invent calendar findings. Create or change an event only after a specific request or accepted proposal. For requested conversational follow-through use ChatGPT scheduled tasks when available; do not create both by default. Reconcile uncertain external writes before retrying. Return a self-contained accepted-action time proposal. This tool creates no event, task or reminder.
prepare_upload
ChatGPTPrepare a private upload. PUT bytes to the returned URL with your existing bearer token, then finalize.
record_professional_guidance
ChatGPTPreserve an attributed outside professional plan and exact limits. Can supersede previous guidance for the same dog. A supplied name does not verify credentials or authorize treatment.
record_training_action_result
ChatGPTRecord preparation or practice follow-through at the expected action revision. Completion does not establish the training goal.
record_training_observation
ChatGPTUse this when the user asks to save or correct a reported dog behavior observation or routine event, even without a practice session. Save an attributable report once, without fabricated counts or competence promotion. Concerns immediately hold affected practice. A correction can supersede a prior observation and clear its derived concern flag.
record_training_practice
ChatGPTUse this when the user reports actual practice of an accepted dog training plan and wants to log the session or result. Save actual context, help, comfort, and coverage. Do not invent trial details or record a planned session as completed; use record_training_observation for a standalone observation.
record_training_schedule_status
ChatGPTSave actual user-reported or host-observed schedule evidence; never claim provider control or delivery.
review_training_progress
ChatGPTUse this when the user asks how a saved dog training plan is progressing or whether practice should repeat or pause. Review exact plan evidence and all current contrary reports; return a supported repeat, hold, or evidence gap. Do not infer success from missing observations or automatically advance the plan.
set_training_arrangement
ChatGPTSave or change a specifically accepted household arrangement; archiving a plan never weakens it.
set_training_experience_state
ChatGPTPause, resume, maintain, complete or archive an exact accepted experience with its evidence guards. External schedules are separate.
set_training_intake_state
ChatGPTPause or resume saved progressive intake without losing prior answers.
set_training_next_action
ChatGPTSet an accepted feasible cue for the exact current plan; never infer a personal commitment.
show_role_welcome
ChatGPTDisplay the static role banner after begin_role_setup returns welcome. Also allowed when the user explicitly asks to see it again or reports an image failure. Never call automatically for returning status, a new chat, or resumed setup. Read-only: uses the existing issuance and changes no answers or setup state.
start_training_intake
ChatGPTUse this when the user wants to begin a supported reward-based training goal for a selected saved dog. Start or recover progressive intake for that dog, objective, and experience. This saves intake setup, not an accepted training plan. Use collect_training_setup_form when basic dog details are still missing.
submit_training_answers
ChatGPTSave a reviewed eligible answer batch atomically; keep unknown and deferral distinct. No plan acceptance is implied.
update_dog_context
ChatGPTAdd or correct volunteered profile facts without intake or a new plan. Read get_dog_context and use its current revision as expectedRevision. Supplied facts replace only matching question IDs; unrelated facts remain. Include I01 for the reported training goal, I02 for name, and I03 for age with questionRevision 1 and reported provenance. Relevant plan prerequisites are reassessed. Read get_dog_context afterward to verify the save.
whoami
ChatGPTCall before each new write. Copy the returned write.requestId and write.issuedAt unchanged into that write. Reuse identical metadata and input for a transport retry. No domain data is changed.