---
title: "Update Live"
method: POST
path: "/v1/deployments/{deployment_id}/update"
tags: ["deployments"]
---

# Update Live

`POST /v1/deployments/{deployment_id}/update`

## Request

- Base URL: `https://api.usekeel.io` (the document declares no server; this is the origin it was published from)
- URL: `https://api.usekeel.io/v1/deployments/{deployment_id}/update`
- Auth: none declared

## Path parameters

- `deployment_id` string, required

## Headers

- `Authorization` string, nullable
- `X-Org-Id` string, nullable

## Request body

- UpdateLiveRequest: Optional body for update-live endpoint. ``tranching`` and ``execution_style`` are THREE-WAY (Q-0439). Pydantic's ``model_fields_set`` distinguishes an omitted key from an explicit ``null``, and the three states mean different things: - **key absent** → CARRY the deployment's current block forward unchanged. This is the historical behaviour and stays the default, so an update that only advances the commit can never disturb an operator block. - **key present with a document** → CHANGE to that block (grant-gated and parsed exactly as on deploy; the schedule is re-derived from it). - **key present and ``null``** → REMOVE the block. Disarming needs no grant — an operator may always turn a capability off. Before this, the only way to change either block was to stop the deployment and deploy again, which is precisely the path that silently drops tranching (the wizard does not carry it, see the DeployLiveRequest note).
  - `schedule` string, nullable: Optional cron spelling; it must equal the selected commit's derived schedule. Omit to reconcile automatically. On a deployment whose tranching block is being changed in this same request, the derived schedule is the one implied by the NEW block.
  - `intent_token` string, nullable: Optional update-intent token when this update completes a handoff flow. Telemetry-only, mirroring DeployLiveRequest.intent_token: a valid token bound to this exact deployment makes the update emit handoff_completed (kind=update); it never gates or alters the update and carries no configuration.
  - `tranching` object, nullable: Three-way (see class docstring): omit to carry the current block, send a block ({"mode": "equal_split", "count": N, "window"?: M} or {"mode": "custom", "times": ["HH:MM", ...], "window"?: M}) to change it, or send null to remove it and return the deployment to a single fire per period. Setting a block requires the org's 'tranching' execution grant (403 without it); an invalid block is a 422 with the exact reason.
  - `execution_style` object, nullable: Three-way (see class docstring): omit to carry the current style, send the closed-world maker document to change it, or send null to disarm back to market. Setting a style requires the org's 'maker' execution grant (403 without it); an invalid document is a 422 with the exact parse reason.

## Response `200`

Successful Response

- DeploymentResponse
  - `deployment_id` string, required
  - `strategy_id` string, required
  - `name` string, required
  - `status` string, required
  - `schedule` string, nullable
  - `evaluation_grid` EvaluationGrid: The candle closes a deployment's runs evaluate (UX-8). Closes sit at ``offset_minutes`` past 00:00 UTC and every ``period_minutes`` after it: the deployment's cron inverted under its persisted dispatch generation (``utils.schedule.evaluation_grid``), so a pre-arm cron that fires at 11:59 reads as the 12:00 close it evaluates.
    - `period_minutes` integer, required
    - `offset_minutes` integer, required
  - `source_hash` string, nullable
  - `s3_key` string, nullable
  - `org_id` string, nullable
  - `account_id` string, nullable
  - `deployed_at` string, nullable
  - `stopped_at` string, nullable
  - `created_at` string, nullable
  - `updated_at` string, nullable
  - `deployed_commit_id` string, nullable
  - `deployed_version_string` string, nullable
  - `config_id` string, nullable
  - `total_pnl` number, nullable
  - `position_count` integer, nullable
  - `venue` VenueBlock: Which venue a deployment or account runs on, and its money class. Paper Trading v2, S01 §3.10 / IC-01-9. ``capital_class`` is ``trading.accounts.capital_class`` (``real`` or ``practice``); ``code`` is the venue's ``trading.venues.code`` (``HL``, ``PAPER``). On a deployment whose ``account_id`` names no account row (no foreign key refuses one, ``V6.0__deployments.sql:16``) the block reads ``{code: null, capital_class: "real"}``: such a row predates paper and keeps its pre-paper meaning. Clients key on this block, never on a probe.
    - `code` string, nullable
    - `capital_class` string, required
  - `remaining` object, nullable
  - `universe_warnings` UniverseWarning[], nullable
    - `symbol` string, required
    - `reason` 'DELISTED' | 'STALE', required
    - `detail` string, required
  - `account_warnings` AccountWarning[], nullable
    - `kind` 'EQUITY_BELOW_VENUE_MINIMUM' | 'ORDERS_SKIPPED_MIN_NOTIONAL' | 'EXITS_BELOW_VENUE_MINIMUM' | 'POSITION_STRANDED_BELOW_MINIMUM', required
    - `account_value` number, required
    - `min_notional` number, required
    - `detail` string, required
    - `orders_skipped` integer, nullable
    - `orders_total` integer, nullable
    - `exits_skipped` integer, nullable
    - `window_days` integer, nullable
    - `exit_symbols` string[], nullable
  - `execution_style` 'market' | 'maker', nullable
  - `paused_reason` string, nullable
  - `paused_at` string, nullable
  - `tranching` object, nullable
  - `health` 'running' | 'degraded' | 'stalled' | 'stopped', nullable
  - `health_detail` HealthDetailResponse: IC-04-2's ``health_detail`` (S04 §3.10.2): the first predicate that matched, the count of most recent consecutive missing bars, the newest COMPLETE bar, and the bar label the current state began at.
    - `reason` 'consecutive_missing' | 'latest_eval_failed' | 'latest_execution_failed' | 'dispatch_exhausted' | 'execution_missing' | 'rail_blocked' | 'funding_stalled' | 'equity_stale' | 'failure_rate' | 'silence' | 'halted' | 'schedule_unreadable', nullable, required
    - `consecutive_missing` integer, required
    - `last_complete_at` string, date-time, nullable, required
    - `since` string, date-time, nullable, required
  - `parity` ParityBlock: IC-03-2: the record's latest non-day-0 snapshot, else its first day-0 row.
    - `residual_bps` number, nullable, required
    - `status` 'tracking' | 'drifting' | 'under_review', required
    - `window_start` string, date-time, nullable, required
    - `window_end` string, date-time, nullable, required
    - `as_of` string, date-time, nullable, required
    - `eval_count` integer, required
    - `meter_eligible` boolean, required
    - `day0_backtest_run_id` string, nullable, required
    - `is_day0` boolean, required
    - `paper_return_pct` number, nullable, required
    - `backtest_return_pct` number, nullable, required
  - `graduation` GraduationBlock: IC-03-5: the graduation criteria state, computed on read. ``record`` and ``carry_over`` are null only for a deployment with no deployed commit (the record has no commit to name).
    - `ready` boolean, required
    - `ready_since` string, date-time, nullable, required
    - `nudged_at` string, date-time, nullable, required
    - `evaluated_at` string, date-time, required
    - `record` GraduationRecord, required: IC-03-5 ``record``: the record the criteria run over (R-27).
      - `started_at` string, date-time, required
      - `commit_id` string, required
      - `commit_seq` integer, required
      - `commit_changed_at` string, date-time, nullable, required
    - `thresholds` GraduationThresholds, required: IC-03-5 ``thresholds``: frozen at create from the defaults version (IC-03-6).
      - `min_age_days` integer, required
      - `min_evals` integer, required
      - `parity_band_bps` number, required
      - `backtest_max_dd_pct` number, nullable, required
      - `dd_floor_pending` boolean, required
      - `defaults_version` integer, required
    - `criteria` GraduationCriteria, required
      - `age` AgeCriterion, required
        - `met` boolean, required
        - `days` integer, required
        - `required_days` integer, required
      - `evals` EvalsCriterion, required
        - `met` boolean, required
        - `count` integer, required
        - `required` integer, required
      - `parity` ParityCriterion, required
        - `met` boolean, required
        - `residual_bps` number, nullable, required
        - `band_bps` number, required
        - `as_of` string, date-time, nullable, required
        - `eval_count` integer, required
        - `meter_eligible` boolean, required
        - `stale` boolean, required
      - `drawdown` DrawdownCriterion, required: Both drawdowns are positive magnitudes; the floor is null while pending.
        - `met` boolean, required
        - `paper_max_dd_pct` number, nullable, required
        - `backtest_max_dd_pct` number, nullable, required
    - `carry_over` CarryOver, required: IC-03-5 ``carry_over`` (S03 §3.9.5, R-32): no simulated capital.
      - `strategy_id` string, required
      - `commit_id` string, required
      - `schedule` string, nullable, required
      - `source_paper_deployment_id` string, required
      - `head_commit_id` string, required
      - `commit_ahead` boolean, required
      - `commit_seq` integer, required
      - `head_commit_seq` integer, required
      - `simulated_capital` string, nullable
      - `min_order_notional` string, nullable
  - `money_freshness` MoneyFreshness: Coverage state of the money facts behind a performance read (spec 04 §5.1, M4). Freshness is part of the answer: the number beside this block is only as fresh as the capture coverage that produced its inputs, and this block says exactly how fresh — * ``COVERED`` — coverage reaches the requested interval's end; ``fresh_as_of`` is the weakest fact class's reach. * ``SYNCING`` — the interval is managed and post-epoch but coverage has not reached its end yet; ``synced_through`` is how far it has (null when no coverage row speaks for the interval at all — absence never reads as fresh). * ``LEGACY`` — the interval predates the account's first coverage row (the epoch cutover): pre-capture history, served ungated but labeled. Until the capture lane goes live, every account reads this. * ``UNMANAGED`` — the interval overlaps no managed window; strategy performance is not defined there. Phase 1 labels only — no numeric field elsewhere in the response changes because of this block.
    - `state` 'COVERED' | 'SYNCING' | 'LEGACY' | 'UNMANAGED', required
    - `fresh_as_of` string, nullable
    - `synced_through` string, nullable
  - `serving_warnings` object[], nullable
  - `known_issues` KnownIssueRef[], nullable
    - `id` string, required
    - `summary` string, required
  - `live_window_warnings` LiveWindowWarning[], nullable
    - `code` 'LIVE_WINDOW_SHORTER_THAN_TRADES', required
    - `message` string, required
    - `backtest_run_id` string, required

## Other responses

- `422`: Validation Error

---

[API](https://skmtc.dev/usekeel/apis/keel-api.md) · [All operations](https://skmtc.dev/usekeel/apis/keel-api/llms.txt) · [OpenAPI document](https://skmtc.dev/usekeel/apis/keel-api/revisions/3191ca10b0ab?raw)