---
title: "Set Auto Model Bias"
method: PATCH
path: "/api/auth/workspace/auto-model-bias"
tags: ["auth"]
---

# Set Auto Model Bias

`PATCH /api/auth/workspace/auto-model-bias`

## Request

- Base URL: `https://api.advisorppc.com` (the document declares no server; this is the origin it was published from)
- URL: `https://api.advisorppc.com/api/auth/workspace/auto-model-bias`
- Auth: HTTP bearer

Set how hard composer Auto may spend for the caller's workspace.

Scoped to ``ctx.tenant_id``, so a caller cannot change a workspace they are
not currently in. Admin-gated for the same reason the rename is: it changes
what every member of the workspace is billed for on an Auto turn.

This grants no entitlement. Every lane the bias can select is one the
workspace's plan already permits, the plan gates in ``routers/chat`` still
run after Auto resolves, and the free plan ignores the bias outright — so
the worst a member can do with it is choose a costlier model they were
already allowed to pick by hand.

## Headers

- `X-Tenant-Id` integer, nullable

## Request body

- AutoModelBiasIn: Set how hard composer Auto may spend for the caller's workspace. Validated against the stored vocabulary rather than sanitized: a client sending an unknown value is stale or buggy, and silently storing "economy" for a reader who asked for "performance" would look like the setting simply did not save. The read path still sanitizes, because a hand-edited row must not break chat.
  - `bias` string, required

## Response `200`

Successful Response

- AutoModelBiasOut
  - `tenant_id` integer, required
  - `auto_model_bias` string, required

## Other responses

- `422`: Validation Error

---

[API](https://skmtc.dev/advisorppc/apis/advisorppc-api.md) · [All operations](https://skmtc.dev/advisorppc/apis/advisorppc-api/llms.txt) · [OpenAPI document](https://skmtc.dev/advisorppc/apis/advisorppc-api/revisions/3441bd729e0c?raw)