Add Gemini Pro freebuff model (#566)

Co-authored-by: James Grugett <jahooma@gmail.com>

Author: brandonkachen

agents/base2/base2-gemini-no-editor-evals.ts

@@ -0,0 +1,13 @@
+import { createBase2 } from './base2'
+
+const definition = {
+  ...createBase2('free', {
+    noAskUser: true,
+    model: 'google/gemini-3.1-pro-preview',
+    providerOptions: {},
+  }),
+  id: 'base2-gemini-no-editor-evals',
+  displayName: 'Buffy the Gemini Evals Orchestrator',
+}
+
+export default definition

agents/base2/base2.ts

@@ -12,29 +12,37 @@ export function createBase2(
     hasNoValidation?: boolean
     planOnly?: boolean
     noAskUser?: boolean
+    model?: SecretAgentDefinition['model']
+    providerOptions?: SecretAgentDefinition['providerOptions']
   },
 ): Omit<SecretAgentDefinition, 'id'> {
   const {
     hasNoValidation = mode === 'fast',
     planOnly = false,
     noAskUser = false,
+    model: modelOverride,
+    providerOptions,
   } = options ?? {}
   const isDefault = mode === 'default'
   const isFast = mode === 'fast'
   const isMax = mode === 'max'
   const isFree = mode === 'free' || mode === 'lite'
 
   const isSonnet = false
-  const model = isFree ? 'z-ai/glm-5.1' : 'anthropic/claude-opus-4.7'
+  const model =
+    modelOverride ?? (isFree ? 'z-ai/glm-5.1' : 'anthropic/claude-opus-4.7')
+  const defaultProviderOptions = isFree
+    ? {
+        data_collection: 'deny' as const,
+      }
+    : {
+        only: ['amazon-bedrock'],
+      }
 
   return {
     publisher,
     model,
-    providerOptions: isFree ? {
-      data_collection: 'deny',
-    } : {
-      only: ['amazon-bedrock'],
-    },
+    providerOptions: providerOptions ?? defaultProviderOptions,
     displayName: 'Buffy the Orchestrator',
     spawnerPrompt:
       'Advanced base agent that orchestrates planning, editing, and reviewing for complex coding tasks',
@@ -150,8 +158,6 @@ Use the spawn_agents tool to spawn specialized agents to help you complete the u
         isMax &&
         `- IMPORTANT: You must spawn the editor-multi-prompt agent to implement the changes after you have gathered all the context you need. You must spawn this agent for non-trivial changes, since it writes much better code than you would with the str_replace or write_file tools. Don't spawn the editor in parallel with context-gathering agents.`,
         isFree &&
-        '- Implement code changes using the str_replace or write_file tools directly.',
-        isFree &&
         '- Spawn a code-reviewer-lite to review the changes after you have implemented the changes.',
         '- Spawn bashers sequentially if the second command depends on the the first.',
         isDefault &&

cli/src/components/freebuff-model-selector.tsx

@@ -5,6 +5,7 @@ import React, { useCallback, useEffect, useMemo, useState } from 'react'
 import { Button } from './button'
 import {
   FALLBACK_FREEBUFF_MODEL_ID,
+  FREEBUFF_GEMINI_PRO_MODEL_ID,
   FREEBUFF_GLM_MODEL_ID,
   FREEBUFF_MODELS,
   getFreebuffDeploymentAvailabilityLabel,
@@ -25,8 +26,15 @@ import {
 import type { KeyEvent } from '@opentui/core'
 
 const FREEBUFF_MODEL_SELECTOR_MODELS = [
+  ...FREEBUFF_MODELS.filter(
+    (model) => model.id === FREEBUFF_GEMINI_PRO_MODEL_ID,
+  ),
   ...FREEBUFF_MODELS.filter((model) => model.id === FREEBUFF_GLM_MODEL_ID),
-  ...FREEBUFF_MODELS.filter((model) => model.id !== FREEBUFF_GLM_MODEL_ID),
+  ...FREEBUFF_MODELS.filter(
+    (model) =>
+      model.id !== FREEBUFF_GEMINI_PRO_MODEL_ID &&
+      model.id !== FREEBUFF_GLM_MODEL_ID,
+  ),
 ]
 
 /**

common/src/__tests__/freebuff-models.test.ts

@@ -1,11 +1,32 @@
 import { describe, expect, test } from 'bun:test'
 
 import {
+  FREEBUFF_GEMINI_PRO_MODEL_ID,
+  FREEBUFF_MODELS,
   getFreebuffDeploymentAvailabilityLabel,
   isFreebuffDeploymentHours,
+  isFreebuffModelAvailable,
 } from '../constants/freebuff-models'
 
 describe('freebuff model availability', () => {
+  test('includes Gemini 3.1 Pro as an always-available option', () => {
+    expect(FREEBUFF_MODELS.map((model) => model.id)).toContain(
+      FREEBUFF_GEMINI_PRO_MODEL_ID,
+    )
+    expect(
+      isFreebuffModelAvailable(
+        FREEBUFF_GEMINI_PRO_MODEL_ID,
+        new Date('2026-01-05T18:00:00Z'),
+      ),
+    ).toBe(true)
+    expect(
+      isFreebuffModelAvailable(
+        FREEBUFF_GEMINI_PRO_MODEL_ID,
+        new Date('2026-01-05T12:00:00Z'),
+      ),
+    ).toBe(true)
+  })
+
   test('formats the close time in the user local timezone while deployment is open', () => {
     expect(
       getFreebuffDeploymentAvailabilityLabel(new Date('2026-01-05T18:00:00Z'), {

common/src/constants/free-agents.ts

@@ -1,5 +1,7 @@
 import { parseAgentId } from '../util/agent-id-parsing'
 
+import { FREEBUFF_MODELS } from './freebuff-models'
+
 import type { CostMode } from './model-config'
 
 /**
@@ -15,6 +17,10 @@ export const FREE_COST_MODE = 'free' as const
  * every user's apparent activity.
  */
 export const FREEBUFF_ROOT_AGENT_IDS = ['base2-free'] as const
+const FREEBUFF_ROOT_AGENT_ID_SET: ReadonlySet<string> = new Set(
+  FREEBUFF_ROOT_AGENT_IDS,
+)
+const FREEBUFF_SELECTABLE_MODEL_IDS = FREEBUFF_MODELS.map((model) => model.id)
 
 /**
  * Agents that are allowed to run in FREE mode.
@@ -26,10 +32,7 @@ export const FREEBUFF_ROOT_AGENT_IDS = ['base2-free'] as const
  */
 export const FREE_MODE_AGENT_MODELS: Record<string, Set<string>> = {
   // Root orchestrator
-  'base2-free': new Set([
-    'minimax/minimax-m2.7',
-    'z-ai/glm-5.1',
-  ]),
+  'base2-free': new Set(FREEBUFF_SELECTABLE_MODEL_IDS),
 
   // File exploration agents
   'file-picker': new Set(['google/gemini-2.5-flash-lite']),
@@ -44,16 +47,10 @@ export const FREE_MODE_AGENT_MODELS: Record<string, Set<string>> = {
   'basher': new Set(['google/gemini-3.1-flash-lite-preview']),
 
   // Editor for free mode
-  'editor-lite': new Set([
-    'minimax/minimax-m2.7',
-    'z-ai/glm-5.1',
-  ]),
+  'editor-lite': new Set(FREEBUFF_SELECTABLE_MODEL_IDS),
 
   // Code reviewer for free mode
-  'code-reviewer-lite': new Set([
-    'minimax/minimax-m2.7',
-    'z-ai/glm-5.1',
-  ]),
+  'code-reviewer-lite': new Set(FREEBUFF_SELECTABLE_MODEL_IDS),
 }
 
 /**
@@ -87,6 +84,13 @@ export function isFreeMode(costMode: CostMode | string | undefined): boolean {
   return costMode === FREE_COST_MODE
 }
 
+export function isFreebuffRootAgent(fullAgentId: string): boolean {
+  const { publisherId, agentId } = parseAgentId(fullAgentId)
+  if (!agentId) return false
+  if (publisherId && publisherId !== 'codebuff') return false
+  return FREEBUFF_ROOT_AGENT_ID_SET.has(agentId)
+}
+
 /**
  * Check if a specific agent is allowed to use a specific model in FREE mode.
  * This is the strictest check - validates both the agent AND model combination.

common/src/constants/freebuff-models.ts

@@ -21,6 +21,7 @@ export interface FreebuffModelOption {
  *  the caller's local timezone. The CLI should render
  *  `getFreebuffDeploymentAvailabilityLabel()` instead. */
 export const FREEBUFF_DEPLOYMENT_HOURS_LABEL = '9am ET-5pm PT every day'
+export const FREEBUFF_GEMINI_PRO_MODEL_ID = 'google/gemini-3.1-pro-preview'
 export const FREEBUFF_GLM_MODEL_ID = 'z-ai/glm-5.1'
 export const FREEBUFF_MINIMAX_MODEL_ID = 'minimax/minimax-m2.7'
 const FREEBUFF_EASTERN_TIMEZONE = 'America/New_York'
@@ -40,6 +41,12 @@ interface LocalTimeFormatOptions {
 }
 
 export const FREEBUFF_MODELS = [
+  {
+    id: FREEBUFF_GEMINI_PRO_MODEL_ID,
+    displayName: 'Gemini 3.1 Pro',
+    tagline: 'Deepest, 1/day',
+    availability: 'always',
+  },
   {
     id: FREEBUFF_MINIMAX_MODEL_ID,
     displayName: 'MiniMax M2.7',

common/src/types/contracts/database.ts

@@ -35,6 +35,7 @@ export type GetUserInfoFromApiKeyFn = <T extends UserColumn>(
 
 type AgentRun = {
   agent_id: string
+  ancestor_run_ids: string[]
   status: 'running' | 'completed' | 'failed' | 'cancelled'
 }
 export type AgentRunColumn = keyof AgentRun

common/src/types/freebuff-session.ts

@@ -9,10 +9,9 @@
 /**
  * Per-model usage counter surfaced to the CLI so the waiting-room UI can
  * render "N of M sessions used" alongside queue/active state. Present when
- * the joined model has a rate limit applied (today: GLM 5.1 with 5 admits
- * per 12-hour window). `recentCount` is the number of admissions inside
- * `windowHours` at the time the response was produced — see also the
- * standalone `rate_limited` status for the reject path.
+ * the joined model has a rate limit applied. `recentCount` is the number of
+ * admissions inside `windowHours` at the time the response was produced —
+ * see also the standalone `rate_limited` status for the reject path.
  */
 export interface FreebuffSessionRateLimit {
   model: string
@@ -72,7 +71,7 @@ export type FreebuffSessionServerResponse =
       queueDepthByModel: Record<string, number>
       estimatedWaitMs: number
       queuedAt: string
-      /** Rate-limit quota for rate-limited models (GLM 5.1 today). Absent
+      /** Rate-limit quota for rate-limited models. Absent
        *  for unlimited models or when the status was produced outside the
        *  rate-limit check path (e.g. pure read via GET). */
       rateLimit?: FreebuffSessionRateLimit
@@ -85,7 +84,7 @@ export type FreebuffSessionServerResponse =
       admittedAt: string
       expiresAt: string
       remainingMs: number
-      /** Rate-limit quota for rate-limited models (GLM 5.1 today). Absent
+      /** Rate-limit quota for rate-limited models. Absent
        *  for unlimited models or when the status was produced outside the
        *  rate-limit check path (e.g. pure read via GET). */
       rateLimit?: FreebuffSessionRateLimit
@@ -152,7 +151,7 @@ export type FreebuffSessionServerResponse =
     }
   | {
       /** User has used up their per-model admission quota in the rolling
-       *  window (GLM 5.1: 5 one-hour sessions per 12h). Returned from POST
+       *  window. Returned from POST
        *  /session before the user is placed in the queue. `retryAfterMs` is
        *  the time until the oldest admission inside the window falls off
        *  and one quota slot opens up — clients should show the user when