feat: support custom Claude model#527
Conversation
Co-authored-by: METO <metowolf@users.noreply.github.com>
Co-authored-by: METO <metowolf@users.noreply.github.com>
Co-authored-by: METO <metowolf@users.noreply.github.com>
… with shouldSetClaudeHaikuModelEnv
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
Findings
- No issues found.
Summary
- Review mode: initial
- Reviewed the full diff for runner-spawned Claude model propagation and the NewSession custom model UI. Residual risk: no automated web interaction test covers the custom-model field/validation path.
Testing
- Not run (automation)
HAPI Bot
There was a problem hiding this comment.
Findings
- [Minor] Reuse the shared Claude preset list — this helper now hard-codes the same preset aliases that already live in
@hapi/protocol, so the next preset addition can make runner-spawned preset sessions look like custom models and setANTHROPIC_DEFAULT_HAIKU_MODELunexpectedly. Evidencecli/src/runner/run.ts:28.
Suggested fix:import { CLAUDE_MODEL_PRESETS } from '@hapi/protocol'; const CLAUDE_BUILT_IN_MODEL_ALIASES = new Set<string>(CLAUDE_MODEL_PRESETS);
Summary
- Review mode: follow-up after new commits
- Reviewed the full latest diff, using the previous no-finding bot review and compare diff as context. One minor maintainability issue found; no blocker/major correctness or security issue found in the custom model submit/spawn path.
Testing
- Not run:
bunis not installed in this runner image (bun: command not found).
HAPI Bot
| import { resolveWorkspaceRoot } from '@/utils/workspaceRoot'; | ||
| import { hashRunnerCliApiToken } from './runnerIdentity'; | ||
|
|
||
| const CLAUDE_BUILT_IN_MODEL_ALIASES = new Set(['opus', 'opus[1m]', 'sonnet', 'sonnet[1m]']); |
There was a problem hiding this comment.
[MINOR] This duplicates the canonical Claude preset list from @hapi/protocol (CLAUDE_MODEL_PRESETS). If a new built-in preset is added there, runner-spawned sessions using that preset will be treated as custom here and will set ANTHROPIC_DEFAULT_HAIKU_MODEL unexpectedly.
Suggested fix:
import { CLAUDE_MODEL_PRESETS } from '@hapi/protocol';
const CLAUDE_BUILT_IN_MODEL_ALIASES = new Set<string>(CLAUDE_MODEL_PRESETS);Replace hard-coded preset aliases with the shared constant so new presets are automatically picked up without requiring a parallel update here.
There was a problem hiding this comment.
Findings
No issues found.
Summary
Review mode: follow-up after new commits
Reviewed the full latest diff, with the prior bot review and the 40db34c1f7c27b05187cc913dca6b4b66b962e0c...c2f8107ea5e2044dc312591bd3ff632e12c9e22d compare as follow-up context. The previous minor finding is addressed by using CLAUDE_MODEL_PRESETS; no blocker/major/minor correctness, security, regression, data loss, performance, or maintainability issues found. Residual risk: tests were not executed in this automation pass.
Testing
Not run (automation)
HAPI Bot
There was a problem hiding this comment.
Findings
No issues found.
Summary
Review mode: follow-up after new commits
Reviewed the full latest diff, using the prior HAPI Bot review at c2f8107ea5e2044dc312591bd3ff632e12c9e22d and the compare to ca5222b6fc2f482197692d72da322184ff9f2f34 as follow-up context. No blocker/major/minor correctness, security, regression, data loss, performance, or maintainability issues found. Residual risk: the custom Claude startup path is covered by focused unit coverage, but I could not execute it in this environment.
Testing
Not run: bun test cli/src/runner/buildCliArgs.test.ts web/src/components/NewSession/types.test.ts failed because bun is not installed in the runner environment.
HAPI Bot
Resolve conflicts by combining: - Custom Claude model support (shouldSetClaudeHaikuModelEnv, customModel state, ModelSelector props) - Upstream multiple workspace roots support (workspaceRoots plural) - Upstream OpenCode model selector (OpencodeModelSelector conditional)
There was a problem hiding this comment.
Findings
No issues found.
Summary
Review mode: follow-up after new commits
No blocker/major/minor issues found in the latest diff. Residual risk: the custom Claude model flow only has unit-level coverage changes here, and I could not execute the Bun test suite in this environment.
Testing
Not run (automation)
HAPI Bot
2 participants
Summary
Changes
custommodel type for Claude agent, allowing users to input any model name (e.g., custom model versions or preview builds)customoptionTest plan