Skip to content

v3.1.6 — Configurable output-token budget

Choose a tag to compare

View all tags
@wezell wezell released this 23 Jun 21:41
v3.1.6
129eddd
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Configurable output-token budget

The model output-token budget is now configurable through the orchestrator, and the bedrock-generic default is raised so reasoning-model reviews stop truncating mid-thought.

Changes

  • New max_output_tokens orchestrator input (default 32000), forwarded to the bedrock-generic and codex-mantle paths. The Anthropic path manages its own budget and ignores it.
  • bedrock-generic-executor default raised 2048 -> 32000. For reasoning models (e.g. DeepSeek R1) the budget also covers reasoning tokens, so the old cap could be exhausted before a complete review was emitted.

Override example

uses: dotCMS/ai-workflows/.github/workflows/claude-orchestrator.yml@v3.1.6
with:
  model_id: us.deepseek.r1-v1:0
  max_output_tokens: 48000   # optional; defaults to 32000, keep under the model's ceiling

Compatibility

Additive and backward-compatible — new optional input; the generic-path default cap was only raised, never lowered. Anthropic path unaffected.

Full changelog: v3.1.5...v3.1.6

🤖 Generated with Claude Code

Assets 2
Loading