Skip to content

feat(llm): add speed option for provider fast mode - #8

Merged
leofmarciano merged 4 commits into
mainfrom
devin/1793682093-fast-mode
Nov 3, 2026
Merged

leofmarciano merged 4 commits into
mainfrom
devin/1793682093-fast-mode

Conversation

@devin-ai-integration

@devin-ai-integration devin-ai-integration Bot commented Nov 3, 2026 •

Copy link
Copy Markdown
Contributor

What and why

Adds speed= next to thinking= to run the editing LLM in the provider's fast mode: same model, faster output, about 2x the price.

wsp.improve(model, "...", llm="anthropic:claude-opus-5-5", speed="fast")
  • Session, wsp.read/fix/write/improve and the backends take speed (None = standard, True/"fast", "standard", or a tier like "ultrafast"). CLI: --speed / --fast. Saved in result.json.
  • Anthropic sends speed; OpenAI and Open Responses send service_tier.
  • Opus 5.5 has always-on thinking and rejects thinking.type="disabled", so thinking=False now falls back to effort="low" (plus the act-directly prompt) on such models. usage.thinking_tokens is recorded so you can see how much it still thought.
  • experiments/exp8/run_edit.py takes --speed fast. Docs, README, llms.txt and the changelog are updated.

How it was tested

  • ruff check . && ruff format --check .
  • pytest (27 passed; new tests for speed on both providers, on Session/Result, and the low-effort fallback)
  • Added a CHANGELOG.md entry (if user-facing)
  • mypy src/weightspace
  • Real Opus 5.5 call with speed="fast", thinking=False: edited a numpy vector correctly, response reported speed='fast', 0 thinking tokens.

Co-Authored-By: Leonardo Marciano <leonardo.marciano@toolzz.me>
@devin-ai-integration

Copy link
Copy Markdown
Contributor Author

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

  • Disable automatic comment, CI, and merge conflict monitoring

devin-ai-integration Bot and others added 2 commits November 3, 2026 05:05
Co-Authored-By: Leonardo Marciano <leonardo.marciano@toolzz.me>
Co-Authored-By: Leonardo Marciano <leonardo.marciano@toolzz.me>
@devin-ai-integration devin-ai-integration Bot changed the title feat(exp8): add --speed flag for provider fast mode feat(llm): add speed option for provider fast mode Nov 3, 2026
@leofmarciano

Copy link
Copy Markdown
Owner

Lgfm

@leofmarciano
leofmarciano merged commit da729dd into main Nov 3, 2026
7 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant