emileriksenkeev adds a Harvey Specter button for harder legal questions

The sydOS fork adds a named chat setting for users who want the AI to spend more effort on a response.

chat-uiworkflow

"Harvey Specter Mode" is a selectable chat option that runs Claude Fable 5 at its maximum reasoning setting. Claude is an AI model used here to generate responses.

It is a small interface move, but a practical one: users can deliberately choose a more intensive setting when a matter warrants extra thought, without having to understand the underlying model controls.

So what Teams already offering AI chat may like the clearer choice between everyday speed and a more considered answer.

View this fork on GitHub →

Spotted something wrong? Or know the PR text has fresher detail than the writeup above?

Commits in this thread

1 commit from emileriksenkeev/sydOS, oldest first. Source extracted verbatim from the harvested git log.

SHA Subject Author Date

efa041e6 Harvey Specter Mode: Claude Fable 5 at max effort in the chat picker emileriksenkeev 2026-07-17 ↗ GitHub

SHA	Subject	Author	Date
`efa041e6`	Harvey Specter Mode: Claude Fable 5 at max effort in the chat picker	emileriksenkeev	2026-07-17	↗ GitHub
commit body Adds claude-fable-5 as a main-chat model ("Harvey Specter Mode" in the UI). Fable's API differs from Opus-tier, so the Claude adapter handles it specially: thinking is always on with the summarized display (the default returns empty thinking text), effort defaults to "max", output headroom is raised to 64K, and requests opt into the server-side refusal fallback to claude-opus-4-8 (Fable's safety classifiers can decline a request; the fallback answers it in the same call). A whole- chain refusal surfaces as a visible note instead of a silent dead stream, and blocks from a declined mid-stream attempt are neither executed as tools nor echoed back on the next turn. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

commit body

Adds claude-fable-5 as a main-chat model ("Harvey Specter Mode" in the
UI). Fable's API differs from Opus-tier, so the Claude adapter handles
it specially: thinking is always on with the summarized display (the
default returns empty thinking text), effort defaults to "max", output
headroom is raised to 64K, and requests opt into the server-side
refusal fallback to claude-opus-4-8 (Fable's safety classifiers can
decline a request; the fallback answers it in the same call). A whole-
chain refusal surfaces as a visible note instead of a silent dead
stream, and blocks from a declined mid-stream attempt are neither
executed as tools nor echoed back on the next turn.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Capture this thread into my fork

Download a single Markdown prompt that tells Claude how to port every commit above into your working tree — adapting paths and structure to match your repo. Run it via claude -p < capture-thread-916.md from inside the repo you want the changes in.

⬇ Download capture-thread-916.md