Repository navigation
Write the ADR for chat modes #2220
Description
No ADR records the chat-modes decision built in #2207:
- Basic (Q&A) and Agentic are chosen per chat, and a chat keeps its mode.
- No measured score keeps a model out of Agentic. Only technical gates do: no opencode, a stated no on tool calls, a stated window under 32,768 tokens.
- The default is Agentic for a model measured at
agenton a remote host, and Basic otherwise, local copies included. - A pass holds only where it was measured; expensive flagships are
assumed.
The file-agent README's planned ADR 0044 ("What a model may do comes from a reviewed, measured capability manifest") describes the opposite of what was built, so it needs rewording or superseding. What is built is in docs/architecture/model-capabilities.md.
Done when: an ADR under docs/adr/ records the decision and why, and the planned 0044 entry matches it.
Activity
Cedric921 commentedon Oct 14, 2026
I'd like to take this. I'll write the chat-modes ADR from what #2207 built and
docs/architecture/model-capabilities.mddescribes:agenton a remote host, and Basic otherwise;assumed.The file-agent README's planned 0044 entry describes the opposite, so I'll reword it to match. A planned ADR has no accepted text to supersede.
docs(adr): record the chat-modes decision
MODSetter commentedon Oct 24, 2026
Fixed by #2266, merged into
dev, by @Cedric921: ADR 0044 records the decision, with the proposal's planned row matching it.Closing by hand: GitHub only auto-closes on a merge to the default branch, and we merge through
dev.