Octa chat — exact editable surface

Verbatim copy of the Odysseus chat CSS (real .msg/.role/.body classes + rendered markdown), n2-preset vars resolved. Tuning here maps 1:1 to odysseus/static/style.css.

Message bubbles & roles

→ .msg.msg-user / .msg.msg-ai / .role / .body

Isiah

How do I switch the default chat model to something lighter?

Octa

Set the default to gpt-5.4-mini — fast and low-token while staying tool-capable. There's no gpt-5.5-mini.

Markdown body — kitchen sink

→ .msg .body h1-h6, strong, em, code, pre, blockquote, ul/ol, a, table

Octa

Heading 1

Heading 2

Heading 3

Heading 4

Body with bold, italic, ~~strikethrough~~, an inline link, and inline code.

async function chat(msg) {
  const res = await fetch('/api/chat', {
    method: 'POST', body: JSON.stringify({ msg })
  });
  return res.json(); // → { response }
}

A blockquote — for cited or set-apart text.

Unordered item one
Unordered item two

Ordered item one
Ordered item two

Model	Speed	Cost
gpt-5.4-mini	fast	low
gpt-5.5	slow	high

Thinking block

→ .thinking-section

Octa

Thinking

gpt-5.4-mini fits: fast, low token cost, still tool-capable. Confirm it's on the endpoint first.

Agent block

→ .agent-thread / .agent-thread-cmd / .agent-thread-node

Octa

✓ bash · ls /app/static/fonts/custom

FragmentMono-Regular.woff2
PP-NeueMontreal-Medium.woff
Svatopluk-Regular.woff2

Chat input bar

→ .chat-input-area / .chat-controls-row / .chat-input-form

Message Octa…

AgentChat

Welcome screen

→ .welcome-name / .welcome-sub

Octa

Get more out of real life.