Slice 1 polish: archive dormant code, merge work-os into primitives, add futures

- Archive dormant apps (daemon, desktop, native, tui) and superseded
  packages/server into repos/ for reference
- Archive 21 dead web components (chat page shell, work-os cluster, strays)
  into repos/web/; keep reused message-rendering primitives in components/chat
- Merge @code/work-os into @code/primitives; work.ts absorbed via ./work
  subpath and barrel export; update all four importers
- Add docs/futures.md tracking audio/video (blocked at Flue + provider),
  web layout cleanup, and message rendering quality
- Repoint dev:zopu script to @code/agents (the live Flue server)
- Note repos/ reference-code convention in AGENTS.md

docs/futures.md

@@ -0,0 +1,22 @@
+# Futures
+
+Work that is intentionally deferred from the current slice. Each item records enough context that a future thread can pick it up without re-investigating.
+
+## Audio and video input
+
+Status: blocked at two layers; not viable through the current transport.
+
+MiMo V2.5 lists audio and video in its model spec, but the live stack cannot carry either modality today:
+
+- Flue agent transport exposes only `images: AgentPromptImage[]` on `AgentPromptOptions`, with `AgentPromptImage.type` pinned to `'image'`. There is no `audio` or `video` field, so the client cannot attach either.
+- The Cheaptricks gateway rejects non-image content parts upstream. Probed directly with OpenAI-style `input_audio` and `video_url` parts, both returned `"Error from provider (Console Go): Upstream request failed"` while a plain text probe on the same endpoint returned a normal completion.
+
+Cheapest real path, if this is revisited: video via client-side frame extraction sent as image attachments (Flue already supports image input). True audio/video would need a Flue provider adapter (Provider API) plus an upstream that actually accepts the modality, which the current probe suggests it does not.
+
+## Web layout cleanup
+
+Polish pass on the mobile-first layout: spacing, alignment, responsive edge cases, and removing dead or vestigial UI introduced during the slice-one build. Goal is a tightened, consistent layout before slice two adds more surfaces.
+
+## Message rendering and streaming quality
+
+Improve the read experience for streamed assistant messages: markdown fidelity, reasoning-trace presentation, code block and Mermaid rendering stability, and the perceived smoothness of token streaming. Covers both correctness of the rendered output and the feel of the live update.