serversdown/project-lyra
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity

docs: roadmap — mark monolith deleted, classifier hardened, Phase C flip-ready

Browse Source 
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
serversdown
2026-07-05 03:30:10 +00:00
parent 800cab8d36
commit 56fb6d9a85
1 changed files with 11 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/ROADMAP.md
+11 -8
View File
@@ -27,14 +27,17 @@ Spec: `docs/superpowers/specs/2026-07-01-poker-prompts-design.md`
`classify(msg, roster_handles)` (READ|HAND|TABLE|MENTAL|STATUS|LOG|CHAT), lean `classify(msg, roster_handles)` (READ|HAND|TABLE|MENTAL|STATUS|LOG|CHAT), lean
always-on `BASE`, per-type `FRAGMENTS`. Wired into `build_messages`; the always-on `BASE`, per-type `FRAGMENTS`. Wired into `build_messages`; the
~100-line `_CASH_CARD` monolith is sharded out (`CASH.card=""`). ~100-line `_CASH_CARD` monolith is sharded out (`CASH.card=""`).
- ⬜ **Delete the dead `_CASH_CARD`** in `modes.py` once the fragments are proven - ✅ **Deleted the dead `_CASH_CARD`** monolith (modes.py 256→161 lines).
in a live session (kept now only as the distillation source). - ✅ **Classifier hardening (round 1).** Fixed 6 real gaps found by probing
- ⬜ **Classifier hardening.** It's a heuristic — will miss weird phrasings. live-style phrasings (-ing action forms, "the whale" bare descriptor, player
`classify()` is the swappable seam; upgrade to an LLM / fine-tuned-MI50 classifier departures→TABLE, "stack" leaking questions into LOG, thin MENTAL lexicon). 18
IF real-session misses justify it. Tune against actual transcripts first. unit tests. Still a heuristic + swappable seam — upgrade to an LLM/MI50
- ⏸ **Phase C — MI50 tool-calling.** Flip `TOOL_BACKENDS` to include `mi50`. classifier only if live misses justify it; keep tuning against real transcripts.
Blocked on the MI50 llama.cpp server running with `--jinja` + a tool-capable - 🔨 **Phase C — MI50 tool-calling (code done, flip pending).** Enabling it is now
model loaded. a config toggle: `cfg.tool_backends` (env `TOOL_BACKENDS`, default "cloud"). To
turn on: MI50 llama.cpp up with `--jinja` + a tool-capable model, then set
`TOOL_BACKENDS="cloud,mi50"`. NOT flipped — MI50 is currently offline, `--jinja`
unconfirmed. No code change needed when ready.
## Persona (the "person" layer) ## Persona (the "person" layer)