m's reframing 2026-05-07 20:56: Paliadin is "mostly for myself now
but can be expanded — monitoring use." Two-stage shape replaces the
single-PR production-v1:
- Phase 0 (PoC): tmux-Claude pattern lifted from goldi/mVoice
(mVoice/server.py:250-380). Claude Code window in a long-lived
tmux session, prompts via tmux send-keys -l, response via
/tmp/paliadin/{turn_id}.txt tail-f → SSE relay. Single user (m),
m's laptop only (PALIADIN_ENABLED=false on prod). ~600-900 LoC,
~1 day. Migration 057 (PoC variant) stores full prompt + response
for monitoring — no redaction at this scope.
- Phase 1 (production v1): the original §2-§6 Anthropic API design,
GATED on PoC success per §0.5.7 expansion criteria (≥3 turns/wd,
≥50% tool-use rate, 4 weeks).
§0.5 (new) inserted as the load-bearing PoC spec. §7 leads with the
two-stage frame. §8.5 questions split into PoC-relevant (Q-PoC-1..6)
and production-v1-deferred. youpc case-law lookup promoted to
Q-PoC-6: m himself does case-law research, so include it from day
one (cross-schema SELECT into data.judgments is technically trivial
since paliad and youpc share the same Postgres).
What we drop for PoC: Anthropic API client, BYO-AI, rate limit,
token caps, multi-user RLS edge cases, /admin cost dashboard,
compliance disclosure, most i18n keys.
What we keep: system prompt voice, citation discipline (best-effort),
visibility gate (Claude is required to use paliad.can_see_project()
in queries), /paliadin surface, SSE shape, audit table.
The two-stage shape protects against the t-145 pattern: ship cheap,
observe, decide. No 4500-LoC investment based on m's gut feel about
adoption.
Inventor design pass for the Paliadin: a Claude-backed conversational
assistant grounded in the user's own paliad data + paliad's static
reference (courts, glossary, deadline rules, Fristenrechner concept
tree). Long-lived in-process Go service that calls Anthropic's
Messages API directly with tool use; every tool is a thin shim over
an existing service (Dashboard / Project / Deadline / Appointment /
Court / Glossary / DeadlineRule). RLS / visibility inherited from
those services — Paliadin literally cannot see what the caller cannot.
Five coordinated sub-designs answer the issue's 20 open questions:
A. LLM architecture + tool-use + prompts (§2)
B. Data access + RLS + PII (§3)
C. UX (§4)
D. Token budget + cost + audit (§5)
E. Phasing (§7)
Phase 1 v1: /paliadin full page + sidebar entry, SSE stream of
Anthropic, 7 read-only tools, session-only history, 30/hour user cap
+ 1000/hour global cap, audit row per turn (metadata only — no
transcript), 4k input + 2k output token caps, no avatar/mascot, no
proactive onboarding. Migration 057 introduces paliadin_turns +
paliadin_rate_limit. Single PR, ~3500-4500 LoC.
mlex / /lex-* reuse: shape (system-prompt voice, tool-catalog idea,
citation style) — NOT code. mLex is a workspace, not a Go/TS repo;
the /lex-* skills drive Claude against youpc's MCP and cannot be
embedded in a paliad service.
Premise verifications surfaced one CLAUDE.md doc-bug (the
ANTHROPIC_API_KEY "Reserved for Phase H — do not set" row needs to
flip in the implementation PR — Paliadin un-defers it).
12 open questions for m in §8.5 — Anthropic key choice (personal vs
HLC enterprise), default model (Sonnet vs Haiku), surface
(/paliadin page vs drawer), mascot phase, 2-PA sanity check before
locking scope, etc. Same adoption-risk concern that just parked
t-paliad-145 — Paliadin's edge over open-Claude-in-another-tab is
data grounding, which only works if v1 makes it visible (citation
chips + tool-call evidence + tagline).
STOP after design. Awaiting m go/no-go before coder shift.
