feat(supplier,gateway,buyer): N-slot chat suppliers + entry-ticket settle mode

Two structural fixes for demo concurrency (3 single-slot suppliers thrashed
under Open WebUI's ~3 conversations per message):

Multi-slot suppliers (MAX_CHAT_SESSIONS, default 1):
- SupplierState is now capacity-aware: a Set of active escrow refs with
  free-until-full status semantics, ref-scoped release(ref) (also fixes the
  latent bug where handler catch-alls released another request's slot), and
  a walletMutex serializing every wallet-spending chain op — Claim, Submit,
  and wallet-health consolidation queue instead of racing the single
  working/collateral UTxO. /status gains active_sessions/max_sessions.

Entry-ticket settle mode (CHAT_SETTLE_MODE=ticket, default full, both sides):
- The buyer's escrow post at session open is the ONLY chain op. The supplier
  verifies the Open escrow as an entry ticket but never Claims/Submits
  (/v1/chat/start → {status:"ticket"}; /v1/chat/end → receipt-less
  {status:"ended"}); no hard-cap timer, so sessions may outlive deliver_by.
  Used tickets can't reopen sessions (records retained; 409 session_exists).
- Buyer SDK: startChat/endChat are settle-mode aware, driven by the
  supplier's response — mixed-mode rollout windows settle correctly.
- Gateway: ticket closes bill cost 0; the sweeper Reclaims Open tickets
  after deliver_by WITHOUT closing live sessions (the conversation and its
  affinity mirror survive their reclaimed ticket), gates reclaims on
  deliver_by, and keeps an in-process settled-refs memory so terminal
  indexer rows aren't retried every sweep forever.

Deployment (kimi-chat trio + gateway): ticket mode, 8 slots per supplier,
idle policy gateway 10 min / supplier 15 min (fixes the inversion where the
supplier idle-closed before the gateway janitor).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
david
2026-07-09 20:48:52 +02:00
parent 9a4f9f0699
commit 2f7b0194a5
34 changed files with 1451 additions and 235 deletions

View File

@@ -113,6 +113,12 @@ All on-chain steps run inside the per-key `mutex.run(...)`.
- **close** → `endChat` (Submit + Accept, awaits) → settle + bond refund; record usage; wallet-health.
- Sessions are reclaimable by the sweeper if abandoned before close.
### Settle modes (`CHAT_SETTLE_MODE`, gateway + supplier, default `full`)
- **full** — the lifecycle above: supplier Claims at start, Submits a transcript receipt at end, gateway Accepts (buyer charged).
- **ticket** — the buyer's escrow post at open is the ONLY chain op. The supplier verifies the Open escrow as an entry ticket (`/v1/chat/start` → `{status:"ticket"}`) but never Claims/Submits; `/v1/chat/end` → `{status:"ended", settle_mode:"ticket"}` with no receipt; the sweeper Reclaims the Open escrow after `deliver_by`, returning all locked funds (net cost ≈ tx fees). Usage rows record `cost_lovelace: 0`. Sessions may outlive `deliver_by` (no hard-cap) — a live session whose ticket got reclaimed keeps running; only the idle janitor/close ends it. Close behavior is driven per-session by the supplier's `/end` response, so mixed-mode rollout windows settle correctly.
- Ticket caveats (accepted, demo-grade): a supplier restart forgets used tickets, so an unexpired un-reclaimed escrow ref could reopen a session (bounded by `deliver_by` + the sweep); `/account.locked_in_escrow_lovelace` over-reports because the indexer never updates spent rows.
- Suppliers additionally take `MAX_CHAT_SESSIONS` (default 1) — concurrent session slots; `/status` reports `active_sessions`/`max_sessions` and stays `free` until all slots are taken.
## Sweeper & wallet-health
- **Sweeper** (interval ~60s): per key, classify escrows. Open/Claimed past `deliver_by` → `reclaim`. Submitted **with a verified receipt** still in accept-window → retry `acceptResult`. Submitted past accept-window → leave for supplier Release (log). Never blind-`reclaim` Submitted.
- **Wallet-health tick**: periodically + after each settle, `planConsolidate` → `consolidateWallet` so every wallet keeps a ≥5 ADA collateral UTxO. Skip if already healthy.