The SDK dropped minTotalTokens/maxTotalTokens from AggregateUsageOptions
(routstr-sdk c98de6b), so replace the five aggregate() calls with a single
list() and bucket entries in-process.
routstrd providers list --refresh now:
- Re-fetches Nostr kind 38421 provider discovery events
- Re-fetches Nostr kind 38423 routstr21 model list
- Re-fetches Nostr kind 38425 review events (applies LGTM-based disable)
- Fetches models from all discovered providers
- Syncs fresh provider list and disabled status into the store
Usage: routstrd providers list --refresh
- Import ModelManager from @routstr/sdk/bun instead of @routstr/sdk
- Import storage helpers from @routstr/sdk/storage/bun instead of @routstr/sdk/storage
- Fix createBunSqliteUsageTrackingDriver call: async and no longer needs
manual bun:sqlite import (now handled internally by the sdk entrypoint)
- Verified sharded discovery adapter setup matches SDK pattern in
scripts/routstr-daemon.ts at 6077aa7
- Add optional contextWindow field to PiModelEntry type
- Map model.context_length to contextWindow when available
- Add context_length to RoutstrModel type in registry
Adds an authUrl field to RoutstrdConfig so CLI management commands
(npubs, clients, usage) can be directed to the routstrd-auth proxy
instead of the daemon. This fixes 'routstrd npubs list' failing with
'Only POST is supported' when the auth proxy and daemon run on
separate ports (e.g., Cloudron setup: proxy on 8008, daemon on 8009).
Changes:
- config.ts: add authUrl to RoutstrdConfig
- daemon-client.ts: add getAuthBaseUrl(), callAuth(), refactor callDaemon
- cli.ts: npubs/usage commands use callAuth, remote command accepts --auth-url
- clients.ts: clients commands use callAuth
The logger utility in this project is a file-only logger — every call to
logger.log() / logger.error() writes to ~/.routstrd/logs/YYYY-MM-DD.log
and produces no terminal output whatsoever. This is intentional and correct
for the background daemon (daemon/, start-daemon.ts), which has no attached
terminal.
However, the onboarding and integration code was mistakenly using logger.log
instead of console.log for all of its user-facing output:
- setupIntegration() printed the 1-6 integration menu with logger.log, so
the user only ever saw the bare prompt 'Select integration [1]: ' with no
options above it. The selection appeared to be broken.
- Every install*Integration function (opencode, openclaw, claudecode, hermes,
pi) reported all progress and results with logger.log, so after picking an
integration nothing appeared to happen.
- initDaemon() used logger.log for 'Initialization complete!' and the
follow-up wallet instructions, so those were also silently dropped.
- installCocodOrExit() used logger.log for the 'cocod not found, installing'
message and the success/failure result.
Fix: replace logger.log/logger.error with console.log/console.error in all
CLI-facing code paths (cli.ts, integrations/index.ts, and all five
integration installers). Remove the now-unused logger import from the five
integration files. The logger import in integrations/index.ts is kept because
refreshModelsAndIntegrations() is called by the daemon's internal scheduler
and correctly logs to file there.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add cooldownMinutes field to /nwc/status and /nwc/auto-refill responses
- Change auto-refill log from ms to minutes display
- Keep cooldownMs internally for calculations
Two bugs caused the auto-refill loop to retry every 5s regardless of outcome:
1) Backoff used lastRefillAt which was only set on success. On failure
it stayed at 0, so now - 0 < backoffInterval never blocked once the
daemon had been running > 30s. Switched to a separate lastAttemptAt
that is updated on every attempt (success or failure).
2) preimage.slice() crashed when NWC returned a payment result without
a preimage field. This turned successful payments into exceptions,
triggering retries and causing the wallet to be charged repeatedly.
Guard preimage with an if-check before calling .slice().
- Add reconnect() method to wallet adapter that tears down old
WalletConnect/RelayPool and creates a new one on the fly
- Auto-refill loop now uses a getWallet() getter so it always
references the latest wallet instance after reconnect
- POST /nwc/connect and /nwc/disconnect handlers call reconnect()
after saving config, no restart needed
- Remove 'Run routstrd restart' messages from CLI connect/disconnect
Changed startAutoRefillLoop to accept a config getter instead of a static
config object. The getter reads config from disk on every check cycle, so
CLI changes (nwc auto-refill on/off) take effect immediately.
- auto-refill.ts: config parameter replaced with getConfig() callback
- wallet/index.ts: always start loop when wallet exists, use getter
- daemon/index.ts: pass getAutoRefillConfig reading from disk sync
- config-store.ts: added loadDaemonConfigSync() for sync disk reads
- http/index.ts + cli.ts: removed 'Restart daemon to apply' messages
- Add step-by-step logging to fundFromNWC (initial/final balance, invoice,
payment status, fees) matching pay_invoice.mts pattern
- Remove invoice and preimage from POST /nwc/fund response
- Add final wallet balance to the response after funding completes
- auto-refill.ts: use WalletConnect instance directly instead of
NwcClient + RelayPool; check wallet.service for connection state
- wallet/index.ts: create single shared WalletConnect for auto-refill,
fundFromNWC, and getNwcStatus; remove NwcClient entirely
- daemon/index.ts: strip all NwcClient creation/connect/logging/cleanup;
auto-refill now gates on connectionString alone
- add applesauce-wallet-connect dependency
- Encryption auto-detection: reads wallet info event (kind 13194) on connect
to choose NIP-04 vs NIP-44 per NIP-47 spec
- Subscription readiness: wait for EOSE before resolving connect()
- Dual NIP-04 / NIP-44 encrypt/decrypt support
- Proper encryption tags on NIP-44 requests
- test_nwc.ts: standalone NWC test script for make_invoice + pay_invoice
Two bugs caused NWC responses to never arrive:
1. getPublicKey() in nostr-tools v2 already returns a hex string, but
the code wrapped it with Buffer.from().toString('hex'), producing a
double-encoded 128-char string. The relay rejected the REQ filter
with 'filter item too large' because the #p tag value was invalid.
2. The 'secret' from the NWC connection string (the pre-shared client
private key per NIP-47) was parsed but never used. Instead a new
random keypair was generated on every restart, so the wallet service
would see requests from an unexpected pubkey.
Also removed 'since' from the subscription filter (simpler, some relays
choke on it) and added debug logging for the REQ filter.
redshift
redshiftandGitHub
Bilthon
Jeroen UbbinkandClaude Sonnet 4.6