Repository navigation
Logbook October 2026
Week of Monday 28th — Building the policy-and-asset query on the utxo-indexer socket, from first answer to a trustworthy one. Thursday landed the query itself, then made every answer state its network, coverage and freshness so an empty result is not read as absence, and made the asset index survive restart and upgrade of older stores. Friday closed the loop with a store that records its build coverage, fault-injection checks, and atomic multi-read views over one indexed snapshot. Alongside, the devnet harness got isolated working directories and a restart hook.
The utxo-indexer gained a query for live outputs by policy ID and asset name, answered on its socket from a new index of outputs by asset (#204). Answers now state the network, the coverage of the index and its freshness, so an empty result cannot be mistaken for proof of absence (#206). The asset index is kept across daemon restart, and stores built before the asset index existed are upgraded in place, with a CLI rebuild available (#205). Along the way, stored outputs whose coin needs eight bytes were fixed to decode, and a warm-boot restart on a populated database that never resumed was repaired (#214).
The devnet no longer force-removes a fixed temporary directory, so concurrent runs stop deleting each other's node data; each run gets its own working directory (#209). The restartable node can run a hook while it is down, letting a test mutate the node database between stop and start (#213). The start-point setting is documented as cold-boot only, and the warm-boot no-intersection error names both routes out (#208).
Issues: #207, #203, #210, #220
The store now records the coverage it was built with, so reopening with a different start point or filter cannot misreport it, and asset answers read that record (#215). Fault-injection CI checks were added for asset matching and for snapshot binding (#216). The library gained address and asset reads at one indexed point (#217), and the socket serves several such reads in a single request from one index snapshot (#221). A follow-up PR for acquiring local state queries at a specific chain point is open (#222).