XIO

xWire

Posts from every feed QUEST has found across the web — blogs, news, podcasts, channels and more.

hive activity

Dan Notestein opened issue #67: ollama_embed retries with the same fixed 30s request timeout forever: a slow but healthy embedding server never succeeds (use...

Problem and intended result hivesense_app.ollama_embed (db/ollama.sql) gives every embedding request the same fixed timeout, urlopen(req, timeout=30) at about line 159, on the first attempt and on every retry. The backoff (5s doubling to 120s) only lengthens the sleep between attempts. So when the embedding server is alive but a batch takes longer than 30s, every attempt fails the same way: Sync path (max_retries => NULL, infinite retries): the worker loops forever on that batch. Each attempt…

Dan Notestein commented on issue #66 at hive / hivesense

Live verification (promoted to aidev/integration@548021d): working. On 548021d7, all 13 full steps passed by hand. In both slice and PCA modes, find_nearest_posts_with_embedding now returns limit 2 = [900101,900102] and continuation = [900105,900103], matching the one-shot order. The post-promote verification canary (slot full) passed.

Dan Notestein commented on issue #63 at hive / hivesense

Live verification (promoted to aidev/integration@548021d): working, after #66 (closed). Ran every full step by hand on the landed commit. stack-pins, install (now through install_with_app_lock.py), events-id, install-lock, embed-search, embed-pca and api-smoke passed. embed-slice failed: the new paging check found a real defect in slice/PCA ranking, filed and fixed as #66 (closed). Negative controls: changing UPSTREAM_BRANCH makes stack-pins fail, naming each stale image. Removing the events_id…

Dan Notestein closed issue #66: find_nearest_posts_with_embedding returns slice/PCA results in arbitrary order: reduced-path rerank has no ORDER BY and joins...

Problem and intended result In slice and PCA modes, hivesense_app.find_nearest_posts_with_embedding (the paging search, db/search.sql:347) returns posts in arbitrary order instead of nearest first. #63 (closed)'s new embed-slice check (.aidev/run-checks.sh check_paging) caught this on session/hivesense at 5137236. The query was post 900101's own first chunk, embedded with mxbai-embed-xsmall-v1, slice mode, 256 dims: one-shot order: [900101,900102,900105,900103] FAIL:…

Dan Notestein opened issue #1225: Offer Hive Keychain on the first sign-in page when the extension is installed at hive / denser

Problem and intended result The sign-in dialog (packages/smart-signer/components/auth/form.tsx) always opens on the safe-storage step (Steps.SAFE_STORAGE_LOGIN, methods/safestorage.tsx). It asks for a username, password and private key. Keychain sign-in is only on the second step, "Other sign in options" (methods/methods.tsx). That step already detects the extension with hasCompatibleKeychain() (packages/smart-signer/lib/signer/signer-keychain.ts, which checks window.hive_keychain and…

Dan Notestein opened issue #1224: Safe storage: keep the sign-in's Direct Authority Mode for an added key, and recognise one key that is both posting and act...

Problem and intended result Two gaps in how safe storage (and WIF sessions) learn which keys a session holds, left by #1212 (closed) ("let a session hold both keys"). 1. An added key ignores the sign-in's Direct Authority Mode The safe-storage sign-in (packages/smart-signer/components/auth/methods/safestorage.tsx, also safestorage-key-update.tsx) has a "Direct Authority Mode" checkbox. It is passed to hb-auth as register(..., keyType, strict). Unchecked (non-strict), a key from an account…

Dan Notestein opened issue #66: find_nearest_posts_with_embedding returns slice/PCA results in arbitrary order: reduced-path rerank has no ORDER BY and joins...

Problem and intended result In slice and PCA modes, hivesense_app.find_nearest_posts_with_embedding (the paging search, db/search.sql:347) returns posts in arbitrary order instead of nearest first. #63 (closed)'s new embed-slice check (.aidev/run-checks.sh check_paging) caught this on session/hivesense at 5137236. The query was post 900101's own first chunk, embedded with mxbai-embed-xsmall-v1, slice mode, 256 dims: one-shot order: [900101,900102,900105,900103] FAIL:…

Dan Notestein opened issue #1223: Join dates read "가입일 7월 2020" in Korean: format month-year dates per locale instead of concatenating at hive / denser

Problem and intended result The profile join date is built by concatenation, so every non-English locale gets English word order. In Korean the Blog profile shows "가입일 7월 2020" (seen live on peerverity, guest4test profile, after #1207 (closed) added ko). A Korean reader expects the year first: "2020년 7월", i.e. "가입일: 2020년 7월" or "2020년 7월 가입". Cause: packages/ui/lib/parse-date.ts dateToShow formats dayjs(...).format('MMMM YYYY') in English and then .replace()s the English month name with…

Dan Notestein opened issue #1219: Recurring workflow failure: test/test_failure in test phase (3x) at hive / denser

Recurring workflow failure pattern This appears to be task-specific — the error may be related to the task requirements. Discovery fingerprint: test:test_failure Error category: test Error subcategory: test_failure Failed phase: test Occurrences: 3 in the last 72 hours Affected workflows 1211-blog-es-and-fr-it-ru-wrong-or-sc0cae4ab-v1 1210-blog-wallet-ui-strings-hard-c-sc0cae4ab-v1 1215-blog-after-publishing-a-post-sc0cae4ab-v1 Sample errors Sample 1 full_test failed (attempt 3/3): terminal…

Dan Notestein opened issue #1218: Retry: Issue #1210: Blog/Wallet: UI strings hard-coded in English (404 page, "You Might Also Like", editor (failed: run_fu...

Failed workflow with recoverable code A workflow failed after producing code on its branch. The implementation may be worth retrying with fixes. Workflow ID: 1210-blog-wallet-ui-strings-hard-c-sc0cae4ab-v1 Branch: ai/1210-blog-wallet-ui-strings-hard-c-sc0cae4ab-v1 Original issue: #1210 Error category: test Recovery hint: Review test output for failure details. Finish-up — implementation present Implementation present on the branch; remaining work is the failing test(s) below. Diff: 119 file(s)…

Dan Notestein opened issue #1217: Retry: Issue #1211: Blog (es, and fr/it/ru): wrong or lower-case translations in navigation and sort labels (failed: run_fu...

Failed workflow with recoverable code A workflow failed after producing code on its branch. The implementation may be worth retrying with fixes. Workflow ID: 1211-blog-es-and-fr-it-ru-wrong-or-sc0cae4ab-v1 Branch: ai/1211-blog-es-and-fr-it-ru-wrong-or-sc0cae4ab-v1 Original issue: #1211 Error category: test Recovery hint: Review test output for failure details. Finish-up — implementation present Implementation present on the branch; remaining work is the failing test(s) below. Diff: 7 file(s)…

Michal Zander closed issue #150: perf: /proposals/votes times out (statement timeout) on default page size; ~6s even for small pages at hive / haf_block_expl...

Summary GET /hafbe-api/proposals/votes (hafbe_endpoints.get_proposal_votes, added in #129 (closed)) is too slow to use: it times out (HTTP 500, statement timeout) on the default page size, and even small-page queries take ~6s. This blocks block_explorer_ui#784 (switching the FE's proposal-vote reads from condenser_api.list_proposal_votes to this endpoint). Reproduction (live, testapi.hivescan.info, 2026-10-07) Request Result /hafbe-api/proposals/votes?voter=blocktrades (no page-size → default)…

Michal Zander accepted merge request !529: Fix proposal vote timeouts and separate page selection from enrichment at hive / haf_block_explorer

Proposal-vote requests can scan millions of accounts several times even when pagination selects only two votes. Select the requested page first, collect its distinct account IDs, and resolve names in one bounded lookup so these requests use the existing account indexes. Reorganize the proposal helpers into explicit page-selection and response-enrichment stages, keeping all functions and private page types in backend/endpoint_helpers/proposal.sql: Current votes: select ordered vote keys, fetch…

Dan Notestein closed issue #1212: Let one session hold both the posting and active key: each operation signs with the key it needs, adding the missing key to...

Rewritten 2026-10-15 by Dan: the scope changed from "the Blog should check the key type and offer Switch to posting key" to "a session can hold both keys and each operation uses the one it needs". @akociubinski's reproduction below is unchanged. Problem and intended result On https://denser.discuss.peerverity.info the Blog and Wallet share one session, and a session today counts as one key type. Reported by @akociubinski (develop f3f40a776d, Chrome 155, guest4test, site storage cleared first),…