* feat: add bot index analysis entry
* test: add Bot index online smoke
* fix: guard Windows-only spawn flag in smoke tests for Linux CI
* fix: clean up worker tree on parent interrupt in smoke runner
* fix: surface interrupt cleanup failures in smoke runner
---------
Co-authored-by: zhulinsen <42829555+ZhuLinsen@users.noreply.github.com>
* Exclude litellm 1.99.0 from allowed versions
litellm 1.99.0 (released 2026-09-01T00:42:41Z) now forwards the
prompt_cache_key argument to the OpenAI provider. The provider cache
contract test asserts the key is not passed through without verified
capture, so CI backend-tests (3/3) and the downstream backend-gate
went red on every branch right after the release, unrelated to any
code change.
Local capture server reproduction: with 1.98.0 the captured request
body carries no prompt_cache_key; with 1.99.0 it carries
prompt_cache_key=cache-key. This one-line exclusion follows the
existing quarantine pattern for 1.82.7 / 1.82.8 and keeps the tested
behaviour until the contract is adapted deliberately.
* Bound litellm below the 1.99.0 behavioral change
Pin the dependency window to a verified upper bound (latest 1.98.x)
instead of quarantining only 1.99.0, and sync every doc that hardcodes
the litellm constraint string so docs cannot drift from requirements.txt
again.
* Sync .env.example runtime constraint comment to the litellm upper bound
* feat: make self-hosted SearXNG search timeout configurable
SEARXNG_TIMEOUT_SECONDS (default 10s, min 1) now controls the per-search
timeout for self-hosted SearXNG instances, threaded through all three
SearchService construction sites. Public-instance timeout stays fixed.
* fix: thread SEARXNG_TIMEOUT_SECONDS through subprocess rebuild and Actions
- persist searxng_timeout_seconds in _constructor_kwargs so the bounded
topic-news subprocess rebuild no longer silently falls back to 10s
- map SEARXNG_TIMEOUT_SECONDS (vars||secrets) in 00-daily-analysis.yml
- document the variable in full-guide/full-guide_EN data-source tables,
add CHANGELOG entry, and add regression tests for both paths
* fix: document SearXNG timeout environment contract
* fix: register SearXNG timeout setting
* fix: keep SearXNG timeout env-only
---------
Co-authored-by: zhulinsen <42829555+ZhuLinsen@users.noreply.github.com>
Co-authored-by: zhulinsen <zhuls97@163.com>
* feat: add one-shot index analysis via --stocks with structured AnalysisTarget pipeline
* fix: render index data sources in aggregate reports
* fix: refresh index registry before --stocks parsing and filter indices on CN holidays
---------
Co-authored-by: zhulinsen <42829555+ZhuLinsen@users.noreply.github.com>
* feat: add Futu OpenD as an optional HK realtime and fundamental data source
Add FutuFetcher and FutuFundamentalAdapter behind FUTU_OPEND_HOST/PORT,
register the settings in Config and config_registry so the Web settings
page can expose host, port and HK realtime priority, and route HK
realtime quotes through a configurable futu/longbridge/akshare/yfinance
order while keeping A-share priority untouched. Include offline tests
for the adapter, config schema and HK routing/fallback, plus docs and
CHANGELOG entries.
* fix: wire Futu fundamentals into HK pipeline and restore quote supplementation
- _fetch_offshore_fundamental_bundle() prefers the Futu fundamental
adapter for HK when FUTU_OPEND_HOST is configured, and falls back to
yfinance when Futu is absent or returns no usable content.
- HK realtime priority loop now supplements missing quote fields
(volume_ratio / turnover_rate / pe/pb / market cap) from later
configured sources instead of returning after the first non-empty
quote, matching the US path's _supplement_quote behavior.
- capital_flow / boards blocks are filled from the Futu bundle for HK
instead of being hard-coded not_supported; status and missing_fields
aggregation updated accordingly.
- Add regression tests for partial-quote supplementation and Futu
fundamental bundle routing/fallback.
* test: expect boards block ok when bundle provides belong_boards
The Futu integration made the offshore boards block data-driven instead
of hard-coded not_supported; update the existing US/HK fundamental
context test to match (belong_boards from the bundle now surface as an
ok boards block).
* fix: preserve HK fallback_from metadata and normalize Futu quote timestamps
- HK realtime priority loop now records the failed preferred source token
and passes it as fallback_from when a later source takes over, so the
pipeline and analysis context can mark the quote as degraded.
- Futu snapshot update_time is a naive Beijing-time (UTC+8) string; attach
the +08:00 offset before storing provider_timestamp so stale_seconds /
is_stale / provider_timestamp freshness semantics are correct instead of
being parsed as UTC.
- Add regression tests for fallback_from propagation and timestamp
normalization.
* fix: normalize Futu belong_boards to name/type/code contract
OpenD owner_plate returns plate_code / plate_name / plate_type, but DSA
downstream consumers (notification, extract_board_detail_fields, market
structure) only read name/type/code. Map the fields in
FutuFundamentalAdapter._boards so HK Futu boards are actually consumed
instead of silently dropped, and add regression tests including an
end-to-end check through extract_board_detail_fields.
* fix: merge yfinance bundle when Futu fundamental returns partial blocks
Futu partial success (e.g. statements failed but static info worked) used
to short-circuit the whole bundle, silently dropping the growth/earnings
that the existing yfinance path could still provide. Now, when Futu
returns content but is missing growth or earnings, fetch the yfinance
bundle within the remaining budget and merge the missing blocks
(growth/earnings/institution/capital_flow/belong_boards), keeping
Futu-preferred values where both exist. Add regression test for the
partial-success merge path.
* fix: use field-level checks when deciding Futu-vs-yfinance growth/earnings
The previous merge condition only checked dict truthiness, so a truthy
growth/earnings shell (all-None core values or metadata-only keys such
as report_date/period/currency) would skip the yfinance supplement and
silently downgrade existing HK fundamentals. Add _earnings_block_has_values
(a core numeric field or a populated dividend is required) and reuse the
existing _has_meaningful_payload for growth; both the missing_core check
and the merge loop now use these. Add regression test for the
all-None-shell scenario.
* fix: fill HK fundamental field gaps from yfinance instead of block-level checks
Block-level meaningful checks still skipped the yfinance supplement when
Futu hit only part of the growth/earnings fields (e.g. revenue_yoy but
None net_profit_yoy, or earnings with only basic_eps), silently dropping
fields the main branch used to provide. Replace the missing_core decision
with a per-field gap list (growth: revenue_yoy/net_profit_yoy/gross_margin;
earnings.financial_report: revenue/net_profit_parent/basic_eps/gross_profit)
and make the merge field-level: keep Futu values, fill each missing field
from yfinance. Add regression tests for partial-hit and all-None shells.
* fix: normalize Futu dividends to the repo contract and treat dividend gaps as supplement triggers
Futu OpenD dividend_list carries raw fields (statement/ex_date/record_date)
which the notification/data_processing market-structure consumers do not
read; the repo contract is ttm_cash_dividend_per_share,
ttm_dividend_yield_pct and events[].cash_dividend_per_share /
ex_dividend_date / event_date. Normalize events in
FutuFundamentalAdapter._dividends_and_splits, compute TTM count/cash and
yield from the latest quote, and teach _field_gaps/_merge_bundles to treat
a dividend block that does not satisfy the contract as a gap so yfinance
supplements it. Also dedupe FUTU_OPEND_HOST/PORT in full-guide_EN.
* fix: read dividend yield price from UnifiedRealtimeQuote objects
FutuFetcher.get_realtime_quote returns a UnifiedRealtimeQuote dataclass,
not a dict, so the yield branch in _dividends_and_splits that guarded on
isinstance(quote, dict) never ran on the live Futu path, silently dropping
ttm_dividend_yield_pct while the contract check considered the dividend
block complete. Read price via getattr(quote, 'price', None) and keep the
dict fallback for other fetchers; add a regression test driving the real
UnifiedRealtimeQuote shape.
* fix: treat dividend blocks with TTM cash but no yield as supplement gaps
The repo contract consumes ttm_cash_dividend_per_share and
ttm_dividend_yield_pct together. When the Futu dividend path has events
and TTM cash but the extra realtime price snapshot failed (quote None /
no price), ttm_dividend_yield_pct cannot be computed and the block was
previously treated as complete, so yfinance was never consulted and the
notification rendered the yield as N/A.
_dividend_contract_has_values() now requires the paired yield whenever
TTM cash is present, so _field_gaps() triggers the yfinance supplement
and _merge_bundles() replaces the incomplete dividend block.
Add regression tests for the adapter-level gap shape (quote unavailable
leaves no yield) and the manager-level supplement path (Futu cash
without yield pulls yfinance and fills the yield).
* fix: skip unconfigured Futu in HK realtime routing
When FUTU_OPEND_HOST is not configured, the HK realtime priority loop
used to still attempt the futu source, record it as the failed primary,
and attach fallback_from='futu' to a successful quote from the next
enabled source (longbridge/akshare/yfinance). Consumers then wrongly
treated an enabled source's first success as degraded fallback data,
contradicting the documented contract that Futu only participates when
OpenD is configured.
The HK loop now checks FutuFetcher.has_configured_endpoint() once and
skips the futu token entirely when it is disabled, so no fallback_from
is written. Existing configured-Futu routing tests explicitly patch the
endpoint check; a new regression test asserts an unconfigured Futu is
never called and the enriched quote carries fallback_from=None.
* fix: release cached HK Futu fundamental fetcher in DataFetcherManager.close()
The HK Futu fundamental path lazily creates and caches its own
FutuFetcher (an OpenQuoteContext-backed OpenD connection) on
_futu_fundamental_fetcher, but close() only released the TickFlow
fetcher and the default fetchers snapshot. Explicit close / reload
paths therefore left the OpenD connection hanging.
close() now takes the cached _futu_fundamental_fetcher, clears the
reference and calls its close() best-effort. A regression test injects
an observable fetcher into _futu_fundamental_fetcher and asserts
close() invokes it and clears the attribute.
---------
Co-authored-by: BayMax local review <baymax-local@invalid>
* fix: render CJK text in Linux share images
* fix(review-feedback-2270): 补到 PR 附件或评论里，和本次渲染修复说明保持一致
* fix: preserve report language contract in share images
* feat: support custom Tushare Pro endpoint via TUSHARE_HTTP_URL
Add TUSHARE_HTTP_URL so the Tushare data source can point at a self-hosted
or third-party compatible endpoint when the official api.tushare.pro is not
reachable. Defaults to the official host when unset, so behavior is
unchanged for existing users.
- data_provider/tushare_fetcher.py: add _resolve_tushare_http_url() helper
(strip + http(s):// schema validation) and forward the resolved URL into
_TushareHttpClient, with an info log when a custom endpoint is in use
- .env.example + .github/workflows/00-daily-analysis.yml: document and map
TUSHARE_HTTP_URL so the new option is wired into the daily job without
leaving a half-configured state
- tests: cover env parsing (empty/whitespace/http/https/missing schema),
fetcher fall-through to the official host, and end-to-end POST target
- docs/CHANGELOG.md: flat [Unreleased] entries
Fixes#1985
* refactor: drop unnecessary string-literal type hint in _build_api_client
TushareHttpClient is already defined above TushareFetcher in the module
scope, so a string-literal type hint is not needed for forward reference.
Restore the bare type to match the surrounding code style and reduce the
diff against main.
* docs(tushare): 补 TUSHARE_HTTP_URL 在 full-guide 中英版本的用途/默认行为/workflow 映射说明
按 PR #2048 review 反馈补齐：
- 表格内新增 TUSHARE_HTTP_URL 行（中英文版本同步），明确默认 https://api.tushare.pro 与 http(s):// 前缀要求
- 在 GitHub Actions 段落后补充 TUSHARE_HTTP_URL 的 vars/Secrets 优先级与每日 workflow 0映射说明，与现有非敏感配置（TICKFLOW_PRIORITY）一致
- 完整环境变量列表（中英文版本）补 TUSHARE_HTTP_URL 行，默认值列填 https://api.tushare.pro
- 与代码实现一致：00-daily-analysis.yml 已用 vars.TUSHARE_HTTP_URL || secrets.TUSHARE_HTTP_URL 映射
* docs(tushare): align TUSHARE_HTTP_URL default with runtime and clarify vars/secrets precedence
Per review feedback on PR #2048: the per-repo config contract must stay
consistent across runtime, .env.example, workflow priority, tests and
both zh/en guides. Two fixes applied as a single contract update:
1. Default endpoint alignment. data_provider/tushare_fetcher.py:100,
.env.example, and tests/test_tushare_fetcher_http_client.py all keep
the existing official endpoint http://api.tushare.pro, but the zh/en
full-guide rows had drifted to https://api.tushare.pro. Switching the
documented default to HTTPS would silently change the runtime contract
that the feat commit explicitly preserved. Revert both zh and en
guide rows to http://api.tushare.pro so docs match runtime, .env.example
and the test assertions.
2. vars/secrets precedence wording. The workflow uses
'vars.TUSHARE_HTTP_URL || secrets.TUSHARE_HTTP_URL', which means a
non-empty vars entry always wins and Secrets cannot override it. The
zh/en notes previously suggested 'Secrets as a tamper fallback' which
is incorrect under this precedence and can mislead users into thinking
Secrets has override power. Replace with explicit description of the
real semantics: vars wins when non-empty; Secrets is only selected
when the Variable is empty; for a tamper-resistant deployment put the
value only in Secrets and leave Variables empty.
Both zh and en guides are updated together; the same 6 contract surfaces
(runtime / .env.example / workflow priority / tests / zh guide / en guide)
now describe one consistent contract.
* docs(tushare): remove false Secret-as-tamper-guard claim, document real vars/secrets write-permission model
* feat: add reliable home watchlist workspace
* fix(review-feedback-1984): Don't trust stale stock-bar rows after refresh failures
* fix(review-feedback-1984): apps/dsa-web/src/stores/stockPoolStore.ts 的 refreshStockBar
* fix(review-feedback-1984): Clear loading when refresh supersedes initial stock-bar load
* fix(review-feedback-1984): Type stock-bar test fixtures as StockBarItem and 跟进结论 - 结论 ：不接受；最新
* fix(review-feedback-1984): tying the Today fetch to the same refresh path while this tab is
_CLOSING_AUCTION_WINDOW_MINUTES had no tw entry, so `.get(market, 0)` gave a
zero-width window and infer_market_phase("tw") could never return CLOSING_AUCTION:
the last tick before 13:30 stayed INTRADAY and 13:30 flipped straight to POSTMARKET.
TWSE/TPEx run a 5-minute closing call auction (13:25-13:30), so add "tw": 5.
+ a phase-boundary regression test (no lunch break, 13:30 half-hour close):
13:24 INTRADAY / 13:25-13:29 CLOSING_AUCTION / 13:30 POSTMARKET.
* feat(market): surface tw 三大法人 (institutional flows) in the tw report institution block
Wire the merged TwInstitutionalFetcher (#1829/#1841/#1855) into the tw analysis
report: in _build_offshore_fundamental_context, for market == "tw" only, the
institution block now carries the raw 三大法人 net buy/sell figures (foreign /
trust / dealer / total, unit=shares) from TWSE T86 / TPEx instead of the hardcoded
not_supported. v2 scope confirmed by the maintainer on issue #1777.
- tw-only + strictly additive: cn/hk/us/jp/kr offshore flows are byte-identical
(the not_supported loop just excludes institution; non-tw still gets the same
not_supported block). Pinned by tests for us/hk/jp/kr.
- fail-open + default-on: any error or no-data -> not_supported, never interrupts
the analysis. Wiring (import/construct) failures log at error (visible) but still
fail open; fetch failures log at warning.
- status "ok" only when all four core nets are present (a genuine 0 is kept).
- raw figures only: no capital_flow_signal / scoring / weight / schema / Web.
Tests: tests/test_tw_institution_report_wiring.py (9) covers data->ok, genuine-zero,
None/raise/init-raise fail-open, missing-core-net->not_supported, us/hk/jp/kr
byte-identical + fetcher-unused, and no-derived-key. + CHANGELOG.
Dual review (code-reviewer + silent-failure-hunter): the hunter caught the combined
try swallowing wiring (import/construct) errors silently at info -> split into
loud-but-fail-open (error log + still not_supported, honoring #1777's
never-interrupt rule) and guarded "ok" against a null core net.
* fix(market): address Codex review on tw institution wiring (#1863) — stage timeout, overall status, docs
- P1: run the tw institution fetch under the fundamental stage/fetch budget via
_run_with_retry (like the other offshore blocks), so a slow / rate-limited TWSE/TPEx
call fails open at the deadline instead of pushing the analysis past it.
- P2: count a present tw institution toward the OVERALL fundamental status so a report
with only 三大法人 data still surfaces fundamentals (consumers key off the top-level
status). missing_fields keeps the original three blocks, so offshore markets without
institution data stay byte-identical.
- P2 (docs): sync docs/market-support.md (drop the now-false "data-layer only / not
wired into report" clause) + add a tw institution note to docs/full-guide.md and
docs/full-guide_EN.md (AGENTS.md requires user-visible report changes to update guides).
Tests: + stage-timeout (a 2s fetch is abandoned at the ~0.3s budget, fail-open) +
overall-status surfacing assertion. 10 tw-wiring tests + offshore/tw/jp-kr suites green.
* feat(i18n): add Korean to report language label maps
Extend the report language layer with a third code `ko` so reports can
render deterministic labels in Korean. Adds `ko` to the supported list,
aliases (korean/kr/ko-kr), every translation map, the full report label
set, sentiment bands, the config registry enum and `.env.example`.
`zh`/`en` behavior is unchanged and unknown languages still fall back to
the default. Prompt and Web surfaces follow in later changes.
Refs #1614
* feat(i18n): emit Korean output directives in analysis prompts
Make the analysis and market-review prompts produce Korean output when
REPORT_LANGUAGE=ko. The decision agent now appends a Korean
output-language directive (JSON keys and decision_type enum unchanged),
and the market-review prompt reuses the English structural scaffolding
while instructing the model to write the shell, headings and conclusion
in Korean. Market-phase and context-pack prompt sections route ko to the
English structural base. The market-review payload keeps the truthful
language code via a dedicated output-language helper, leaving zh/en
behavior unchanged.
Refs #1614
* fix(i18n): localize phase and market-context guardrails for Korean
Route Korean reports through the English structural scaffolding for
market-phase and market-context prompt sections, and add Korean output,
detection markers, negations and recap patterns to the phase-decision and
daily-market-context guardrails so they operate on Korean model output.
Confidence and operation labels now flow through the shared localize
helpers. zh/en behavior is unchanged.
Refs #1614
* feat(i18n): localize Korean fallback output across analysis pipeline
Add Korean output to the deterministic strings emitted outside the LLM:
per-stock and executor output-language directives, no-API-key / backend
/ parse error fallbacks, hold-watch advice and reasons, market-review
titles and summaries, and history and notification report labels. Fund
flow, trend and confidence values now flow through the shared localize
helpers, and language-keyed advice tables no longer raise KeyError for a
third language. Structural prompt sections route ko to the English
scaffolding. zh/en output is unchanged.
Refs #1614
* feat(web): add Korean report language rendering
Extend the Web ReportLanguage type with ko and add Korean copy to the
report detail surfaces: report text, sentiment labels, market-phase
labels, analysis-context summary, market-review view, diagnostics and
news source. Run-flow chrome that is keyed by UI language falls back to
English for ko. The report-language selector is driven by the backend
config schema, so Korean appears automatically. zh/en rendering is
unchanged.
Refs #1614
* docs(i18n): document Korean report language support
Note that REPORT_LANGUAGE accepts ko in the bilingual guides and the
analyze / market-review request-language parameters, and add a
CHANGELOG entry for Korean report output.
Refs #1614
* fix(i18n): canonicalize Korean values and accept ko in API schemas
Add Korean aliases to the operation-advice, trend, confidence, chip and
bias canonical maps so Korean model output (매수/매도/보유/관망 etc.)
resolves to the correct decision_type and signal level instead of
falling back to hold or a score-band signal. Accept ko in the
analyze, market-review and decision-signal request schemas (and the
static API spec) so the typed client and backend agree and per-request
Korean analysis is not rejected with 422.
Refs #1614
