Extract `_copy_close_to_null_rows()` with explicit mutation
bounds, searchable close-copy comments, and preserved activity
fields. Type and document `maybe_fill_null_segments()` including
its distinction from packed timestamp-gap repair.
Disable relative epoch tolerance when validating synthetic
stamps: an overfull reservation must not duplicate the right
boundary. Cover modern epochs, typed empty replies, packed gaps,
and unchanged rejected inputs.
Prompt-IO: ai/prompt-io/codex/20260924T165757Z_1da9b760_prompt_io.md
(this patch was generated in some part by `codex` using
`gpt-6` (`openai`))
Complete HMDS no-data replies without silently retrying a day
older. Let `open_history_client()` classify the original request
and leave frame stepping to `start_backfill()`.
Distinguish `request_end_dt` from `oldest_shm_dt` so the first
frame after an empty interval retains its unwritten end bar.
Inline the storage time cutoff and drop the unused two-direction
`diff_history()` helper. Explain both dates and filtering rules.
Add request-boundary tests and real SHM/NativeDB regressions:
all 2,004 supplied bars survive, while the closure stays empty.
Prompt-IO: ai/prompt-io/codex/20260923T164656Z_efa72980_prompt_io.md
(this patch was generated in some part by `codex` using
`gpt-6` (`openai`))
- preserve dirty main-tree work while transferring branch ownership
- hold provider, FQME, timeframe, and symptom fixed while debugging
- capture read-only baselines before disposable live runs
- turn each reproduced failure into deterministic coverage
- qualify interruption and restart behavior after the normal lifecycle
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Lay out a correctness-first path from deterministic regressions
through normalized gap detection, structured backfill completion,
and crash-safe `NativeDB` persistence.
Deats,
- separate checker, storage, provider, and annotation gates
- record confirmed current defects and data invariants
- define commit boundaries and timeout-wrapped verification
Prompt-IO: ai/prompt-io/opencode/20260721T215055Z_c6ec3d41_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Override `NonShittyWrapper._setTimer()` to compute packet silence
in seconds from integer epoch nanoseconds, avoiding upstream's
`datetime` subtraction error after `tcpDataArrived()`. Delegate
pre-traffic datetime and sentinel handling upstream.
Preserve the packet timestamp on expiry and clear the timer before
`timeoutEvent` emission so listeners can safely rearm it. Add three
offline regressions under `tests/providers/` for rescheduling, single
expiry, datetime and sentinel states, disabling, and reentrant
listener rearming. Document the wrapper overrides on the class.
This fixes an explicitly armed packet-idle timer; heartbeat, default
arming, and reconnect behavior are outside this change.
Prompt-IO: ai/prompt-io/opencode/20260924_ib_idle_timer.md
(this patch was generated in some part by `opencode` using
`gpt-6-astra` (`openai`))
Propagate the authoritative `disconnectedEvent` as a
`ConnectionError` so proxy callers and trade consumers fail loudly
instead of waiting on a dead socket. Preserve the last `apiError`
text as the reason without treating numeric farm/order errors as
socket failures.
Deats,
- Disconnect the one-shot callback before sending the exception.
- Raise when `recv_trade_updates()` loses its socket, and relay
disconnect exceptions through both Trio-side consumers.
- Exercise real `eventkit` error/disconnect callbacks and preserve
the supplied commission fill while testing bridge termination.
- Record the aborted integration stage as historical provenance;
its test passes do not validate this post-abort boundary.
Prompt-IO: ai/prompt-io/opencode/20260916_ib_connectivity_integration.md
(this patch was generated in some part by `opencode` using
`gpt-6-astra` (`openai`))
Keep each commission callback's exact `Fill` paired with its
report, so overlapping executions on one contract cannot borrow
each other's fees. Cache positions by account and `conId`, queue
commissioned fills by execution ID until a position arrives, and
accept zero commissions. Reject mismatched execution/report IDs.
Keep `Execution.execId` authoritative when flattening ledger
entries; an empty commission ID must not erase a valid fill. Add
regressions for empty commission IDs, reversed fee arrival, and
positions arriving before or after overlapping executions.
The human live-tested fills and reports the accounting fix works.
Exact-boundary automated checks remain pending; the earlier nine
passes covered the mixed working tree. Included Prompt-IO logs
retain whole-session context, including separate roaming work.
Prompt-IO: ai/prompt-io/opencode/20260916T194052Z_749ca1f0_prompt_io.md
(this patch was generated in some part by `opencode` using
`gpt-6-astra` (`openai`))
Normalize structured and raw IB errors without indexing strings as
message dictionaries.
Deats,
- match each error form before request-id handling;
- parse `[code N]` and `[Errno N]` prefixes without crashing;
- cover raw errors through `deliver_trade_events()`.
Prompt-IO: ai/prompt-io/opencode/20260828T063126Z_1606122b_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
IB returns trading samples, not one row per wall-clock period.
Converting elapsed time to a physical prepend offset left 42,720
zero rows in NVDA's 60s buffer even though both valid margins
shared the same timestamp.
Use the actual earliest provider sample as the NativeDB boundary.
After reverse retrieval, prepend older stored rows directly beside
it.
Deats,
- start reverse insertion at the current first index;
- update `ShmArray._first` after every provider prepend;
- trim provider/NativeDB overlap by its timestamp;
- publish the final storage prepend to chart and FSP consumers;
- derive the source `time` field from `ohlc_key_map`;
- cover the full lifecycle with a real multi-frame `ShmArray`.
Prompt-IO: ai/prompt-io/opencode/20260805T195001Z_f0f3c806_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Make `store ls` show every NativeDB FQME and its available sample
periods by default. Keep `store series` as the detailed view of
each exact sub-series, including its period, size and path. Share
the same optional FQME filter and numerically ordered payload.
Use `store shm` as the canonical volatile-buffer ep and make its
default operation a read-only report. Enter the prior persistence
workflow only through `--write-parquet`, require that flag before
`--reload-parquet-to-shm`. Reject read-only formatting options in
repair mode instead of silently ignoring them.
Also,
- update the qualification docs, debug helper and harness refs;
- cover no-flag listings, command help and SHM mode validation;
- retain exact generation and object-name selection for stale SHM.
Prompt-IO: ai/prompt-io/opencode/20260801T015321Z_74ea2152_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Make bare storage groups and eps render help, so the CLI is
discoverable without memorizing `Typer` option conventions. Add
read-only `series` and `shm` eps for durable `NativeDB` series and
live history buffers, with `Rich`-table or `JSON` output.
Deats,
- preserve exact `(fqme, period)` identities for `NativeDB`
series;
- parse exact `(generation, fqme, kind)` names for SHM buffers;
- reconstruct provider dtypes when attaching to live buffers;
- tolerate buffers disappearing between discovery and attach.
Also,
- report expected and observed cadence separately, plus invalid
rows, ordering defects and ranked gaps;
- add exact `ldshm --shm-name` selection and keep disabled write
and reload paths from reading undefined state or entering `pdb`;
- document the triage layers and cover help, identity parsing,
read-only diagnostics, dtype selection and `ldshm` flow.
Prompt-IO: ai/prompt-io/opencode/20260731T165215Z_0846cbd4_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Keep reverse history and stored SHM layout aligned so sparse IB
frames do not expose permanent zero-time reservations.
Deats,
- omit inclusive query endpoints from SHM but retain full storage
- serialize and bound farm resets, blanks, and cancellations
- run the final null sweep after reverse retrieval completes
- require valid timestamp boundaries for synthetic gap rows
- cover the live QQQ 653-row failure and reset interleavings
Prompt-IO: ai/prompt-io/opencode/20260729T204046Z_ad6c560f_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Do not pass unpublished or corrupt timestamp rows into dedupe,
`NativeDB` replacement writes, or live SHM reload.
Deats,
- infer cadence from finite positive observed steps
- choose the smaller observed step when frequency ties
- skip invalid or degenerate snapshots as a whole
- cover the QQQ zero-slot failure with mutation-failing spies
Prompt-IO: ai/prompt-io/opencode/20260729T185406Z_d359b1cb_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Filter known zero-initialized epoch slots before persisted history
hydrates SHM or merges with a fresh provider frame.
Deats,
- reindex the repaired read view for same-session chart publication
- treat sentinel-only files as empty history for fresh backfill
- reject new first-slot sentinels but preserve modern zero-price bars
- classify raw sentinel evidence separately in `store audit`
Prompt-IO: ai/prompt-io/opencode/20260729T041723Z_ce33deb6_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Serialize aware reverse-fill boundaries as `YYYYMMDD-HH:MM:SS`
before handing them to `ib_async`.
This avoids Gateway error 10314 from its trailing-`UTC` format and
drops the unfinished EST comparison breakpoint from `Client.bars()`.
Prompt-IO: ai/prompt-io/opencode/20260729T034106Z_922a5df8_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Drive known-bad, fresh, restart and append checks from a private
disposable NativeDB root.
Deats,
- bind child config and imports to the marked worktree root
- preserve checksum-matched reports, snapshots and case metadata
- compare timestamp coverage, seam values and persisted structure
- document manual chart runs, gap triage and the expansion matrix
Prompt-IO: ai/prompt-io/opencode/20260728T003247Z_ed85721c_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Inspect Parquet bytes directly without opening storage or normalizing
malformed evidence.
Deats,
- report canonical schema, timestamp, index and finite-value state
- retain exact cadence aggregates with bounded endpoint details
- snapshot bytes before parsing and clean up partial copy failures
- reject unsafe FQMEs, links, collisions and non-regular sources
Prompt-IO: ai/prompt-io/opencode/20260728T005034Z_ed85721c_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Keep incoming provider frames strict while dropping non-positive rows
from persisted history before merge validation.
This lets known-bad MNQ history repair itself on the next valid IB
frame without losing valid old bars or publishing noncanonical indexes.
Prompt-IO: ai/prompt-io/opencode/20260728T232718Z_ed85721c_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Track successful `create=True` calls through Tractor's
current-process factory instead of scanning `/dev/shm` or trusting
the attachment cache.
- cooperate with normal actor unlink and restore token-cache state
- verify POSIX object identity before fallback cleanup
- fail after cleaning exact surviving segments
- reproduce the pre-`current_actor()` allocation leak
Prompt-IO: ai/prompt-io/opencode/20260727T215906Z_689df816_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Require positive, increasing timestamp bounds before querying a SHM
gap or applying synthetic fallback. Defer unbounded edge runs
without entering the debugger or rewriting their rows, and continue
checking later segments.
Cover leading, trailing and entirely null frames with provider,
debugger and notification calls forbidden; assert the complete
frame stays unchanged.
(this patch was generated in some part by `Codex` using
`GPT-6` (`OpenAI`))
IB history frames omit the derived `index` and append `count`.
Normalize provider data to durable names, order, and dtypes before
merge, and reject fractional epoch truncation.
Also,
- map canonical storage fields into provider-specific SHM on reload
- select Polars columns by name instead of position
- exercise first-start and restart through actor, SHM, and Parquet
Prompt-IO: ai/prompt-io/opencode/20260727T212454Z_689df816_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
`datad`'s persistent feed task already owns each series writer.
The actor-local and file locks duplicated this SC contract and left
`.parquet.lock` sidecars that upstream indexed as data.
- preserve validated temp writes, `fsync()`, and atomic replacement
- keep exact-suffix indexing for legacy lock and temp sidecars
- prove full and incremental writes create no lock artifacts
Prompt-IO: ai/prompt-io/opencode/20260727T204711Z_84a6d47b_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
- merge provider deltas without dropping unaffected history
- validate reopened temp parquet before atomic replacement
- serialize writers with actor-local and cancellable file locks
- publish cache only after disk replacement
- persist full startup/backfill frames before SHM publication
- bound sampler notifications and avoid MarketStore duplicates
Prompt-IO: ai/prompt-io/opencode/20260722T024134Z_3a4f6737_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
- keep positive missing-history gaps distinct from ordering faults
- return duplicate and reversed timestamp rows as structured data
- warn before normalization removes the original ordering evidence
Prompt-IO: ai/prompt-io/opencode/20260722T024133Z_3a4f6737_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
- recompute datetime and delta cols after final timestamp ordering
- detect only positive sample-step gaps
- document inclusive null grouping with an updated visual map
- check all small null layouts against an independent test oracle
- honor selected null cols and reject discontinuous frame indexes
- share exact null grouping across NumPy and Polars
Prompt-IO: ai/prompt-io/opencode/20260722T001357Z_16f3cd6f_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Pre the (datad|brokerd)-split this was done by `brokerd`'s
daemon fixture, so ALSO `get_console_log()` the provider
backend's mod subtree (eg. `piker.brokers.ib.*`) in
`_setup_persistent_datad()`; without it all backend records
emitted in the `datad` actor fall through to the stdlib's
bare (non-colorized) `logging.lastResort` handler.
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Follow-up to the lazy order-req qualify: pre-qualify (and
cache) contracts for ALL already-open pps and orders during
`open_trade_dialog()` startup so live submissions NEVER pay
a first-request qualification delay; the warmup runs before
the order handler task starts so early reqs just buffer in
the ems IPC stream. Any brand-new mkt still lazily
qualifies on its first submission.
Deats,
- factor the `Client` lookup-table writes out of
`.symbols.get_mkt_info()` into a new `cache_contract()`
helper which now ALSO keys `_contracts` by `mkt.fqme`
(read by the fill-time pp-update path in
`emit_pp_update()`) alongside `mkt.bs_fqme` (read by
`Client.submit_limit()`); resolves the old "post-split
mktmap lookup" TODO.
- explicitly `cache_contract()` in
`handle_order_requests()` since the lifo-cache may
deliver a hit (body skipped) when another acct/client
already qualified the fqme.
- filter `None` entries (ambiguous contracts) from
`qualifyContractsAsync()` results in
`Client.find_contracts()` before any attr access + raise
a better "use a (more) venue-qualified fqme" error msg.
- relay ALL (non-cancel) errors from the aio method-relay
task to the `trio`-side caller instead of crashing the
whole proxy/dialog; critical post-`datad`-split where eg.
qualification failures are expected to be caught
per-request by order/warmup code.
- handle inline `('event', ...)` api-farm status msgs in
`MethodProxy._run_method()` at info-level instead of the
"UNKNOWN IB MSG" warning.
- only `log.setLevel()` in `open_trade_dialog()`;
attaching a handler via `get_console_log()` double-prints
every record since the daemon fixture already enables the
console handler on the parent subsys logger.
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Post (datad|brokerd)-split the trading actor's
`Client._contracts` cache is never warmed by in-proc
feed setup (that now happens in the `datad.ib` sibling)
so ALL live submissions failed with "no live feed?" at
`Client.submit_limit()`; `brokerd` must be able to
submit orders without any feed registered in its own
subactor.
Deats,
- thread the acct `proxies` table into
`handle_order_requests()` and, on a `_contracts`
cache-miss for the req's fqme, lazily run the same
`get_mkt_info(fqme, proxy=...)` symbology ep the
feed-side uses; it writes the `mkt.bs_fqme` key
`submit_limit()` looks up (and warms `_cons2mkts`
which the position-audit path also needs) on exactly
the same aio `Client` instance.
- guard `submit_limit()` w/ a try/except ->
`BrokerdError` relay so a single bad submission
degrades to an EMS error msg instead of crashing the
dialog (and causing the `TrioTaskExited` teardown
storm seen in testing).
- fix the (non-f-string..) raise msg in
`Client.submit_limit()` and doc the new lazy-qualify
contract; the bug was foretold by the TODO in
`.symbols.get_mkt_info()` B)
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Prompt-IO: ai/prompt-io/claude/20260610T213549Z_f084e899_prompt_io.md
Commit the AI-authored design doc driving this branch's
7-commit `brokerd` -> (`datad` + `brokerd`) split; all
prior commits' `Prompt-IO:` entries reference its
stages so this makes those refs resolvable in-repo for
PR review.
Also,
- add the doc's own prompt-io entry pair (scope: docs)
incl. a 4-item implementation-deviation log vs. the
plan as written.
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Prompt-IO: ai/prompt-io/claude/20260610T173309Z_f15f8178_prompt_io.md
Caps-sec tightening now that `brokerd` is trading-only: NO
`piker.data.*` (feed) mods are RPC-enabled in the (live,
credentialed) trading actor anymore.
Deats,
- drop `_data_mods` for a minimal `_brokerd_service_mods`
(just `piker.brokers._daemon`); dedup-compose with the
backend's set in `spawn_brokerd()`.
- `broker_init()` reads the backend's `_brokerd_mods`
(fallback: `__enable_modules__` for flat backends).
- fail fast in `spawn_brokerd()` via `validate.get_eps()`
when a backend offers NO live order-ctl eps (eg.
`kucoin`, `deribit`) -> tells the caller to use
paper-mode instead of booting a dead actor; analogous
warning in `datad_init()` for datad-ep-less backends.
- offset `ib`'s default `client_id` per daemon-kind in
`load_aio_clients()`: post-split BOTH `datad.ib` and
`brokerd.ib` connect to the same gw/tws endpoint and the
shared default (6116 + linear retry incrs) would collide
and burn the full conn-timeout retry cycle; datad gets
+16, ad-hoc (test/cli) actors +32.
- drop the import-cleanup leftovers (`exceptiongroup`,
`_FeedsBus` type-only import) and the now-resolved
"expose datad" TODO in `.cli`.
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Prompt-IO: ai/prompt-io/claude/20260610T171344Z_eee19de0_prompt_io.md
The topology flip: all data-feed consumers now route to the
new `datad.<broker>` sibling daemon; `brokerd` becomes
trading-only and is ONLY ever booted lazily by `emsd`'s
`open_brokerd_dialog()` (see prior commit). Chart-only and
paper sessions run with zero (live, credentialed) `brokerd`
procs B)
Deats,
- `open_feed()` -> `maybe_spawn_datad()` (NB: imported
relative-direct from `._daemon` to dodge a partial-init
cycle via `piker.service`); flip the `open_feed_bus()`
actor-name assert to `'datad'`; comment sweep.
- slim `_setup_persistent_brokerd()` to a trading-only
fixture: console logging + pinned-open ctx; the feed-bus
alloc moves to `_setup_persistent_datad()` and backend
`open_trade_dialog()` ctxs own their own task trees.
(the `piker ledger` ad-hoc actor enters this same slimmed
fixture - exactly what it needs.)
- repoint data-flavoured spawn sites to `maybe_spawn_datad`:
`.ui._app` symbol-search (+ rename
`install_brokerd_search` -> `install_datad_search`),
`.brokers.core.symbol_search()`, `.brokers.cli`
`brokercheck`/`record`, legacy kivy `.ui.cli` +
`option_chain`'s `wait_for_actor()`.
- invert `tests.test_services` expectations: feed/EMS-paper
flows must spawn `datad.kraken` and `paperboi.kraken`
with an explicit negative assert that NO `brokerd.kraken`
service task exists.
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Prompt-IO: ai/prompt-io/claude/20260610T171259Z_59d5d9a6_prompt_io.md
Drop the ONE coupling that forces feed + trading eps into
the same actor: `Router.open_trade_relays()` pulling its
trades-dialog portal from `feed.portals[brokermod]`.
Instead `open_brokerd_dialog()` now (maybe) spawns/finds
`brokerd.<broker>` itself via `maybe_spawn_brokerd()` and
ONLY when a live trades-ep will actually be opened; the
paper-mode short-circuit never touches it, so post
feed-cutover paper sessions will run with zero `brokerd`
procs.
Pre-cutover this is a pure refactor: the registry lookup
just finds the same feed-spawned daemon.
Deats,
- new `open_brokerd_dialog()` sig: portal acquisition moves
inside via an `acquire_live_portal()` helper; keep an
explicit `portal: Portal|None` override for the
`piker ledger` cli which boots its own ad-hoc actor.
- `Router.maybe_open_brokerd_dialog()` drops its `portal`
param; `open_trade_relays()` drops the `feed.portals`
lookup entirely.
- `.accounting.cli`: pass `portal=` by keyword.
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Prompt-IO: ai/prompt-io/claude/20260610T171226Z_64181219_prompt_io.md
First half of the `brokerd` split: a new per-provider
data-feed-only daemon-actor `datad.<broker>` to (soon) host
all `validate._eps['datad']` eps (live quotes, history
loading, symbology search) leaving `brokerd` for live order
ctl only. Purely additive; nothing routes through it yet.
Deats,
- new `piker.data._daemon` mod mirroring the
`.brokers._daemon` conventions (and the `samplerd`
sub-daemon precedent):
- `_setup_persistent_datad()` lifetime fixture owning the
actor-global `_FeedsBus` alloc.
- `datad_init()` building `enable_modules` from the
backend's `_datad_mods` (falling back to
`__enable_modules__` for not-yet-split backends) and
copying `_spawn_kwargs` (critical for `ib`'s
`infect_asyncio`).
- `spawn_datad()`/`maybe_spawn_datad()` wrapping
`Services` + `maybe_spawn_daemon()`.
- add `piker.data._daemon` to `_root_modules` so `pikerd`
can run `spawn_datad()` requests.
- re-export the spawn eps from `piker.service`.
- add `test_datad_spawn` verifying actor boot + service
registration via `ensure_service('datad.kraken')`.
Note the `Services`-based impl style deliberately mirrors
`spawn_brokerd()` so the eventual `tractor.hilevel`
`ServiceMngr` port (see the `service_mng_to_tractor`
branch's d8c21d44 prep work) lands symmetrically on both.
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Prompt-IO: ai/prompt-io/claude/20260610T171142Z_119d2c04_prompt_io.md
Prep for the `brokerd` -> (`datad` + `brokerd`) actor split
by having each (split-style) backend declare which of its
submods host which daemon-kind's eps, exactly per the
`piker.data.validate._eps` groupings; `ib` already had
`_brokerd_mods`/`_datad_mods` so extend the convention to
`kraken`, `binance` and `deribit` (and add `'api'` to ib's
datad set since both kinds need the `Client` layer).
`__enable_modules__` stays as the (deduped) union so this
is a ZERO behavior change; flat backends (`kucoin` etc.)
just don't declare the split yet.
Also,
- add `validate.get_eps()` returning a backend's defined
eps per daemon-kind for spawn-time introspection.
- import `NoBsWs`/`open_autorecon_ws` from
`piker.data._web_bs` directly in `.kraken.broker` (they
were only re-exported via `.kraken.feed`) so the trading
mod doesn't depend on the feed mod for ws primitives.
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Prompt-IO: ai/prompt-io/claude/20260610T171105Z_bc6e18d7_prompt_io.md
The `proxy_msgs()` deadline covers both `ws.get_message()` and
`snd.send()`. Repeated socket-ready reports with no first relay
could therefore mean either a silent source or a blocked local
consumer; the old generic reset log could not distinguish them.
Track the currently awaited operation and report its phase, URL,
deadline, accumulated timeout count and reset threshold on each
timeout. Snapshot `NoBsWs._timeout` once per iteration so the log
reflects the deadline that actually elapsed. Keep the existing
shared deadline, counter behavior and reconnect policy intact.
Add two real loopback regressions: a silent websocket server and
one payload blocked on an unconsumed zero-capacity Trio relay.
Require the matching timeout phase and URL before the reset budget
is exhausted. The 15 selected Binance/websocket cases pass.
The human observes this after invalid `.usdm.` startup followed
by corrected `.usdtm.` input. Metadata accepts the typo but later
history lookup expects canonical keys. Record that investigation
as context; this diagnostic patch does not fix FQME validation,
failed-startup cleanup or prove the cause of corrected-feed stalls.
Prompt-IO: ai/prompt-io/opencode/20261028T180856Z_032580bf_timeout_phase.md
(this patch was generated in some part by `opencode` using
`gpt-6.1-sol` (`openai`))
An INFO pikerd still hid websocket connection-success messages.
The shared ._util.log emitter is piker.data, but open_feed_bus()
configured piker.data.feed. Errors reached bare stderr while
INFO diagnostics remained disabled on the actual emitter.
Configure the shared logger by its exact log.name and prefer
the actor's loglevel over a quieter per-client request for these
actor-wide transport diagnostics. Preserve the selected client
level passed to provider feed startup.
Also,
- retain the existing handshake-success report and add URL-tagged
readiness after fixture setup and first-message relay once per
connection, including replacement sockets.
- include the URL in receive-timeout reset errors so Binance's
independent bookTicker and aggTrade routes are distinguishable.
- prove real console INFO visibility through production feed
startup, stopping before provider lookup, then restore logging
state; the original code fails with the emitter at WARNING.
- extend the loopback recovery test to verify both readiness
reports and the first relay from the replacement connection.
First-message relay is raw websocket traffic, not proof that
parsed market data or chart updates have recovered. The 18
logging/Binance/websocket cases pass; live reporting qualification
remains separate.
This is a local follow-up, not a downstream cherry-pick. The
54da2973 websocket cleanup and 5adf77a9 namespace fix do not
configure this shared emitter. Relevant local histories and
dirty IB-history/TIPC deltas contain no equivalent wiring fix.
Prompt-IO: ai/prompt-io/opencode/20261028T165638Z_b95e7c79_reconnect_logging.md
(this patch was generated in some part by `opencode` using
`gpt-6.1-sol` (`openai`))
A Binance Ping racing local socket closure makes trio-websocket's
reader send a Pong while wsproto is LOCAL_CLOSING. That reader
nursery error escapes proxy_msgs(), then tears down brokerd and
pikerd instead of restoring quote delivery.
Classify NoBsWs.recon_errors around the entire websocket context,
with collapse_eg() outside it so a singleton reader exit group
reaches the reconnect handler. Invalidate the stale connected
event and socket reference before retrying. Keep retry sleep
unshielded so parent cancellation stops recovery promptly.
Deats,
- Add real loopback websocket coverage: TCP, handshake, Ping,
reader dispatch and wsproto's actual Pong validation.
- Control only the first client Ping's protocol-state ordering,
withholding its Close frame to prevent a peer response racing
the deterministic LOCAL_CLOSING reproducer.
- Require fresh data on a replacement connection, plus exactly
one connection when cancelling at the reader or retry boundary.
- Use RuntimeError for the unexpected-failure control:
LocalProtocolError is now a recoverable connection failure.
- Preserve the original unexpected exception with bare raise,
and assert its identity in the test. Log recoverable errors
with the websocket URL and the underlying traceback.
This is a local follow-up, not another downstream cherry-pick.
The earlier wrapper suppressed escaped communication failures;
97eb8347 deliberately re-raised them. Retain that outer error
visibility rather than restoring suppression without recovery.
The isolated baseline reproduces the live Pong failure; all 13
selected Binance/websocket cases pass with this fix. Human live
qualification remains separate from these loopback regressions.
Follow-up-to: 97eb834799
Original-equivalent: d50ab6124c
Related-history: 3d83b61f3f
Prompt-IO: ai/prompt-io/opencode/20261028T160846Z_7242dcb9_websocket_close.md
(this patch was generated in some part by `opencode` using
`gpt-6.1-sol` (`openai`))
- reject non-single-symbol feeds with `FeedInitializationError`
- route book and trade data across the correct USD-M endpoints
- accept the first matching L1 or trade quote as feed readiness
- preserve transport and startup-timeout failure context
- cover provider msg schemas and both readiness classes
Prompt-IO: ai/prompt-io/opencode/20260829T175542Z_e09e306a_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
- restart handled handshake failures before touching cancel state
- propagate `trio.Cancelled` without shielded retry throttling
- bound initial connection waits and expose transport failures
- cover real reconnect task paths in `test_websocket.py`
Prompt-IO: ai/prompt-io/opencode/20260901T024235Z_e09e306a_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Binance retired its legacy futures websocket and split quote
traffic across `/public` and `/market`. Route `bookTicker` and
`aggTrade` through dedicated URLs, merge both sockets, and accept
the new `st` stream-type field.
Bound connection and first-quote startup so failed or silent
streams raise `DataUnavailable` instead of hanging `datad` and
chart consumers.
Also,
- derive feed mode from resolved `FutesPair` metadata
- keep direct URL subscriptions stable across reconnects
- cover USD-M routes, payloads, and startup timeouts
Prompt-IO: ai/prompt-io/opencode/20260826T222600Z_5f631e19_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
Advance the `piker_pin` lock from `b64c89f0` to the published
`8fe574c1`, retaining the evaluated Trio 0.34.0 release. Require
Trio >=0.34.0 for the updated cancellation APIs. Restore the Git
source used before local editable evaluation; no environment sync
is included.
Record the human's overlapping charts, Gateway data reset and
close/reopen qualification. Keep the full Gateway restart string
error and Tractor teardown callback assertion explicit as separate
recovery limits. Offline lock validation resolves 126 packages;
this is not a fresh-install qualification claim.
Prompt-IO: ai/prompt-io/20261106T201746Z_495c0915_612dfc_prompt_io.md
(this patch was generated in some part by `codex` (`openai`))
Partially backport only the `get_hist()` exception catch from
0846cbd4a4, authored by goodboy
<goodboy_foss@protonmail.com>. Keep the existing gap fallback.
A timed-out IB query raises `NoData` from null-segment repair.
Letting it escape cancels sibling history and live quote tasks
while `pikerd` can remain alive. Catch `NoData` and
`DataUnavailable` locally and stop querying this repair pass.
Add a regression using real null detection and a waiting sibling
in the same nursery. Both provider failures must return without
canceling the sibling. The two cases failed before this catch;
the focused history files passed 17 tests with it.
Partial-backport-of: 0846cbd4a4
Prompt-IO: ai/prompt-io/20261106T185628Z_495c0915_215e9b_prompt_io.md
(this patch was generated in some part by `codex` (`openai`))
Use `msg.setdefault('brokerd_msg', msg)` instead of blind assignment and
log a warning when the field was already populated.
Specifically, this avoids a self-reference field recursion which causes
crashes when using `tractor.devx.pformat.ppfmt()`..
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
(cherry picked from commit 8cefc1bdf8)
- let structured nursery joins own child-task completion
- stop empty provider frames before timestamp endpoint indexing
- cover provider exhaustion and empty frames with bounded tests
Prompt-IO: ai/prompt-io/opencode/20260722T001358Z_16f3cd6f_prompt_io.md
(this patch was generated in some part by `opencode` using `gpt-5.6-sol` (`openai`))
(cherry picked from commit 931f22d235)
Record exact keep/backout/import provenance for the #94 candidate,
including the quote-cache lineage correction and the distinction
between file-equivalent adapter deltas and partial source commits.
Deats,
- Preserve the original raw prompt record and add the human's
later scope decisions, corrections and live-test contributions
to its structured partner.
- Record 38 prior unique offline passes and the known-red bridge
callback-cleanliness case, without claiming clean teardown.
- Record live QQQ backfill, reattachment and overlapping chart
close/reopen results. Startup needed debugger continuation;
the local breakpoint cleanup was applied after that run.
- Retain the cached-client daemon shutdown failure as a human-
accepted follow-up outside #94, with its exact propagation path.
- Require disposable downstream rebase rehearsals before landing,
preserving omitted SHM/storage fixes and local regressions.
The commit package records exact per-boundary project checks.
Its structural replay/preflight does not substitute for those
checks, the prior live observations or rebase qualification.
Prompt-IO: ai/prompt-io/opencode/20261022T212248Z_dc618229_ib_recovery_prompt_io.md
(this patch was generated in some part by `opencode` using
`gpt-6-astra` (`openai`))
Avoid interactive startup pauses when `get_null_segs()` sees one
zero-time row. Remove the explicit breakpoint and its now-unused
`RuntimeError` handler, preserving the legacy `None` result.
LOCAL UX CLEANUP -- NOT the downstream null-grouping fix.
The related source below removes this branch as part of a much
larger grouping/deduplication rewrite. None of that algorithm,
its index validation or its oracle coverage is backported here.
Downstream replay,
Prefer the complete downstream implementation. Do not preserve
this legacy single-null early return against the later rewrite.
The small inline comment records that temporary discrepancy.
Related-downstream: 976db6c33c
Prompt-IO: ai/prompt-io/opencode/20261022T212248Z_dc618229_ib_recovery_prompt_io.md
(this patch was generated in some part by `opencode` using
`gpt-6-astra` (`openai`))
Import the source-equivalent EMS code and its original wrapper
regression. `clear_dark_triggers()` owns a non-strict child from
`quote_stream.subscribe()` for the lifetime of its clearing loop,
instead of changing a shared receiver's private lag policy.
LOCAL ADDITION,
Add a second regression using the production clearing loop and a
real broadcaster. Deliver a tick to EMS and a sibling, cancel EMS,
and prove the sibling remains usable with strict lag policy.
No order trigger or live broker is involved in this test.
Downstream replay,
The EMS code has the source patch's stable patch ID. The test file
also contains the local real-broadcaster case, so the complete
commit is not patch-identical. Preserve that extra coverage when
reconciling an already-applied EMS delta.
Backport-of: 12b0d66e85
Source-author: goodboy <goodboy_foss@protonmail.com>
Prompt-IO: ai/prompt-io/opencode/20261022T212248Z_dc618229_ib_recovery_prompt_io.md
(this patch was generated in some part by `opencode` using
`gpt-6-astra` (`openai`))