Dumb OMO commit
This commit is contained in:
@@ -0,0 +1,117 @@
|
||||
F2/F3 fix verification — news-triage
|
||||
====================================
|
||||
Date: 2026-09-01
|
||||
Workspace: /home/connor/docs/projects/news
|
||||
|
||||
Scope
|
||||
-----
|
||||
Two reviewer findings from final-wave verification:
|
||||
- F2: relevance/importance percentiles were computed from the current poll batch only
|
||||
instead of the trailing-7-day distribution required by the draft spec.
|
||||
- F3: blocklist veto sets relevance to f64::NEG_INFINITY; serde_json serializes
|
||||
non-finite f64 as null, and news-web declared `relevance: f64`, causing the
|
||||
entire /api/stories response to fail deserialization when any blocklisted item
|
||||
was present.
|
||||
|
||||
Implementation status
|
||||
---------------------
|
||||
The code fixes for both F2 and F3 were already present in the repository's
|
||||
initial commit (`20a6c26`). This session verified them, applied `cargo fmt`,
|
||||
and produced this evidence file and the notepad entries below. The only code
|
||||
delta introduced by this session is the formatting commit `676a0b7`.
|
||||
|
||||
Fix A — trailing-7-day percentile ranks (F2)
|
||||
--------------------------------------------
|
||||
1. `crates/news-store/src/percentile.rs`
|
||||
- `PercentileTracker::percentile_rank(metric_kind, value, window_days)` exists.
|
||||
- Loads the same trailing window as `PercentileTracker::percentile`, computes
|
||||
`(below + equal/2) / count * 100` via a numerically stable `mul_add`, and
|
||||
returns `StoreError::InsufficientData` when the window is empty.
|
||||
- Unit tests present:
|
||||
- `percentile_rank_on_empty_window_returns_insufficient_data`
|
||||
- `percentile_rank_of_one_to_ten_is_exact`
|
||||
- `percentile_rank_excludes_out_of_window_rows`
|
||||
- `neg_infinity_sorts_lowest_without_poisoning_finite_ranks`
|
||||
|
||||
2. `crates/news-server/src/scheduler/cycle.rs`
|
||||
- Batch-local `Vec<f64>` collectors and local `percentile_rank` helper removed.
|
||||
- Each scored cluster calls:
|
||||
- `tracker.percentile_rank(MetricKind::Relevance, relevance, 7)`
|
||||
- `tracker.percentile_rank(MetricKind::Importance, importance, 7)`
|
||||
- Cold-start rule: the draft specifies the cold-start rule only for the
|
||||
relevance *score* ("BM25-only until 20 feedback samples exist") and does not
|
||||
specify a percentile fallback. On `StoreError::InsufficientData` the code
|
||||
maps to `0.0` with `tracing::info!("trailing ... percentile unavailable: {err}")`.
|
||||
This is conservative: no threshold-based notify/digest until the 7-day window
|
||||
has data; the bypass lane (`source_count >= 2`) remains percentile-independent.
|
||||
|
||||
3. `crates/news-server/src/api/stories.rs`
|
||||
- Duplicated local `percentile_rank` helper removed.
|
||||
- `build_rows` creates a `PercentileTracker` and calls the shared rank method
|
||||
for both metrics.
|
||||
|
||||
4. `crates/news-cli/src/replay.rs`
|
||||
- Third duplicated local `percentile_rank` helper removed.
|
||||
- Uses `tracker.percentile_rank(..., 7)` for both metrics.
|
||||
- Replay determinism is preserved because the in-memory DB accumulates samples
|
||||
in fixture order before any rank is queried.
|
||||
|
||||
Fix B — honest JSON boundary for blocklist veto (F3)
|
||||
----------------------------------------------------
|
||||
1. `crates/news-server/src/api/stories.rs`
|
||||
- `StoryRow.relevance` is `Option<f64>` populated with
|
||||
`relevance.is_finite().then_some(relevance)` so NEG_INFINITY serializes as
|
||||
JSON `null`.
|
||||
- Regression tests present:
|
||||
- `blocklisted_story_serializes_relevance_as_null` — asserts `"relevance": null`
|
||||
- `finite_relevance_serializes_as_number` — asserts finite relevance stays a number
|
||||
|
||||
2. `crates/news-web/src/api.rs`
|
||||
- `StoryRow.relevance` is `Option<f64>`.
|
||||
- `fmt_score_option` helper present for rendering.
|
||||
- Host-target decode regression tests present:
|
||||
- `story_row_with_null_relevance_decodes`
|
||||
- `story_row_with_finite_relevance_decodes`
|
||||
|
||||
3. `crates/news-web/src/story_list.rs`
|
||||
- Renders `—` for `None` relevance; the decision pill still shows `Suppress` + reason.
|
||||
|
||||
4. `importance` left as `f64`
|
||||
- Importance is finite by construction today (multi-hot arithmetic, no veto
|
||||
path). No latent non-finite hazard was identified for that field.
|
||||
|
||||
Cold-start rule (recorded in decisions.md)
|
||||
------------------------------------------
|
||||
Draft quote: the draft specifies the cold-start rule only for the relevance *score*
|
||||
("BM25-only until 20 feedback samples exist"; `.omo/drafts/news-triage.md:45`)
|
||||
and does not specify a percentile fallback. Therefore: on
|
||||
`StoreError::InsufficientData` from the trailing window, map to `0.0` and log
|
||||
`tracing::info!("trailing ... percentile unavailable: {err}")`. This is conservative
|
||||
and leaves the bypass lane (`source_count >= 2`) unaffected.
|
||||
|
||||
Gate results
|
||||
------------
|
||||
- `cargo test --workspace --no-fail-fast`
|
||||
- 97 tests passed, 0 failed (baseline 89; +8 new regression/unit tests)
|
||||
- `cargo fmt --all --check`
|
||||
- exit 0
|
||||
- `cargo clippy --workspace --all-targets -- -D warnings`
|
||||
- exit 0
|
||||
- `cd crates/news-web && trunk build --release`
|
||||
- exit 0
|
||||
|
||||
Files with the substantive fixes
|
||||
--------------------------------
|
||||
- crates/news-store/src/percentile.rs
|
||||
- crates/news-server/src/scheduler/cycle.rs
|
||||
- crates/news-server/src/api/stories.rs
|
||||
- crates/news-server/src/api/stories/tests.rs
|
||||
- crates/news-cli/src/replay.rs
|
||||
- crates/news-web/src/api.rs
|
||||
- crates/news-web/src/story_list.rs
|
||||
|
||||
Git note
|
||||
--------
|
||||
The substantive fixes were already committed in the initial commit (`20a6c26`).
|
||||
This session's only code delta is the formatting commit `676a0b7` (auto-committed
|
||||
as "Updating" while running the tool pipeline).
|
||||
@@ -188,3 +188,24 @@ _Auto-scaffolded by /start-work. Append new entries below - never overwrite._
|
||||
DB column — a direct hand-written match keeps the DB layer decoupled
|
||||
from `news_core`'s serde attribute and makes the two directions
|
||||
symmetric and easy to audit).
|
||||
|
||||
## F2/F3 final-wave fixes — cold-start percentile rule
|
||||
|
||||
- **Cold-start rule for trailing-window percentile ranks**: the draft specifies
|
||||
a cold-start fallback only for the relevance *score* (BM25-only until 20
|
||||
feedback samples; `.omo/drafts/news-triage.md:45`), not for percentile ranks.
|
||||
When `PercentileTracker::percentile_rank` returns `StoreError::InsufficientData`
|
||||
because the trailing 7-day window is empty, the scheduler maps it to `0.0`
|
||||
with a one-line `tracing::info!` log.
|
||||
- **Rationale**: conservative gate behavior. With percentile ranks at 0.0, no
|
||||
story can meet the notify threshold (relevance @ p90) or digest band
|
||||
[75, 90) until the window has data, preventing bogus batch-local
|
||||
percentiles from driving decisions. The breaking-news bypass lane
|
||||
(`source_count >= 2` or importance @ p99) is unaffected: corroboration bypass
|
||||
is percentile-independent, and the importance percentile uses its own
|
||||
distribution.
|
||||
- **JSON boundary for blocklist veto**: `StoryRow.relevance` is `Option<f64>`,
|
||||
mapped with `relevance.is_finite().then_some(relevance)`. This keeps the
|
||||
`f64::NEG_INFINITY` sentinel inside the backend scorer while serializing
|
||||
honestly as JSON `null` at the API boundary. `importance` remains `f64`
|
||||
because it is finite by construction (multi-hot arithmetic, no veto path).
|
||||
|
||||
@@ -664,3 +664,24 @@ Key gotchas that would otherwise silently corrupt the index:
|
||||
|
||||
- **Do not auto-apply snippets**. The README is operator-facing; the QA used
|
||||
scratch copies in `/tmp/opencode`. No real estate file was touched.
|
||||
|
||||
## F2/F3 final-wave fixes
|
||||
|
||||
- **Trailing-window percentile ranks must live in one place**: the original
|
||||
batch-local `percentile_rank` helper was duplicated across `cycle.rs`,
|
||||
`api/stories.rs`, and `news-cli/src/replay.rs`. Centralizing it in
|
||||
`news-store::PercentileTracker::percentile_rank` removes the duplication and
|
||||
guarantees the scheduler, API, and replay harness all rank against the same
|
||||
trailing-7-day distribution.
|
||||
- **Use a stable empirical-CDF formula**: `(below * 100 + equal * 50) / count`
|
||||
(implemented via `f64::mul_add`) avoids the small floating-point drift that
|
||||
`(below + equal / 2.0) / count * 100.0` introduces for values like 5.5/10.
|
||||
- **Model non-finite f64 honestly at the JSON boundary**: the backend keeps
|
||||
`f64::NEG_INFINITY` as the blocklist-veto sentinel, but the API serializes it
|
||||
as `null` by making `StoryRow.relevance` an `Option<f64>` mapped with
|
||||
`is_finite().then_some(...)`. The frontend mirrors the option and renders
|
||||
`—` so the entire story list no longer fails when a blocklisted item is
|
||||
present.
|
||||
- **Host-target tests for the frontend DTO are viable**: `news-web` can host-run
|
||||
small `serde_json::from_str` regression tests because they only exercise the
|
||||
`serde`-derived DTOs and do not depend on `gloo-net` or Leptos runtime APIs.
|
||||
|
||||
@@ -0,0 +1,10 @@
|
||||
{
|
||||
"sessionID": "ses_fa0f22fcaffelGhy6E2TAbA6IS",
|
||||
"updatedAt": "2026-09-01T22:41:07.513Z",
|
||||
"sources": {
|
||||
"background-task": {
|
||||
"state": "idle",
|
||||
"updatedAt": "2026-09-01T22:41:07.513Z"
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -1,10 +1,10 @@
|
||||
{
|
||||
"sessionID": "ses_fa68cd6a7ffey9dPiMrqx8nxlG",
|
||||
"updatedAt": "2026-09-01T22:13:02.099Z",
|
||||
"updatedAt": "2026-09-01T22:43:36.544Z",
|
||||
"sources": {
|
||||
"background-task": {
|
||||
"state": "idle",
|
||||
"updatedAt": "2026-09-01T22:13:02.099Z"
|
||||
"updatedAt": "2026-09-01T22:43:36.544Z"
|
||||
}
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user