Skip to content

Promote edge to stable: eMudhra distrust, every eligible comment routable, simultaneous joins, the vote path, unknown ids answer 404, seed CSV fixes (migrations 000043, 000044) - #3268

Merged
colinmegill merged 31 commits into
stablefrom
promote-2026-10-24
Oct 24, 2026

Conversation

@colinmegill

Copy link
Copy Markdown
Member

Promotes edge to stable (merge commit; the tree equals edge at 1c6e874). 30 commits.

Database: two migrations apply during the deploy.

  • 000043: three indexes for the derived conversation activity reads, built concurrently (writes are not blocked).
  • 000044: the participant-number lock held to commit. One-time: DROP TRIGGER takes ACCESS EXCLUSIVE on participants, bounded by the 5 s lock timeout; joins sent meanwhile wait and are stored.

Security

Participants

Admins

No user-visible change

🤖 Generated with Claude Code

colinmegill and others added 30 commits October 19, 2026 13:29
…switch (#3229)

Rust read route only, off by default (POLIS_API_DELPHI_LOGS_READS=1 plus the delphiLogs arm of RUST_API_ROUTES); no Node handler, schema or write change. GET /api/v3/delphi/logs?job_id=… served by polis-api with the run-record route's moderator rule (never an identity inferred from an unverified token), byte-faithful to Node against the Postgres-mode recording baseline, including stored log text carrying lone UTF-16 surrogate escapes, which Node accepts. A 502 hands the request to Node. Reviewed independently: publish as-is with the correction (folded in).


Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…ind its own switch (#3230)

Rust read route only, off by default (its own switch plus a RUST_API_ROUTES arm); stacked on the job-log route PR (merge that first; retargeted to edge after). GET /api/v3/delphi/visualizations with the job filter, order, archived and live-work behaviours byte-faithful to Node against the Postgres-mode baseline. When the liveness query fails, the route hands the request back to Node instead of answering with live-work true (the reviewer's required edit, with its test), so a Rust statement timeout can never report a run as live where Node says otherwise.


Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…own switch (#3231)

Rust read route only, off by default (its own switch plus a RUST_API_ROUTES arm); stacked on the visualizations route PR (merge order: logs, visualizations, topics; retargeted to edge as each merges). The public topics-by-layer read with moderation state, byte-faithful to Node against the Postgres-mode baseline; it passes stored topic names through unchanged. No write, schema or Node change. Reviewed independently: publish as-is after the rebase.


Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…behind its own switch (#3232)

Rust-only read of GET /api/v3/delphi/reports, off by default (its own switch plus a RUST_API_ROUTES arm); byte-faithful to Node against the Postgres-mode baseline (stored timestamps are copied strings on both sides); a query string the legacy parser would decode differently is handed to Node, and an unmatched section filter falls back to the section Node names; a 502 hands the request to Node.


Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…tive statement (#3233)

CLI only (34 files); no server change. The participant command family is renamed: statement list, view and next become comment list, view and next; report statements list becomes report comments list; polis comment (submitting) and the collective statements commands are unchanged. Help text, README, tests and the differential census follow, including every user-visible error and help string that still said statement for a participant's comment. Reviewed independently: publish with edits (applied).


Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…greement or divisiveness (#3234)

CLI only (seven files), stacked on the naming change (merge that first; retargeted to edge after). polis comment stats --conversation ID [--sort agree|disagree|pass|divisive|consensus|important|votes] [--by-group] [--json]: per comment the vote count, agree, disagree and pass percentages over votes cast, the important count when the site serves it, a divisiveness measure stated in the help (1 minus the absolute difference between agree and disagree over their sum), per-group columns on request, and the text. Counts print as integers; a conversation without opinion groups yet is said so, with a test. Source: the math read's per-group tallies and the comment list; input validated before any credential is read. It answers: which comments divided the room the most? Reviewed independently: publish with edits (applied).


Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…for each (#3235)

CLI only (seven files), stacked on comment stats (merge order: naming, stats, questions). polis questions prints, with no sign-in, the ten patterns an organizer looks for in a conversation (what everyone agreed on; what divided the room; the groups and what defines each; what each group alone believes; what is important to people; what is new; what lacks votes; how it moved over time; the collective statements per topic; one participant against the room) with the exact command for each, plus a root help section and a JSON guide. All examples use the renamed commands. Reviewed independently: publish as-is.


Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…te, site add/list/use/remove (#3236)

CLI only. pol.is stays the default with no configuration and every existing command is unchanged without one. A named site (a local stack, a self-hosted or regional instance) is added once with polis site add NAME URL, selected per command with --site NAME or made the saved default with polis site use NAME; polis site use default returns to pol.is. Sign-ins are kept per site, the config write is atomic and locked, and the active site is named in output whenever it is not pol.is.


Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
)

CLI only. Every command has --json; the 77 tabular reads gain --csv (RFC 4180 quoting and line endings, a header row, byte-stable for the same response). No output gains an identity field a route does not return; participant-visible reads carry no xid; owner and moderator outputs keep xid where the route gives it. The README shows an organizer joining their own member list to the votes export on xid.


Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
* nginx: switch a Rust route on or off without an outage

Observable changes: none for any request by default. Every Rust route stays
off unless RUST_API_ROUTES names it, as before.

What changes when an operator flips RUST_API_ROUTES:
- nginx regenerates its Rust map and reloads in place; the nginx-proxy
  container is never recreated. Before this, a flip recreated nginx-proxy and
  polis-api together (about 11 s), and every route through nginx, votes and
  comments included, failed for that long.
- polis-api starts or stops without touching nginx; nginx tolerates the
  upstream being absent while every route is off (no failed startup, no 502
  on Node-served routes).
- A container start always obeys its RUST_API_ROUTES environment: unset or
  off overrides any previously persisted selection.
- Rendering holds a kernel flock that dies with its process, so a hard stop
  can never leave a stale lock that blocks nginx startup.

Proof on a local stack: four 50 ms probe streams on participationInit,
nextComment, POST votes and POST comments plus pca2 and testConnection, with
real votes and comments arriving, across pca2 on and off: 3152/3152 requests
succeeded; nginx, Node and dependency container ids unchanged; only polis-api
had lifecycle events. Hard-stop-with-stale-lock and env-off-after-persisted-on
restarts both come back serving Node. Local checks: nginx-routing.sh 62/62;
file-server/nginx/test-rust-reload.sh 10/10 (2238 stub probes, 0 failures).

Files: docker-compose.yml, file-server/nginx.Dockerfile, the nginx entrypoint
and two render scripts, test-rust-reload.sh, file-server/nginx/README.md,
scripts/rust-api-switch.sh. No Node change. No schema change. How a
production deploy would carry the reload path is documented separately as a
proposal and is not part of this change.

Review round 1 by the second reviewer; corrections applied and re-proven.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* differential: the CLI check builds the production nginx image and switches routes by reload

The differential harness mounted the split entrypoint into a stock nginx
image without the route helper it now calls, so the proxy exited before
nginx started, in the Node phase too. The harness now builds the image
from file-server/nginx.Dockerfile, gives the proxy its own routing volume,
and enters each phase through the reload helper instead of recreating the
container; the access log is truncated after old workers drain so each
phase's counts stay independent. No product code, expectation or bound
changed.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
#3239)

* votes: a typed contract for POST /api/v3/votes with 140 recorded cases

The contract in polis-contracts names the request, every response Node
produces on this pin and the vote convention it speaks through; the
characterization harness records and replays 140 cases of the real route
against unchanged Node with zero differences. Three opt-in probes stay red
on purpose: they document the deferred bookkeeping writes failing after a
durable vote, which a separate change fixes. No served behaviour changes.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* votes: drop two unused bindings in the recorder

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…h that no longer stores vote_count (#3241)

The vote-path recorder derives a participant's count from the votes table
when settling and in its recorded participants query; the routing replay's
substitute pg gains a notification passthrough and extracts updateVoteCount
only when the server still declares it. Harness only; goldens unchanged.

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
… derived at read (#3240)

Node's deferred bookkeeping after every vote (a participant recount, the
conversation's modified time, the notification state) ran on the two
participation write pools and, behind a held row lock, starved votes,
comments and nextComment until they timed out. The vote and comment paths
no longer write vote_count or modified; readers derive both from committed
history. Notification state moves to its own bounded pool and fails soft.
A participation-under-load integration test holds the realistic lock
(NO KEY UPDATE on the conversation and participant rows) and asserts every
request answers within 5 s. No schema change.

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…ion makes no request (#3242)

The updater resolves the newest release at the configured release location,
verifies the asset's SHA-256, replaces the binary atomically and keeps a
backup; every failure has its own exit code and leaves the installation
untouched, and a sync failure after replacement reports an unknown outcome.
--version prints the version and commit, plain or JSON, offline.

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…, as deployed (#3119)

The web security group no longer opens SSH (22), HTTP (80) or HTTPS (443)
to the internet in code. Production has run this way since 2026-10-11:
the web instances take port 80 from the load balancer's security group
only, and nothing else reaches them from outside the VPC. Until now the
code still added the three 0.0.0.0/0 rules, so the next deploy from this
branch would have reopened them.

Also brought over from the deployed stack, unchanged:

- SSH stays off unless asked for, and turning it on now requires naming
  a range (CDK_SSH_ALLOWED_IP_RANGE); there is no 0.0.0.0/0 default.
- cdk/originAccess.ts: two controls that let only the CDN reach the load
  balancer, both off by default. `-c restrictAlbIngress=true` replaces
  the load balancer's 80/443-from-anywhere rules with a prefix list or a
  list of CIDRs (config/cloudflare-ips-v4.json holds the CDN's published
  IPv4 ranges); `-c requireOriginHeader=true` makes both listeners answer
  403 unless the request carries the CDN's origin header, read from
  Secrets Manager by name. dns.ts takes the listener settings from it.
- test/originAccess.test.ts covers the web group, SSH, and both controls
  on and off.

With every flag off, the load balancer's rules and listeners are exactly
as before. The whole-stack snapshot changes only by the three web rules.

The other deployed edits were already here and are untouched: the Delphi
small box on r7i.2xlarge, the operational alarms, the ops dashboards
policy, the large memory class (now the worker class table), the backup
function packaging and the build box.

Not changed here, and not yet in the deployed stack: the inactive
coordinator service (always synthesized) and the reworded description
of the CI worker's security group (synthesized with enableCiEc2; a
description change replaces the group).

Checked with `cdk synth -c enableCiEc2=true --no-lookups` of this branch
and of the deployed tree: apart from those two and the CDK metadata
resource that lists the constructs used, the templates are identical,
including the backup function's asset hash. cdk tests: 10
suites, 182 passed; tsc clean.


Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
… contract; concurrent index builds get a restricted header (#3244)

The replays, the coordinator tools, several test fixtures and db-reset sent
each migration file to PostgreSQL as one query, an implicit transaction that
refuses anything which must run outside one. A shared Node helper and a
shared Python helper now hand header-marked files to polis-migrate, which
gains a restricted concurrent-indexes header: one autocommit per plain
btree index under the deploy's advisory lock, verified valid before history
is written, everything else refused with a test. No migration is added.

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…, built concurrently (#3243)

Adds votes(zid, created), comments(zid, created) and comments(zid, modified)
under the concurrent-indexes header so the activity time derived at read is
an index probe instead of a scan of the conversation's votes. Selected for
the next deploy; declared additive; listed in the deploy report and the
upgrade notes.

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…3245)

The first visit to an embedded page creates its conversation after looking
up which account owns the page's site id. When no account does, the route
reported a server failure; it now answers 404 with the code the handler
already used for that case. Nothing is created in either case and every
other path keeps its status.

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…me, staged off (#3247)

* polis-api: typed read routes for GET /api/v3/votes and /api/v3/votes/me, staged off

Two contracts and two routes with their own bounded read pool; they serve
the demonstrated empty and single-row results byte-exactly against Node's
recordings and hand everything else back to Node. Off by default behind
the nginx switch; nothing is served until switched on.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* differential: the routing census carries the two vote read routes; contracts crate formatted

The CLI differential's route inventory now lists the votes and votes/me
selector with scenarios and exact serving counts, and the contracts
crate's lib.rs is rustfmt-clean.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
… from stable in every build (#3246)

* version: one Polis version, embedded with the commit and the distance from stable in every build

The root VERSION file holds the Polis version; the server, the three
browser clients and the CLI embed it with the full build commit and, off a
tagged stable build, the distance from the last stable, rendering
2.x.y+dev.N.<commit>. polis --version, a server version route and the
admin's About page report the same string; a source run without the
generated build file falls back to the VERSION file. No tag, release or
workflow change; the updater keeps its contract.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* version: the bootstrap number is 2.41.0 under the effective-selection rule

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…tag and the promoted range (#3248)

polis-version applies the one rule (a promotion that ships a migration is
a minor bump, otherwise a patch; the major never moves; hotfixes follow
the same rule; a redeploy gets no new number), prints the plan as text or
JSON, dry-run by default, and extends the seeded history of promotions
since 2.0.0 to the given stable commit. Nothing calls it yet.

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
The updater resolves the newest release tagged vMAJOR.MINOR.PATCH, the
Polis release on the stable merge commit, and takes the CLI's eight
assets from it; asset names, checksum verification, atomic replacement
and every failure case are unchanged. A cli-v tag is ignored with a
clear message, and a development build never replaces itself unasked.

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
GET /api/v3/participation reported a server failure for a well-formed
conversation id that matches no conversation; it now answers 404 with
the code the resolver already used for that case, through an optional
error responder on the shared parameter helper that only this route
calls. Malformed ids keep their 400 codes and auth failures their 401;
the success path is byte-identical.

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
…image, with tests (#3253)

* tls: distrust the eMudhra intermediates in every runtime, bundle and image, with tests

Eight eMudhra intermediate CA keys were recoverable; browsers block them
but file-based trust stores do not until updated. Node images move to
22.23.4 / 24.22.1 (compiled-in SPKI blocklist), certifi to 2026.10.19,
webpki-roots to 1.0.10 and rustls-webpki to 0.103.16, base image CA
packages updated, and ci/tls asserts offline that each trust store
rejects the sixteen affected certificate objects, direct and cross-signed.
No served behaviour changes.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* tls: the review's corrections: hash-checked Lambda payload, all sixteen objects with native verifiers, final runtime images and the probe base's local distrust

The backup Lambda's certifi wheel is hash-checked and its real package is
tested. ci/tls now checks all sixteen affected certificate objects by
fingerprint and SPKI, including the code-signing intermediates, and
proves the result with each runtime's native verifier (Node's compiled
blocklist, Python's ssl/certifi/botocore paths, Java PKIX). The image
gate covers the fifteen final runtime images. The probe base's Amazon
Linux 2023 ca-certificates package has no fix yet, so it distrusts the
affected roots locally by exact fingerprint through the p11-kit
blocklist, to be removed once the vendor package ships. The legacy
commentgraph image, which is not built or deployed, stays selectable but
outside the required gate.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
… the form (#3258)

The bulk seed upload's rejection handler was never attached to the
request, so a failed POST /api/v3/comments-bulk left the form on
"Saving..." with nothing shown. It now dispatches the existing error
action: the form leaves its loading state and shows the server's message,
and a retry works. A component test covers success and failure.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…st 999 (#3261)

Ordinary routing ordered a participant's eligible, unvoted comments by
creation and kept the first 999 before the weighted draw, a default limit
from 2013 meant for the comments list. In a conversation with more than
999 such comments, later ones could never be offered until the
participant had voted on older ones. Node and the Rust route now draw the
same weighted lottery over the ids of every eligible comment and then read
only the chosen row, so a vote does not pull every candidate's text. The
public comments list keeps its existing default. Node and Rust stay
byte-identical for the same database state and draw.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…d cases (#3256)

* comments: a typed contract for POST /api/v3/comments with 153 recorded cases

The contract in polis-contracts names the request, every response Node
produces for POST /api/v3/comments on this pin, the durable-before-response
boundaries and the vote convention the author's explicit vote speaks
through. The characterization harness records and replays 153 serial cases
and four concurrent ones against unchanged Node with zero differences:
anonymous, standard, OIDC first use, XID, Treevite and owner/moderator
identities; first and repeat membership; absent, null, empty, long and
duplicate text; closed and strictly moderated conversations; the comment
number held to commit under concurrency; and the moderation email to the
owner and site moderators, captured by a local SES endpoint with an
explicit expected-attempt policy. A separate declared observation profile
waits out the language-detection budget for the timeout cases only. A
golden guard keeps product and recordings from changing together, and a
plan for qualifying a second writer is written down. No served behaviour
changes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* comments: recorder artifact paths are safe for upload; an unused argument is marked

A recorded case whose id contains a quote wrote a directory name the
artifact upload refuses; case directories are now percent-escaped
reversibly. The upload action is pinned by commit like the other
workflows, and the control's unused argument is named for the linter.
No recording changes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…3257)

* polis-api: typed read routes for /iip and /iim, each behind its own switch

Recorded byte for byte against unchanged Node and replayed; the public
conversation id is resolved by one indexed lookup per request. Off by
default: nothing is served until its switch is named.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* polis-api: a typed read route for GET /api/v3/participants, behind its own switch

Recorded byte for byte against unchanged Node and replayed; the public
conversation id is resolved by one indexed lookup per request. Off by
default: nothing is served until its switch is named.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* polis-api: a typed read route for GET /api/v3/ptptois, behind its own switch

Recorded byte for byte against unchanged Node and replayed; the public
conversation id is resolved by one indexed lookup per request. Off by
default: nothing is served until its switch is named.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* polis-api: a typed read route for GET /api/v3/treevite/myInvites/csv, behind its own switch

Recorded byte for byte against unchanged Node and replayed; the public
conversation id is resolved by one indexed lookup per request. Off by
default: nothing is served until its switch is named.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* polis-api: typed read routes for GET /api/v3/bid and /api/v3/locations, each behind its own switch

Recorded byte for byte against unchanged Node and replayed; the public
conversation id is resolved by one indexed lookup per request. Off by
default: nothing is served until its switch is named.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* polis-api: typed routes for the conversation pages (root, explore, share, summary, ot), each behind its own switch

Recorded byte for byte against unchanged Node and replayed; the public
conversation id is resolved by one indexed lookup per request. Off by
default: nothing is served until its switch is named.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* differential: scenarios and serving assertions for the twelve new read routes

Each new selector is inventoried with its serving counts with the switch
off and on: the five API routes through the real CLI, the seven HTML
routes through GET, HEAD and query hand-back scenarios comparing complete
bodies, status and type. Test-only.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…essage (#3262)

* comments: a seed CSV without its header row is refused with a clear message

A seed CSV's first row must be the header comment_text, optionally with
original_id. A file without it answered 200 after treating its first
comment as column names and skipping the rest, and some wrong headers
stored comments anyway. The upload is now checked before any participant
or comment is created and refused with 400 polis_err_csv_header, "CSV
header must be comment_text, optionally with original_id.", which the
admin upload form shows. Well-formed one- and two-column uploads are
unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* comments: a seed CSV saved with a byte-order mark is read normally

Excel's 'CSV UTF-8' export starts with a byte-order mark, which made the
first header read as '\ufeffcomment_text' and the upload be refused.
The parser now drops the mark, so those files upload as before. Four
cases cover both header shapes and quoting with a leading mark.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…cipants joining at once each get a number on the first try (migration 000044) (#3266)

The pid_auto trigger numbers a conversation's participants MAX(pid) + 1
under a per-conversation advisory lock, but took it as a session lock that
pid_auto_unlock released when the INSERT ended, before COMMIT. A second
join could read the same MAX(pid) and fail with 23505 on (zid, pid); one
retry hid most of these, and two collisions in a row reached the caller
as HTTP 500, for example a Treevite invite accept that left the invitation
unused. Migration 000044 takes pg_advisory_xact_lock on the same key and
drops pid_auto_unlock and its trigger, the participant twin of 000042.
Numbering, participant_count and the down migration are as before.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…able, simultaneous joins, the vote path, unknown ids answer 404, seed CSV fixes (migrations 000043, 000044)
@github-actions

Copy link
Copy Markdown
Contributor

Delphi Coverage Report

File Stmts Miss Cover
polismath/init.py 2 0 100%
polismath/benchmarks/bench_pca.py 128 107 16%
polismath/benchmarks/bench_repness.py 81 65 20%
polismath/benchmarks/bench_update_votes.py 38 28 26%
polismath/benchmarks/benchmark_utils.py 34 0 100%
polismath/components/init.py 1 0 100%
polismath/components/config.py 165 70 58%
polismath/conversation/init.py 2 0 100%
polismath/conversation/conversation.py 1293 264 80%
polismath/conversation/group_vote_counts.py 38 2 95%
polismath/conversation/manager.py 131 42 68%
polismath/database/init.py 1 0 100%
polismath/database/dynamodb.py 392 176 55%
polismath/database/postgres.py 414 95 77%
polismath/delphi_storage/init.py 1 0 100%
polismath/delphi_storage/codec.py 295 22 93%
polismath/delphi_storage/embedding_spool.py 58 6 90%
polismath/delphi_storage/golden_corpus.py 60 9 85%
polismath/delphi_storage/import_report.py 51 3 94%
polismath/delphi_storage/legacy_import.py 171 16 91%
polismath/delphi_storage/postgres.py 49 22 55%
polismath/delphi_storage/production_import.py 612 131 79%
polismath/delphi_storage/resource.py 179 32 82%
polismath/delphi_storage/scan_pacing.py 77 0 100%
polismath/delphi_storage/writer.py 196 13 93%
polismath/empty_output.py 29 2 93%
polismath/engine_adapter.py 317 265 16%
polismath/job_child/init.py 364 51 86%
polismath/job_child/census.py 74 9 88%
polismath/job_child/input_bundle.py 101 7 93%
polismath/pca_kmeans_rep/init.py 5 0 100%
polismath/pca_kmeans_rep/clusters.py 257 38 85%
polismath/pca_kmeans_rep/corr.py 98 17 83%
polismath/pca_kmeans_rep/group_k_smoother.py 28 0 100%
polismath/pca_kmeans_rep/legacy_kmeans.py 197 8 96%
polismath/pca_kmeans_rep/pca.py 162 19 88%
polismath/pca_kmeans_rep/repness.py 226 5 98%
polismath/poller/init.py 4 0 100%
polismath/poller/admission.py 318 15 95%
polismath/poller/backfill.py 936 32 97%
polismath/poller/capacity.py 485 27 94%
polismath/poller/capacity_queue.py 248 13 95%
polismath/poller/coordinator_bridge.py 317 317 0%
polismath/poller/math_writer.py 113 3 97%
polismath/poller/promotion.py 112 4 96%
polismath/poller/readiness.py 368 13 96%
polismath/poller/rebuild_child.py 133 5 96%
polismath/poller/service.py 756 54 93%
polismath/poller/worker_pool.py 140 0 100%
polismath/queue/executor.py 221 111 50%
polismath/regression/init.py 4 0 100%
polismath/regression/clojure_comparer.py 188 20 89%
polismath/regression/comparer.py 891 640 28%
polismath/regression/datasets.py 126 27 79%
polismath/regression/recorder.py 36 27 25%
polismath/regression/utils.py 130 85 35%
polismath/replay/init.py 2 0 100%
polismath/replay/attribution_capture.py 49 0 100%
polismath/replay/axis_continuity.py 347 24 93%
polismath/replay/certify.py 999 59 94%
polismath/replay/crosslang.py 128 2 98%
polismath/replay/decision_ties.py 51 8 84%
polismath/replay/diagnostics.py 50 7 86%
polismath/replay/driver.py 73 1 99%
polismath/replay/dsn_admission.py 36 1 97%
polismath/replay/equiv_query.py 10 0 100%
polismath/replay/event_ingress.py 95 1 99%
polismath/replay/fixture_bundle.py 815 62 92%
polismath/replay/fixture_config.py 266 11 96%
polismath/replay/fixture_extract.py 285 12 96%
polismath/replay/fixture_generate.py 103 1 99%
polismath/replay/fixture_samples.py 157 14 91%
polismath/replay/fixture_selection.py 119 0 100%
polismath/replay/fixture_survey.py 147 11 93%
polismath/replay/legacy_pca.py 29 17 41%
polismath/replay/polarity.py 319 25 92%
polismath/replay/poller_equiv.py 776 99 87%
polismath/replay/prodclone.py 193 2 99%
polismath/replay/real_data.py 146 10 93%
polismath/replay/schedule.py 205 10 95%
polismath/replay/served_math.py 159 4 97%
polismath/replay/shard_bench.py 189 65 66%
polismath/replay/stagecompare.py 890 161 82%
polismath/replay/stages.py 256 40 84%
polismath/replay/stepcompare.py 86 1 99%
polismath/replay/store.py 120 11 91%
polismath/replay/tie_capture.py 107 19 82%
polismath/replay/types.py 58 6 90%
polismath/run_math_pipeline.py 284 93 67%
polismath/tools/engine_rebuild.py 230 11 95%
polismath/types.py 6 0 100%
polismath/utils/cli_flags.py 15 0 100%
polismath/utils/clj_hash.py 44 0 100%
polismath/utils/env_flags.py 13 0 100%
polismath/utils/general.py 63 39 38%
polismath/utils/output_profile.py 66 2 97%
polismath/utils/serialization.py 10 5 50%
polismath/utils/vote_convention.py 159 5 97%
umap_narrative/700_datamapplot_for_layer.py 514 470 9%
umap_narrative/701_static_datamapplot_for_layer.py 310 310 0%
umap_narrative/702_consensus_divisive_datamapplot.py 435 275 37%
umap_narrative/801_narrative_report_batch.py 666 218 67%
umap_narrative/803_check_batch_status.py 268 148 45%
umap_narrative/llm_factory_constructor/init.py 2 0 100%
umap_narrative/llm_factory_constructor/model_provider.py 232 106 54%
umap_narrative/narrative_data.py 178 55 69%
umap_narrative/numerical_stages.py 80 14 82%
umap_narrative/polismath_commentgraph/init.py 1 0 100%
umap_narrative/polismath_commentgraph/cli.py 268 268 0%
umap_narrative/polismath_commentgraph/core/init.py 3 3 0%
umap_narrative/polismath_commentgraph/core/clustering.py 108 108 0%
umap_narrative/polismath_commentgraph/core/embedding.py 104 104 0%
umap_narrative/polismath_commentgraph/lambda_handler.py 219 219 0%
umap_narrative/polismath_commentgraph/schemas/init.py 2 0 100%
umap_narrative/polismath_commentgraph/schemas/dynamo_models.py 80 6 92%
umap_narrative/polismath_commentgraph/tests/conftest.py 17 17 0%
umap_narrative/polismath_commentgraph/tests/test_clustering.py 74 74 0%
umap_narrative/polismath_commentgraph/tests/test_embedding.py 55 55 0%
umap_narrative/polismath_commentgraph/tests/test_storage.py 87 87 0%
umap_narrative/polismath_commentgraph/utils/init.py 9 7 22%
umap_narrative/polismath_commentgraph/utils/converter.py 283 50 82%
umap_narrative/polismath_commentgraph/utils/group_data.py 332 279 16%
umap_narrative/polismath_commentgraph/utils/storage.py 609 481 21%
umap_narrative/reset_conversation.py 162 53 67%
umap_narrative/run_pipeline.py 355 203 43%
umap_narrative/topic_naming.py 148 11 93%
Total 25609 7377 71%

@colinmegill
colinmegill merged commit 833fd0e into stable Oct 24, 2026
56 of 62 checks passed
@colinmegill
colinmegill deleted the promote-2026-10-24 branch October 24, 2026 18:20
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant