Skip to content

Delphi shadow: a job graph that cannot finish no longer holds up other conversations' full-pipeline runs; optional same-host graph-stage daemon, off by default - #3447

Merged
colinmegill merged 8 commits into
edgefrom
delphi/job-graph-one-machine-recovery
Nov 18, 2026
Merged

colinmegill merged 8 commits into
edgefrom
delphi/job-graph-one-machine-recovery

Conversation

@colinmegill

Copy link
Copy Markdown
Member

Observable changes

None with today's settings. DELPHI_SHADOW_GRAPH_CONVERSATIONS is empty and DELPHI_SHADOW_GRAPH_ENABLED is not set, so the new code makes no query, opens no connection and writes no log line, the scheduler's state file is byte-identical, and the deploy hook skips the new block. Rendered with docker compose config under every profile with production-shaped settings, this branch differs from edge only by the new service delphi-shadow-graph (its own profile, off) and its volume delphi-shadow-graph-state; networks are identical, and the hook's existing shadow settings check names nothing.

An operator who turned them on would see:

  • DELPHI_SHADOW_GRAPH_ENABLED=1 (locally: the delphi-shadow-graph profile beside delphi-shadow): a delphi-shadow-graph container that claims the four graph steps (graph_embed, graph_cluster, graph_topics, graph_narrative) in the graph environment, one step at a time, as PID 1 of its own namespace, offline, with topic model none and no provider key. Defaults: 6 GiB (DELPHI_SHADOW_GRAPH_CONTAINER_MEMORY), 2 CPUs (DELPHI_SHADOW_GRAPH_CONTAINER_CPUS), the same lowest CPU weight and highest OOM score as the full-pipeline shadow. Its attempt journal lives on the named volume, bound to the host by /etc/machine-id (or the file named in DELPHI_SHADOW_GRAPH_HOST_ID_FILE); a journal bound to another host, or holding entries written before the binding existed, is refused (journal_host_mismatch, journal_host_unrecorded), and a second daemon cannot take the journal's lock. It receives DELPHI_EMBED_MODEL_BY_ENV exactly as the scheduler and the full-pipeline shadow do. At deploy, scripts/check-shadow-graph.py validates the rendered settings and prints names only, never values.
  • The conversation list set as well: a graph node still running at DELPHI_SHADOW_JOB_DEADLINE_SECONDS (default 14,400) is cancelled through the existing fenced cancel, logged once as graph_recovery_pending, and followed in the background. Other conversations' full-pipeline runs resume. That conversation stays out of the full-pipeline shadow while its graph scope is held, and no new graph is admitted while one is deferred. Publication is accepted only when the served root is this graph's own, read back after every outcome. A publish that fails with a transient database error before commit is tried up to three times in all (graph_publish_retry), then counted failed (graph_publish_failed); a served read that itself fails keeps the graph to be resolved on a later poll. A conversation refused admission is looked at again once per build, so every deploy re-evaluates each listed conversation once.
  • Rollback to a release without this code, once any graph has run: drain every graph, deferred ones included, and confirm its scope was released first. Older schedulers drop the deferred fields from their state and cannot keep those exclusions.

No schema change

No migration, no SQL function added or replaced, no table or role touched. No Node handler and no Rust API route changed.

What it does

The Delphi shadow is a background scheduler that runs the Delphi pipeline on chosen conversations into shadow result environments that readers never serve. It can also run a listed conversation as a four-node job graph in a separate shadow environment. That path was already in edge but switched off. This branch fixes two failure modes in it and adds a way to run its steps on the same host:

  1. A graph that never finishes no longer holds up the scheduler. Before, a graph node whose process was gone was waited on indefinitely, and the scheduler admitted nothing else. Now the foreground wait ends at the job deadline. The overdue graph is cancelled and deferred, its scope stays held, and it is polled one per cycle in the background. A scope is released only by the existing database reconciliation, which requires every attempt's exit to be confirmed; the clock never releases a scope.
  2. An uncertain admission does not fall through to a full-pipeline run of the same conversation. The graph's environment is written to the scheduler's state before admission. Graphs in every remembered environment are recovered when the scheduler starts, in each cycle, after an admission error and before the comparison step, and every full-pipeline candidate is checked against held graph scopes. So a committed graph whose reply was lost is found, and its conversation stays excluded.
  3. Publication is recognised by what is served. After every publish outcome the served bundle is read on a fresh connection and accepted only if its root artifact belongs to this graph at the next generation. A partial or superseded bundle is still refused by SQL.
  4. An optional graph-stage daemon on the same host, its own Compose service and profile, as above. Its recovery after a restart or a replaced container uses the retained journal: on one host, a replaced PID 1 container's processes are gone, and a restarted container is a fresh namespace. A lost journal or host is never taken as proof that a process exited.
  5. Recovery log lines say why, and the recovery procedure is in docs/configuration.md ("Same-host graph executor and recovery").

Test harness: the deploy-hook test runner now copies scripts/check-shadow-graph.py, so the graph executor's Compose tests run in the Delphi test container. The repository check that every graph admission goes through the shared entry now lists delphi/tests/shadow/graph_recovery.py beside the other graph proofs, because that proof admits one test-model node directly through SQL. That proof needs the test build of the Delphi image.

The vote, comment and next-comment paths

Nothing here is on them. No Node or Rust API code changed, and there is no migration. With the list empty, the new code makes no query at all. Once it is set, the scheduler's new database work is read-only, on its existing read connection, one connection at a time, REPEATABLE READ, with statement_timeout and lock_timeout of 1 s, against queue and graph tables only. The graph daemon writes only through the existing fenced queue functions, under its own queue login. No connection pool is shared with votes, comments or the next comment, and no lock is taken on conversations, votes, comments or participants. The graph daemon shares the full-pipeline shadow's CPU weight and OOM score, which every other service outranks, and runs within its own memory and CPU ceilings: it yields first, though that reserves no capacity for participation. No vote is read or written. The vote-sign lint reports 0 new sites, and the recovery proof asserts the votes table is empty. The Delphi shadow integration tests in queue-rs/tests/delphi_shadow.rs, which include the tests that the scheduler's sessions never make a vote, a comment or the next comment wait, pass.

Checks

At this branch's head, on an arm64 Docker host that other work was also using (load 10 to 47). Counts overlap across suites.

Check Result
make check-vote-sign-lint pass: 32 unittest; 581 sites, 0 new
make check-lint pass: 0 errors, 1 warning (an existing pending Cypress test)
python3 ci/check_graph_registry.py and its unittest pass: 0 errors; 6 tests
make check-queue-rs (includes the Delphi shadow scheduler and daemon integration tests) pass: Rust 323 passed, 0 failed, 1 ignored (18 binaries; delphi_shadow 20, jobs_integration 27); Clippy in both feature sets
make check-job-graphs first run failed under load in the existing test jobs::journal::tests::one_daemon_per_journal_directory ("free again once released", queue-rs/src/jobs/journal.rs:236). Re-run alone: pass, Rust 303 passed, 1 ignored; Jest 48; 18 proof records
make check-polis-cli pass: Rust 620; unittest 142
Delphi image, --target final and --target test built
Compose render and deploy-check probe, production-shaped settings, against edge pass: only delphi-shadow-graph and its volume differ. The check accepts the graph on with defaults and with a list. It refuses the graph environment equal to the shadow's or to a served one, CPUs 0, the profile not rendered, and production's settings with the graph off
make check-embeddings-table pass: real-model graph, write and read-back, re-run adds only new, switch-off parity; COMPARE OK
make check-dynamo-writers pass: Rust 111; Jest 93 passed, 7 skipped, and 7 passed; every scenario PASS
delphi/tests/shadow/graph_recovery.py (test image; PostgreSQL 17 with this release's migrations) pass: 7 PASS: container replaced, same container restarted, journal lost (second daemon refused in each; the exact attempt recovered in the first two; in the third, nothing invented exit proof or released the scope); no votes
delphi/tests/shadow/graph_scheduler_limits.py (final image, 256 MiB and 0.25 CPU) pass: 4 assertions (restart, list cleared, other conversation admitted, orphan scope kept, the limits read back from Docker)
delphi/tests/shadow/rehearsal.sh, unchanged The first run stopped at its offline check under load. Internet was blocked and the download was refused, but the probe (docker run … 2>&1 | tail -1, line 119) took a Hugging Face retry warning as its last line instead of refused:OSError. Run alone, the same probe on the same image answers refused:OSError. Re-run: pass. 18 of 18 shadow jobs succeeded; granite embeddings 1,208 rows in 18 sets; the served pointer is unchanged; internet blocked at start and end; image and by-name vectors identical over 441 texts; no graph_ log line from the scheduler
make check-delphi-python-tests deploy-hook tests 525 passed, 67 skipped. Shards 1 to 5: 15, 1,118, 387, 887 and 1,520 passed, 0 failed, 0 errors. Shard 6: 783 passed, 1 error; re-run alone in the same run: 782 passed, 2 errors. Every error is a setup in tests/poller/test_promotion_postgres.py failing set_payload_type's assertion (tests/poller/test_backfill_postgres.py:55, [('json',), ('jsonb',)]): the shared-table race between those tests' json and jsonb cases, already red on edge and outside this change. 5,064 tests collected

Known limits

The graph list is empty in production and the graph daemon's profile is off. Each of these is a condition to meet before the list is ever set:

  1. A lost journal holds the graph path for good. If the host is replaced, or the journal volume is removed, while a graph node runs, that attempt's exit can never be confirmed. Its scope is never released, no new graph is admitted in any graph environment, and that conversation stays out of the full-pipeline shadow indefinitely. Changing the graph environment does not help, and admission logs nothing about why it is holding. Needed: an operator action for a lost journal, recorded as its own kind of evidence and never as a process exit, plus a log line naming the reason whenever admission holds for a deferred graph.
  2. The host binding is a file's contents. A root volume restored on a second machine, or an image that ships a fixed /etc/machine-id, passes the binding. A different boot then counts as exit proof for an attempt that may still be running. Writes stay safe: finalize is fenced by owner, attempt and epoch and needs that attempt's exit proof, each node has one terminal attempt, and graph children hold no database setting. The harm is a false record and two machines computing the same node. Needed: an entry from another boot is confirmed only after the attempt's lease has expired by database time.
  3. Some recovery behaviours have no test that fails without them: recovery after an admission error whose reply was lost; no new graph while one is deferred; the graph environment saved before admission; recovery before the comparison step; a deferred graph's finish not taking the rest pause; and recovery searching remembered environments after the list is emptied. Needed: Rust tests in queue-rs/tests/delphi_shadow.rs for at least the lost admission reply, the emptied list with no flight in state, and the hold while a graph is deferred.
  4. No check runs the two Docker proofs. graph_recovery.py and graph_scheduler_limits.py are run by hand, and the limits proof cannot reuse its output directory. Needed: one make target that builds the images, applies the release to an owned PostgreSQL and runs both, plus a row in docs/local-checks.md.
  5. A single small machine is not documented. The default ceilings are 16 GiB for the full-pipeline daemon, 6 GiB for the graph daemon and 256 MiB for the scheduler: 22.25 GiB before PostgreSQL, the math engine and the web services. These are configured ceilings, not measured residency. The full-pipeline daemon's setting, DELPHI_SHADOW_CONTAINER_MEMORY, is not in docs/configuration.md or example.env, and macOS Docker Desktop has no /etc/machine-id. Needed: a short section naming both memory settings, a measured pair that works, and that the graph daemon is optional.
  6. A missing graph daemon stalls the full-pipeline shadow. With the list set but the daemon absent or unable to start, each listed conversation is admitted as a graph that nothing claims. The scheduler then admits no full-pipeline run until that graph's job deadline. Separately, an invalid graph setting makes the deploy start neither shadow service. Needed: admit a graph only when the graph environment has a ready daemon (or give an unclaimed graph a short deadline), and start the full-pipeline shadow when only the graph settings are refused.
  7. Publication retries are back to back. The three tries run without a pause, so a database failover outlasts them. The graph is then counted failed, and because its snapshot is remembered, it is not published again while the conversation is unchanged. Needed: keep the flight and retry on later polls, with the count carried in the flight.

🤖 Generated with Claude Code

https://claude.ai/code/session_01KvftUAdasWzxmMYdioZJWs

colinmegill and others added 8 commits November 18, 2026 02:12
…d holds only its own conversation

When a graph's attempt cannot be proven finished (its scope stays held), the
scheduler no longer waits on it forever. After the job deadline the graph moves
to a deferred list that is polled one per cycle, and the full-pipeline shadow
keeps admitting every other conversation; only the conversation whose graph
scope is held is skipped. The environments the graph path has used are recorded
in the state file before any admission, so a graph admitted under a list that
was later emptied is still found and recovered after a restart.

The job daemon can bind its retained journal to one host identity file
(POLIS_JOBS_JOURNAL_HOST_FILE): a journal written under another host is refused
at start-up rather than read as exit evidence.

Adds Rust tests for a lost attempt after a restart with the list emptied and for
a committed admission whose reply was lost, and two Docker process checks for
recovery and for the scheduler under its deployed memory and CPU limits.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KvftUAdasWzxmMYdioZJWs
…ournal volume and deploy check

Adds the delphi-shadow-graph Compose service (profile delphi-shadow-graph, off
unless DELPHI_SHADOW_GRAPH_ENABLED=1): the job daemon restricted to the four
graph stages at concurrency one, with its own journal volume, a read-only host
identity file bound in for the journal's host binding, and its own memory and
CPU ceilings (6 GiB and 2 CPUs by default). The daemon refuses to start unless
it runs as PID 1 of its own container with exactly those stages, concurrency
one and a host identity file set.

The deploy hook starts it only when the full-pipeline shadow is selected and the
setting is on, and first checks the rendered Compose settings with
scripts/check-shadow-graph.py (names only); if they are wrong, neither shadow is
started. With the setting off, nothing in the deploy changes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KvftUAdasWzxmMYdioZJWs
… retried a bounded number of times; refusals are reconsidered by build

A finished graph's publication is now confirmed by reading the served bundle
and checking that its root artifact is this graph's leaf at the next
generation, not by the publish call's reply alone. A publish whose commit is
durable but whose reply was lost is recognised as published; a transient
database error before commit (connection, serialization, deadlock, lock or
too-many-connections classes) is retried at most three times, each on a fresh
connection, and then counted failed. A definite refusal is not retried. If the
served bundle cannot be read, the graph stays in flight and is resolved on a
later poll.

A refused conversation's recorded observation now carries the admission rule
and the build's code version, so a later build looks at it again once. A graph
admission that errored is followed by a recovery read, so an admission that
committed before its reply was lost is picked up before any full-pipeline run.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KvftUAdasWzxmMYdioZJWs
… the recovery runbook is documented

A graph waiting past its deadline now logs the reason it is still held (nodes
still active, a process exit not yet confirmed, or reconciliation pending) and
the count of attempts with no confirmed exit, instead of one fixed reason. A
job daemon with a host-bound journal logs one line per retained journal entry
saying whether the attempt's exit was proven and why. docs/configuration.md
gains a section on the same-host graph executor: its profile, host identity
file, journal volume, deadlines, memory ceilings and what to drain before a
rollback.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KvftUAdasWzxmMYdioZJWs
…st container

The graph executor's validator test located scripts/check-shadow-graph.py by
its own path, which inside the Delphi test container resolves to /scripts and
does not exist, so every shard of the Delphi Python tests stopped with a
collection error. It now reads the validator from the checkout the other
Compose tests read ($POLIS_CHECKOUT_DIR), and the deploy-hook runner copies the
validator there with its other inputs.

The test that the shadow yields CPU and memory to every other service failed
once the graph executor was added, because the executor carries the shadow's
own lowest weight. Both shadow daemons must now carry exactly that weight and
OOM adjustment, and every other service must still outrank them; no assertion
about any other service changed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KvftUAdasWzxmMYdioZJWs
Brings in the typed graph step and model registry. Three files conflicted
only because both sides appended at the same end: docs/configuration.md,
example.env and queue-rs/tests/delphi_shadow.rs. Each keeps edge's appended
text first, unchanged, followed by this branch's appended text, unchanged.
docker-compose.yml, queue-rs/src/shadow/graph.rs and scripts/after_install.sh
merged without conflict.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KvftUAdasWzxmMYdioZJWs
…ing model setting as the shadow daemon does

DELPHI_EMBED_MODEL_BY_ENV is now passed through to delphi-shadow-graph the way
the registry passes it to delphi-shadow and the scheduler: listed by name, with
no default of its own, so with the setting unset (today) it is absent from the
container and nothing changes. Each graph's model is still decided and sealed
by the scheduler at admission, and graph children's environment stays the
existing allowlist. The deploy check refuses a graph executor whose setting
differs from the scheduler's; tests cover the check and that both daemons list
the setting the same way.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KvftUAdasWzxmMYdioZJWs
…de directly

delphi/tests/shadow/graph_recovery.py admits one test-model node through the
SQL admission function so that a test-build daemon has a real attempt to lose
and recover, as the graph proofs already do. The check that every product
admission goes through the shared entry now lists it beside those proofs. The
proof's description says it needs a test-build image, since a production
daemon refuses to run a test-only model.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KvftUAdasWzxmMYdioZJWs
@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 614 131 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/embedding_model.py 197 49 75%
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 7 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 120 3 98%
polismath/poller/promotion.py 112 4 96%
polismath/poller/readiness.py 368 13 96%
polismath/poller/rebuild_child.py 133 5 96%
polismath/poller/recompute_line.py 141 8 94%
polismath/poller/service.py 760 50 93%
polismath/poller/worker_pool.py 143 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/step_log.py 95 7 93%
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 657 209 68%
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 97 17 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 616 486 21%
umap_narrative/reset_conversation.py 162 53 67%
umap_narrative/run_pipeline.py 391 228 42%
umap_narrative/topic_naming.py 148 11 93%
Total 26109 7460 71%

@colinmegill
colinmegill merged commit 61a4792 into edge Nov 18, 2026
41 checks passed
@colinmegill
colinmegill deleted the delphi/job-graph-one-machine-recovery branch November 18, 2026 13:32
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