Skip to content

offline profile 4: the offline compose overlay, the image bundle script and the install guide (stacked on #2929, #2937, #2936) - #2938

Open
colinmegill wants to merge 14 commits into
edgefrom
offline-4-overlay-bundle
Open

colinmegill wants to merge 14 commits into
edgefrom
offline-4-overlay-bundle

Conversation

@colinmegill

@colinmegill colinmegill commented Oct 3, 2026 •

Copy link
Copy Markdown
Member

Stacked on #2929, #2937 and #2936. This branch merges those three branches, so until they land the diff shows their changes too. This PR's own changes are the last three commits (bedebc536, 850d05bd8, a373adc55).

This is the fourth offline-profile PR. It adds:

It builds on the sibling flags exactly as they are named:

  • OFFLINE (1/true) for the client build arg, the server and delphi;
  • BAKE_EMBEDDING_MODEL;
  • USE_CPU_TORCH.

Services per tier

Service Tier Offline settings
postgres core profile reset: always on; the only store
server core prod build, OFFLINE=1, trusts the issuer's mkcert CA, mounts the JWT keys
file-server core build arg OFFLINE=1; GA and Datadog RUM args empty
nginx-proxy core —
client-participation-alpha core kept because nginx proxies /alpha/ to it by name and nginx will not start if that name does not resolve
math-python core profile reset; MATH_WORKER_POOL_SIZE=1; 1536m limit; MATH_POLLER_ALLOW_HOSTNAME_IDENTITY=1; MATH_ENV shared with the server and delphi (python)
oidc-simulator core admin issuer for now: built image, no source mount, mkcert certificates; offline PR 9's local issuer replaces it
delphi offline-topics CPU torch, baked embedding model, OFFLINE=1, LLM_PROVIDER=ollama
ollama offline-topics model shipped in the ollama-models volume, never pulled on the box
dynamodb, minio offline-topics stay until P-076 and R-8 land

Not started in either tier: ses-local, localstack, import-worker, coordinator-idle. Every image the overlay builds is named polis-offline/* in place of the ECR and Docker Hub names, so the rendered config names no hosted registry.

The Makefile

make OFFLINE start reads offline.env, or OFFLINE_ENV_FILE; example.offline.env is the template, and offline.env is gitignored. It runs -f docker-compose.yml -f docker-compose.offline.yml, adds --profile offline-topics only when OFFLINE_TOPICS=true, and exports OFFLINE=1 BAKE_EMBEDDING_MODEL=true USE_CPU_TORCH=true.

One side fix: when PROD, TEST or OFFLINE has already chosen the profiles, get_env_vars no longer appends them a second time.

  • make PROD start used to print --profile postgres --profile local-services twice; it now prints them once.
  • make start renders the identical command (checked with make -n).

The bundle

scripts/offline-bundle.sh save [--topics] [--with-models] writes one tar. It holds:

  • images.tar: docker save of every image docker compose config --images lists for the tier;
  • manifest.tsv: image, content id, repo digest;
  • source.tar: git archive HEAD;
  • optionally models.tar, the ollama volume;
  • SHA256SUMS.

load checks the sums, runs docker load, and refuses to finish unless every manifest image is present with the same content id. The images are pinned by that id. In the overlay, minio is named by its tag rather than tag@digest, because docker load does not always restore a repo digest and compose would then try to pull. The manifest still records the digest.

Test

ci/offline_compose_config_test.sh uses docker compose config and make -n only. It checks:

  • the core tier lists exactly the 7 services above;
  • offline-topics lists exactly the 11;
  • the rendered core config has no amazonaws|auth0|anthropic|simpleanalytics|googleapis;
  • topic-tier images name no hosted registry;
  • every core image is polis-offline/*;
  • OFFLINE=1 on server, math-python and delphi; MATH_ENV=python on all three;
  • pool 1 and the 1.5 GB limit; Ollama with no Anthropic key;
  • matching delphi/math-python build args; the file-server OFFLINE arg;
  • that make -n OFFLINE start adds no dev, local-services or ollama profile.

Local result: all checks pass. When the overlay was broken by hand (an ECR image name, pool 4), the test failed with 4 failures. A new Offline compose config workflow runs it on pull requests that touch these files.

Acceptance checklist (docs/offline.md)

The acceptance test: a participant can open a conversation, vote, and see groups and consensus with the network cable unplugged; an admin can create a conversation and read the report.

  • make OFFLINE start: every core service comes up and an egress-deny firewall log stays empty
  • A participant opens /<conversation_id> from a phone on the box's Wi-Fi; every request goes to the box (known exception below)
  • Votes land and math-python writes a new math_main row
  • Groups and consensus show
  • An admin logs in (simulator account), creates a conversation and seeds comments
  • The report renders its graphs from the bundled d3/Plotly; the report CSVs download
  • Topic tier: Delphi names topics with the local model (or keyword labels)
  • After a restart with the cable unplugged, everything returns and nothing is downloaded
  • No third-party host appears in the shipped bundles beyond the known exceptions

Sizing: the core tier needs 4 GB RAM and 2 vCPU (x86-64). With topics and llama3.1:8b, it needs 16 GB.

What is still outstanding

  • Offline PR 9, a local admin issuer. The simulator has fixed accounts and a fixed password. Its HTTPS needs mkcert certificates, so mkcert is downloaded on the build machine and every admin device must trust its CA. PR 9 needs a ruling first.
  • Offline PR 10, the data export path. Admin data export still writes to AWS S3 and fails offline. The report CSVs work.
  • Offline PR 11, Delphi plots on the filesystem (R-8). Plots are in MinIO today.
  • Offline PR 12, P-076. It moves the DynamoDB queue and stores to Postgres. That is a schema change, with per-table rulings.
  • Offline PR 13, an Ollama provider for the narrative report and the collective statement. Both fail offline today.
  • Not yet covered by any PR: the legacy participant page still loads the pol.is logo and favicon as <img> tags (participation.handlebars:53,71, conversationTabs.handlebars:116). Offline they are broken images, so checklist item 2 cannot pass strictly until they are self-hosted.

Not run (laptop budget): no image builds, no stack, no bundle save or load. bash -n passes on the bundle script.

Review fixes (a373adc)

  • Bundle script exit status. save and load exited 1 on success and left their scratch directory behind, because the EXIT trap used a function-local variable. Both now exit 0 and clean up.
    • Checked with a load round trip of a small local image: rc 0, nothing left behind. Before the fix: rc 1, directory leaked.
  • Pinned upstream images. New offline-images.lock pins ollama, dynamodb-local and minio by digest. offline-bundle.sh pull fetches exactly those digests, and save refuses any other copy. save also prints the archive's sha256 so it can be compared out of band.
  • Restart policies. nginx-proxy, file-server, dynamodb and minio now have restart: unless-stopped, so a reboot brings ports 80/443 back.
  • nginx certificate. Port 443 serves the mkcert certificate.
  • Admin login over https. Admins use https://<box>/, because oidc-client-ts needs a secure context. The checklist now says to verify login from a second device.
  • Wider test. The hosted-endpoint grep adds datadoghq, huggingface, ollama.com, cloudflare, jsdelivr and unpkg. The test also checks a restart policy on every service and that every upstream image is pinned. It still passes.
  • Guide additions.
    • commands for checklist items 1, 3, 4, 8 and 9;
    • the box name is fixed at build time;
    • make and bash are needed on the box;
    • the side effects of DEV_MODE=true and how to avoid them;
    • peak disk use during load.

No change to the default stack or production. No schema change.

🤖 Generated with Claude Code

https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s

colinmegill and others added 13 commits October 3, 2026 01:58
The legacy participant page (the default /<conversation_id> page) loaded
jwt-decode@latest from cdn.jsdelivr.net with a synchronous <script> in
<head> (public/index.ejs:36). Offline, that blocks parsing until the
request fails, and hangs where a firewall drops packets.

The global it provided is still used once: index.ejs decodes the OIDC
access token to read the delphi_enabled claim, and
js/views/participation.js:984 hides the donate box when it is set. So
the tag is replaced, not just dropped: a ten-line inline decoder reads
the payload the same way jwt-decode 3.1.2 (what jsdelivr serves for
@latest at this path) does, base64url then UTF-8 then JSON, and throws
on a missing or malformed token as before, which the existing
try/catch logs.

No behaviour change for pol.is: the donate box is hidden for the same
users. No schema change.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
…ith OFFLINE

client-report (public/index.html:15-16) and the legacy participant page
(client-participation/public/index.ejs:21-22) loaded the Simple
Analytics script and noscript image unconditionally, in every build
mode. client-admin already gated them on isProduction
(public/index.html:15-26).

All three templates now test one template option, simpleAnalytics,
which each webpack config sets to: production mode, and OFFLINE is not
1/true. OFFLINE is a new build-time flag, declared as a build ARG in
each client stage of file-server/Dockerfile and documented in
docs/configuration.md under "Client Build Flags". Compose does not pass
it yet; the offline overlay will.

client-report's HtmlWebpackPlugin moves from webpack.common.js into the
production and development configs through a small htmlPlugin(mode)
helper, so the page knows its mode.

No behaviour change for pol.is: the file-server image builds every
client with build:prod and OFFLINE unset, so the tag stays. Development
builds and dev servers of the report and legacy page no longer load it,
matching client-admin. No schema change.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
…FFLINE

parse_env_bool returns the string "true" or "false" (Makefile:15-16),
but the profile lines tested it with $(if ...), which is true for any
non-empty string (Makefile:36-37 and 59-60). So --profile postgres and
--profile local-services were always added, even with
POSTGRES_DOCKER=false or LOCAL_SERVICES_DOCKER=false. All four lines now
use $(filter true,...), as the USE_PRODCLONE lines already do.

Generated `make -n start` commands, before and after:
- .env with POSTGRES_DOCKER=true, LOCAL_SERVICES_DOCKER unset, and
  example.env (both true): unchanged, both profiles. Unset still means
  true. `docker compose ... config --services` with that command lists
  the same fifteen services.
- both false: before, both profiles; after, neither.
- POSTGRES_DOCKER=true, LOCAL_SERVICES_DOCKER=False: before, both;
  after, postgres only.

Adds an OFFLINE target in the PROD/TEST position. Until the offline
compose overlay lands it only prints what `make OFFLINE start` will do,
and exits 2 so the following goal cannot fall through to the
development stack.

No behaviour change for pol.is or for a default development checkout.
No schema change.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
… OFFLINE; ollama behind a profile

On a local stack (DYNAMODB_ENDPOINT set) the image CMD ran
scripts/setup_ollama.sh on every start (delphi/Dockerfile:115-118),
which pulls llama3.1:8b from registry.ollama.ai, even with
LLM_PROVIDER=anthropic. And the ollama service had no profile
(docker-compose.yml:447), so every `docker compose up` started it.

- delphi/Dockerfile: the CMD runs setup_ollama.sh only when
  LLM_PROVIDER=ollama, and never when OFFLINE is 1 or true. Each skip
  logs why. The MinIO setup and the job poller run as before. The
  production branch (DYNAMODB_ENDPOINT unset) is unchanged.
- docker-compose.yml: the ollama service is behind the `ollama`
  profile. The delphi service forwards OFFLINE (fallback empty);
  LLM_PROVIDER is already forwarded with fallback anthropic.
- Makefile: `make start` adds --profile ollama when the env file sets
  LLM_PROVIDER=ollama, so a development stack that uses Ollama still
  starts it.
- docs/configuration.md and example.env describe both.
- delphi/tests/test_compose_math_env.py runs the image CMD with stub
  python and setup_ollama.sh across provider and OFFLINE values, checks
  the production branch never runs the setup, that the ollama service
  is profile-gated and no stack starts Ollama without a profile, and
  that delphi forwards OFFLINE. Nine of the new cases fail against the
  previous Dockerfile and compose file.

No behaviour change for pol.is: production deploys start named
services (scripts/after_install.sh), never ollama, and run the image
with DYNAMODB_ENDPOINT unset. A development stack with the default
LLM_PROVIDER=anthropic no longer starts the ollama container or pulls
a model. No schema change.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
OFFLINE=1 (or true) is for a box with no network (offline profile item 5).
Unset, nothing changes. When set, the server skips:

- Auth0 Management API in isProConvo (routes/comments.ts), reached on every
  participant comment post and on comment lists filtered by mod_gt: it answers
  false without a lookup. False is what the function already returns on every
  error path, so the box keeps today's offline result minus the network
  timeout, and the Gemini / ip-api.com moderation it gates is never reached.
- Akismet verifyKey at module load (server.ts); nothing else calls Akismet.
- dd-trace init in production mode (index.ts), via shouldInitTracer.
- Google Translate: SHOULD_USE_TRANSLATION_API is ignored (comment.ts).
- SES: with SES_ENDPOINT unset, sends become a debug log of the subject only
  and resolve with no message id. A set SES_ENDPOINT (ses-local) still gets
  the mail.

Each skip is announced once at startup (utils/offline.ts), never per request.
Compose forwards OFFLINE to the server service, empty by default.
docs/configuration.md gains an "Offline mode (server)" section.

Tests: __tests__/unit/offlineFlag.test.ts loads each real module with the flag
on and off; delphi/tests/test_compose_offline.py checks the compose passthrough.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
…N tags

The report loaded d3 v4 (d3js.org), d3-scale-chromatic v1 (d3js.org),
Plotly 2.27.0 (cdn.plot.ly) and Font Awesome 4.4.0 (maxcdn) from tags in
public/index.html. With no network the graphs and topic plots fail.

- Pin the versions those URLs served: d3 4.13.0 (was ~4.6.0 in
  package.json, while the page ran the 4.13.0 global), d3-scale-chromatic
  1.5.0 and plotly.js-dist-min 2.27.0 (its plotly.min.js is byte-identical
  to cdn.plot.ly/plotly-2.27.0.min.js).
- src/vendorGlobals.js, the first import of src/index.js, sets window.d3
  (d3 plus the chromatic schemes, as the two UMD scripts did, with d3.event
  kept live) and window.Plotly. The call sites and the unit tests that stub
  those globals stay as they are.
- Drop the webpack `externals: { d3: "d3" }` so d3 is bundled.
- Remove the Font Awesome stylesheet: no fa- class or react-fontawesome
  component is used anywhere in client-report, so no icon set is needed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
public/index.html loaded Space Mono (400, 700, 400 italic) from
fonts.googleapis.com. Ship the same woff2 files Google Fonts served (latin,
latin-ext and vietnamese subsets for each style) under
public/font/space-mono/ with the SIL OFL licence, and declare them with
@font-face in the page's existing <style> block, keeping the same unicode
ranges and font-display: swap. CopyPlugin already copies public/ into the
build, and the server already proxies /font/* to the file server.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
Offline profile item 7. SentenceTransformer downloads all-MiniLM-L6-v2 from
the Hugging Face hub on first use (embedding.py:63-78), so a box with no
network fails its first run.

- Dockerfile (final stage): ARG BAKE_EMBEDDING_MODEL=false and
  ARG SENTENCE_TRANSFORMER_MODEL=all-MiniLM-L6-v2. With true, the model is
  downloaded at build time into the default HF cache (about 92 MB for
  MiniLM-L6); a failed download fails the build. With the default the step
  runs nothing, so the production image contents are unchanged.
- Compose passes both args to delphi and math-python (same image tag), and
  forwards OFFLINE to delphi with fallback empty. That line is byte-identical
  to the one in #2929, so the two merge cleanly in either order.
- polismath/utils/offline_env.py: with OFFLINE 1 or true, the job poller and
  run_delphi.py set HF_HUB_OFFLINE=1 and TRANSFORMERS_OFFLINE=1 on their own
  environment before spawning any pipeline step (the HF libraries read them at
  import), and the embedding engine loads with local_files_only.
- docs/configuration.md: "Offline embedding model (Delphi)", including the
  image-size cost and building with USE_CPU_TORCH=true for an offline box.

Tests (delphi/tests/test_compose_offline.py): compose forwards OFFLINE and the
model name, build args default to false on both services and match, the
Dockerfile defaults match compose, the bake RUN runs python only for true and
fails the build on a failed download, the flag values, the spawners apply the
flags before spawning, and the engine's local_files_only follows OFFLINE.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
…e source

Review follow-up.

- The server parsed OFFLINE with the `boolean` package, which also accepts
  yes/y/t, while the Delphi image and job poller accept only 1 and true, so
  OFFLINE=yes would have switched the server off-network but not Delphi. The
  server now accepts the same two values (any case, trimmed); tests cover
  yes/on/y/t staying off. docs/configuration.md no longer claims yes/on.
- delphi/Dockerfile: the model download moves to its own `embedding-model`
  stage built on the dependency layer (`builder-deps`), before any project
  source is copied, so a code change no longer re-downloads the model. The
  final stage copies /models/ into /root/ ahead of the packages; with the
  default the directory is empty and the copy adds no files. Tests check the
  stage order and the copy.

docs/configuration-env-reference.md is not edited: it is a census pinned to a
source snapshot with counted totals, so a hand edit would make it disagree with
its own snapshot.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
…gin/offline-2-server-flag-baked-model' and 'origin/offline-3-vendor-report-libs' into offline-4-overlay-bundle
docker-compose.offline.yml runs Polis on one machine with no network.
Core tier: postgres, server, file-server, nginx-proxy,
client-participation-alpha (nginx proxies to it by name) and math-python
(pool 1, 1.5 GB limit, hostname identity, the shared MATH_ENV), with the
oidc-simulator as the admin login issuer. --profile offline-topics adds
delphi (CPU torch, baked model, Ollama), ollama, dynamodb and minio.
Built images get local polis-offline/* names; OFFLINE=1 reaches the
server, math-python, delphi and the client build.

make OFFLINE start replaces the placeholder target: it reads offline.env
(example.offline.env is the template), runs the overlay with only the
offline-topics profile when OFFLINE_TOPICS=true, and exports OFFLINE=1,
BAKE_EMBEDDING_MODEL=true and USE_CPU_TORCH=true. Profiles chosen by
PROD/TEST/OFFLINE are no longer appended a second time; make start is
unchanged.

ci/offline_compose_config_test.sh (parse only) checks the services per
tier, that the rendered core tier names no hosted endpoint, and the
offline settings; a small workflow runs it on pull requests.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
scripts/offline-bundle.sh save writes one archive: docker save of every
image the tier runs, a manifest of content ids and repo digests, git
archive of the checkout, optionally the ollama model volume, and
checksums. load checks the checksums, runs docker load and refuses to
finish unless every image is present with the manifest's content id.

docs/offline.md: tiers, machine sizing, build and install steps, the
acceptance checklist, and what still reaches out today (the issuer's
mkcert certificates, the legacy page's pol.is images, data export,
Delphi plots, the emulators, narrative and collective statement).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
- Bundle script: save and load exited 1 on success and left their
  scratch directory behind, because the EXIT trap named a function-local
  variable. The scratch dir is now global, cleaned by a top-level trap,
  and the script exits 0. New `pull` fetches the upstream topic images
  at the digests in offline-images.lock; save refuses any other copy, so
  two bundles carry the same bytes. save prints the archive's sha256;
  the usage text no longer prints the script's first command.
- Overlay: restart: unless-stopped on nginx-proxy, file-server, dynamodb
  and minio, so a reboot brings ports 80/443 back. nginx serves 443 with
  the mkcert certificate.
- Config test: the hosted-endpoint grep also covers datadoghq,
  huggingface, ollama.com, cloudflare, jsdelivr and unpkg; it checks a
  restart policy on every service and that upstream images are pinned.
- Guide: admin login over https (oidc-client-ts needs a secure context),
  verified from a second device; the certificate names the box; the box
  name is fixed at build time; make and bash on the box; commands for
  checklist items 1, 3, 4, 8 and 9; DEV_MODE=true side effects and how to
  avoid them; peak disk during load.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
…ocal narratives

example.offline.env sets OLLAMA_HOST=http://ollama:11434: the server
reads it for the narrative and collective statement and would otherwise
use localhost. docs/offline.md takes the lines the data export PR
(#2940) and the Ollama narrative PR (#2945) left for it: checklist item 6
adds the conversation data export zip, item 7 has the narrative and
collective statement from the local model, and the known gaps drop the
Anthropic-only line and describe the streamed export.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018NVGBuYCk4EZmiz4csUv9s
@github-actions

github-actions Bot commented Oct 3, 2026

Copy link
Copy Markdown
Contributor

Delphi Coverage Report

File Stmts Miss Cover
init.py 2 0 100%
benchmarks/bench_pca.py 128 107 16%
benchmarks/bench_repness.py 81 65 20%
benchmarks/bench_update_votes.py 38 28 26%
benchmarks/benchmark_utils.py 34 0 100%
components/init.py 1 0 100%
components/config.py 165 70 58%
conversation/init.py 2 0 100%
conversation/conversation.py 1290 264 80%
conversation/group_vote_counts.py 38 2 95%
conversation/manager.py 131 42 68%
database/init.py 1 0 100%
database/dynamodb.py 395 178 55%
database/postgres.py 399 101 75%
empty_output.py 29 2 93%
engine_adapter.py 316 264 16%
pca_kmeans_rep/init.py 5 0 100%
pca_kmeans_rep/clusters.py 257 38 85%
pca_kmeans_rep/corr.py 98 17 83%
pca_kmeans_rep/group_k_smoother.py 28 0 100%
pca_kmeans_rep/legacy_kmeans.py 197 7 96%
pca_kmeans_rep/pca.py 162 19 88%
pca_kmeans_rep/repness.py 226 5 98%
poller/init.py 4 0 100%
poller/admission.py 318 15 95%
poller/backfill.py 936 32 97%
poller/capacity.py 444 17 96%
poller/capacity_manifest.py 260 13 95%
poller/coordinator_bridge.py 316 316 0%
poller/large_class.py 149 2 99%
poller/math_writer.py 113 3 97%
poller/promotion.py 127 2 98%
poller/readiness.py 368 12 97%
poller/service.py 641 39 94%
poller/worker_pool.py 140 0 100%
queue/executor.py 223 111 50%
regression/init.py 4 0 100%
regression/clojure_comparer.py 188 20 89%
regression/comparer.py 891 640 28%
regression/datasets.py 135 27 80%
regression/recorder.py 36 27 25%
regression/utils.py 130 85 35%
replay/init.py 2 0 100%
replay/attribution_capture.py 49 0 100%
replay/axis_continuity.py 347 24 93%
replay/certify.py 999 59 94%
replay/crosslang.py 128 2 98%
replay/decision_ties.py 55 8 85%
replay/diagnostics.py 50 7 86%
replay/driver.py 73 1 99%
replay/dsn_admission.py 36 1 97%
replay/equiv_query.py 10 0 100%
replay/event_ingress.py 95 1 99%
replay/fixture_bundle.py 815 62 92%
replay/fixture_config.py 266 11 96%
replay/fixture_extract.py 283 12 96%
replay/fixture_generate.py 103 1 99%
replay/fixture_samples.py 157 14 91%
replay/fixture_selection.py 119 0 100%
replay/fixture_survey.py 147 11 93%
replay/legacy_pca.py 29 17 41%
replay/polarity.py 319 25 92%
replay/poller_equiv.py 777 99 87%
replay/prodclone.py 193 2 99%
replay/real_data.py 146 10 93%
replay/schedule.py 205 10 95%
replay/served_math.py 159 4 97%
replay/shard_bench.py 189 65 66%
replay/stagecompare.py 890 161 82%
replay/stages.py 256 40 84%
replay/stepcompare.py 86 1 99%
replay/store.py 120 11 91%
replay/tie_capture.py 107 19 82%
replay/types.py 71 6 92%
run_math_pipeline.py 264 114 57%
types.py 22 0 100%
umap_narrative/501_calculate_comment_extremity.py 112 53 53%
umap_narrative/502_calculate_priorities.py 135 135 0%
umap_narrative/700_datamapplot_for_layer.py 502 502 0%
umap_narrative/701_static_datamapplot_for_layer.py 310 310 0%
umap_narrative/702_consensus_divisive_datamapplot.py 434 275 37%
umap_narrative/801_narrative_report_batch.py 790 726 8%
umap_narrative/803_check_batch_status.py 185 185 0%
umap_narrative/llm_factory_constructor/init.py 2 2 0%
umap_narrative/llm_factory_constructor/model_provider.py 234 234 0%
umap_narrative/polismath_commentgraph/init.py 1 0 100%
umap_narrative/polismath_commentgraph/cli.py 270 270 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 106 106 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 160 9 94%
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 3 0 100%
umap_narrative/polismath_commentgraph/utils/converter.py 283 237 16%
umap_narrative/polismath_commentgraph/utils/group_data.py 331 289 13%
umap_narrative/polismath_commentgraph/utils/storage.py 584 470 20%
umap_narrative/reset_conversation.py 159 50 69%
umap_narrative/run_pipeline.py 374 230 38%
umap_narrative/topic_naming.py 148 148 0%
utils/clj_hash.py 44 0 100%
utils/env_flags.py 13 0 100%
utils/general.py 63 39 38%
utils/offline_env.py 12 0 100%
utils/output_profile.py 66 2 97%
utils/serialization.py 10 5 50%
utils/vote_convention.py 29 2 93%
Total 22938 8210 64%

This branch has not been deployed

No deployments
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