GitHub engineering analytics for one organization.
DevPulse collects commits, pull requests and reviews from the GitHub GraphQL API, stores them in PostgreSQL, and renders contributor and repository statistics in a server-rendered Bootstrap 5 dashboard.
DevPulse does not automatically analyze every GitHub organization your Personal Access Token can reach. Only the organization configured in
GITHUB_ORGis queried and analyzed. If your token has access toorg-a,org-b, your personal account and a dozen open-source organizations, DevPulse still queries exactly the one you configured — it never enumeratesviewer.organizationsand never discovers organizations on its own.
- What the project does
- Architecture
- Requirements
- GitHub PAT configuration
- GitHub PAT permissions
- Organization configuration
- Docker installation
- Database setup
- Initial synchronization
- Incremental synchronization
- Date filters
- Contributor ranking
- Teams
- PDF reports
- Bot exclusions
- Archived repository handling
- GitHub GraphQL rate-limit handling
- Security considerations
- Troubleshooting
- Metric definitions
- Development
DevPulse answers questions like "who reviewed the most pull requests last quarter?" or "which repositories were active in the past 30 days?" without clicking through GitHub.
It provides:
- Overview dashboard — active contributors, commits, PRs opened, PRs merged, reviews, active repositories, plus time-series charts.
- Contributor ranking — every metric sortable independently.
- Contributor detail pages — summary metrics, charts, repositories, pull requests and reviews.
- Repository list and detail pages — contributors, commits, PRs, reviews, last activity, archived status. Every column header on the list sorts the table; clicking the active column flips the direction.
- Team filter — scope the overview, contributors and repositories pages to a GitHub team.
- Status page — sync health, GraphQL rate limit, stored row counts, run history.
- Global date filter — 7 or 30 days, 3/6/12 months or a custom range, applied consistently across every page.
- PDF reports — download the whole picture for any period as a printable document.
- Light and dark themes — pick Light, Dark or System from the header. The choice is remembered per browser, System follows the operating system live, and the theme is applied before the first paint so navigation never flashes.
DevPulse does not compute a single "performance score" and does not label anyone productive or unproductive. Commits, pull requests and reviews measure different kinds of contribution; they are presented and sorted separately so a human can interpret them in context.
GITHUB_TOKEN (environment variable, never persisted)
│
▼
GitHub GraphQL API rate-limit aware client:
│ cost / remaining / limit / resetAt logged per call
▼
Go collector / sync worker SyncOrganization → SyncRepositories → SyncCommits
│ → SyncPullRequests → SyncReviews → SyncContributors
│ → rebuild affected daily aggregates
▼
PostgreSQL raw entities + contributor_daily_stats read model
│
▼
Go HTTP application html/template handlers; reads PostgreSQL only
│
▼
Bootstrap 5 + Material UI server-rendered pages, Chart.js charts
cmd/server/ entrypoint: config, migrations, worker, HTTP server
internal/config/ environment parsing and startup validation
internal/github/ GraphQL transport, queries, pagination, rate limiting
internal/collector/ sync orchestration, incremental windows, error isolation
internal/database/ pgx pool, embedded migrations, upserts, analytics queries
internal/metrics/ date ranges, metric definitions, sort allow-list
internal/http/ router, handlers, template rendering
internal/models/ shared structs
migrations/ SQL migrations (embedded into the binary)
web/templates/ server-rendered pages
web/static/ CSS and Chart.js wiring
docs/ARCHITECTURE.md records the design decisions in more detail.
| Table | Purpose |
|---|---|
organizations |
The configured organization. last_synced_at drives the status page. |
repositories |
One row per repository, with commits_synced_at / prs_synced_at / reviews_synced_at watermarks for incremental sync. |
contributors |
GitHub accounts, with is_bot. |
commits |
Default-branch commits; unique on (repository_id, github_oid). |
pull_requests |
Unique on github_id and on (repository_id, number). |
pull_request_reviews |
Review submissions; unique on github_id. |
teams |
Organization teams, unique on github_id and on (organization_id, slug). |
team_members / team_repositories |
Team membership and repository access, replaced wholesale on each sync. |
contributor_daily_stats |
Daily read model keyed by (date, organization_id, contributor_id, repository_id). |
sync_runs |
One row per synchronization attempt. |
schema_migrations |
Applied migration versions. |
Every GitHub-derived table carries organization_id, and every dashboard query
filters on it. v1 runs one configured organization per deployment, but the
schema already supports several.
Indexes exist for organization_id, repository_id, contributor_id, the
relevant timestamps and pull_request_id.
- Docker and Docker Compose (the only requirement for running DevPulse).
- A GitHub Personal Access Token with read access to the organization.
- Go 1.25+ only if you want to build or test outside Docker.
- Create a token on GitHub (see permissions below).
- Copy the example environment file and fill it in:
cp .env.example .env
$EDITOR .envGITHUB_TOKEN=github_pat_xxxxxxxxxxxxxxxxxxxx
GITHUB_ORG=my-company
DATABASE_URL=postgres://devpulse:devpulse@postgres:5432/devpulse?sslmode=disable
SYNC_INTERVAL=30m
HISTORY_MONTHS=12
INCLUDE_ARCHIVED=false
EXCLUDED_USERS=dependabot[bot],renovate[bot].env is listed in .gitignore. Never commit it. The token is read from the
environment into memory and is never written to PostgreSQL, logs, HTML or API
responses.
| Variable | Required | Default | Description |
|---|---|---|---|
GITHUB_TOKEN |
yes | — | Read-only Personal Access Token. |
GITHUB_ORG |
yes | — | The single organization login to analyze. A comma-separated list is rejected. |
DATABASE_URL |
yes | — | PostgreSQL connection string. |
SYNC_INTERVAL |
no | 30m |
Worker period. Minimum 1m. |
HISTORY_MONTHS |
no | 12 |
How far back the first sync reaches (1–120). |
INCLUDE_ARCHIVED |
no | false |
Include archived repositories in analytics. |
SYNC_TEAMS |
no | true |
Import organization teams and membership. Needs read:org. |
EXCLUDED_USERS |
no | — | Comma-separated logins excluded from rankings. |
APP_PORT |
no | 8080 |
Host port published by Docker Compose. |
LISTEN_ADDR |
no | :8080 |
Address the server binds inside the container. |
LOG_LEVEL |
no | info |
debug, info, warn or error. |
SYNC_ON_STARTUP |
no | true |
Run one synchronization immediately at startup. |
MIN_RATE_LIMIT_REMAINING |
no | 200 |
Stop syncing below this GraphQL point budget. |
MAX_REPOSITORIES |
no | 0 |
Cap the repositories imported (0 = no limit). |
GITHUB_API_URL |
no | https://api.github.com/graphql |
Override for GitHub Enterprise Server. |
Configuration is validated at startup. Missing or malformed values produce a clear error naming the variable — and never its value:
devpulse: missing required configuration:
- GITHUB_TOKEN must be set (create a read-only Personal Access Token)
- GITHUB_ORG must be set to exactly one organization login
DevPulse only reads. It never requires, requests or performs a write.
- Resource owner: the organization in
GITHUB_ORG. - Repository access: All repositories (or the subset you want analyzed).
- Repository permissions (all Read-only):
Contents— commit history, additions/deletions.Pull requests— pull requests and their reviews.Metadata— mandatory, granted automatically.
- Organization permissions (Read-only):
Members— resolve contributor logins and display names, and read teams and their membership. Without it, setSYNC_TEAMS=false.
read:orgrepo— required only to read private repositories. For a public-only organization,public_repois enough. GitHub's classic tokens offer no read-only variant ofrepo; the fine-grained token above is the least-privilege option and is what we recommend.read:user
If your organization enforces SAML SSO, authorize the token for the
organization after creating it, otherwise GitHub answers 401/NOT_FOUND.
GITHUB_ORG is the sole scope of the deployment.
GITHUB_ORG=my-company- Every GraphQL query passes this login explicitly as a variable.
- The collector never calls
viewer.organizations,searchor any other discovery API. - Repositories whose owner does not match
GITHUB_ORGare skipped and logged, even if GitHub somehow returned them. - A comma-separated value is rejected at startup — one organization per deployment.
- To analyze a second organization, run a second DevPulse deployment with its own database.
git clone <this repository>
cd dev-pulse
cp .env.example .env
$EDITOR .env # set GITHUB_TOKEN and GITHUB_ORG
docker compose up -dThen open http://localhost:8080 (or APP_PORT).
Compose starts two services:
postgres— PostgreSQL 17 with a named volume and apg_isreadyhealth check.app— the DevPulse binary, started only once PostgreSQL reports healthy.
Useful commands:
docker compose logs -f app # follow synchronization progress
docker compose ps # service health
docker compose config # validate the compose file
docker compose down # stop (keeps data)
docker compose down -v # stop and delete the database volumeThe image is a multi-stage build: the Go toolchain stays in the build stage and the runtime image is Alpine plus the static binary (~25 MB), running as a non-root user.
Nothing to do by hand. On startup the application:
- Waits up to 60 s for PostgreSQL to accept connections.
- Takes a PostgreSQL advisory lock, so several replicas can start at once.
- Applies any migration in
migrations/*.sqlnot yet inschema_migrations, each inside its own transaction. - Starts the HTTP server only after migrations succeed, so requests are never served against an incompatible schema.
Migrations are embedded in the binary; the .sql files do not ship separately.
Health endpoints:
| Endpoint | Meaning |
|---|---|
GET /health |
Liveness plus database reachability. 503 when PostgreSQL is down. |
GET /ready |
Readiness: database reachable and schema applied. Reports schema_version. |
curl -s localhost:8080/health
{"status":"ok","database":"ok","organization":"my-company","version":"1.0.0","uptime_seconds":142,"last_sync_at":"2026-09-20T10:30:00Z","sync_status":"success"}Both Compose services declare health checks.
The first run starts automatically at startup (unless SYNC_ON_STARTUP=false)
and:
- Loads the organization named by
GITHUB_ORG. - Pages through its repositories (archived ones are stored too).
- Loads
HISTORY_MONTHS(default 12) of default-branch commits, pull requests and reviews per repository. - Imports the organization's teams, their members and their repositories.
- Persists everything in PostgreSQL with upserts.
- Builds the
contributor_daily_statsaggregates for the days it touched.
For a large organization the first run can take a long time and consume a lot of
GraphQL budget. Watch progress on /status or with docker compose logs -f app.
MAX_REPOSITORIES=5 is handy for a first trial run.
You can also trigger a run from the Sync now button on /status.
Later runs do not re-download the history.
Each repository stores three watermarks — commits_synced_at, prs_synced_at,
reviews_synced_at. On the next run:
- Commits are fetched with
history(since: <watermark − 15 min>). - Pull requests are read
orderBy: {field: UPDATED_AT, direction: DESC}and paging stops at the first pull request older than the watermark. - Reviews use the same early-stop rule, following nested review pagination where a pull request has more reviews than fit on one page.
The 15-minute overlap re-reads a small window so nothing is missed at the boundary.
Reruns are safe. Every write is INSERT … ON CONFLICT (<GitHub key>) DO UPDATE:
| Table | Idempotency key |
|---|---|
organizations |
github_id |
repositories |
github_id |
contributors |
github_id |
commits |
(repository_id, github_oid) |
pull_requests |
github_id |
pull_request_reviews |
github_id |
Running the same synchronization ten times produces exactly the same rows.
Aggregates are recomputed by deleting and re-inserting the affected
(repository, date) window inside one transaction, so they cannot drift or
double-count.
A synchronization failure never affects the web application: the worker runs in its own goroutine, recovers from panics, and the dashboard keeps serving whatever is already in PostgreSQL.
Every page takes the same filter and passes it along in links:
| Selection | URL |
|---|---|
| Last 7 days | /contributors?period=7d |
| Last 30 days | /contributors?period=30d |
| Last 3 months | /contributors?period=3m |
| Last 6 months | /contributors?period=6m |
| Last 12 months | /contributors?period=12m |
| Custom range | /contributors?period=custom&from=2026-01-01&to=2026-03-31 |
Ranges are inclusive and expressed in UTC calendar days. An invalid custom range does not error out: the page renders with the 30-day default and explains what was wrong.
Changing the period only queries PostgreSQL. No GitHub call is made when rendering statistics or changing filters — this is covered by an automated test that fails if a handler contacts GitHub.
Chart granularity follows the range length: ≤45 days daily, ≤200 days weekly, longer monthly.
/contributors ranks contributors by any single metric:
Top Contributors — Period: Last 3 months — Sort by: Reviews submitted
# Name Commits +/- PRs Reviews Active days
1 Erin Schmidt 111 +11.8k −4k 30 86 84
2 Carol Nwosu 112 +11.6k −3.8k 32 84 86
3 Bob Ferreira 110 +11.8k −3.8k 31 83 84
Sortable metrics:
- Code contribution — commits, additions, deletions, changed files
- Pull requests — PRs opened, PRs merged, PRs closed
- Review activity — reviews submitted, approvals, changes requested, review comments, unique PRs reviewed
- Activity — active days, repositories contributed to
Sorting is by ?sort=<key>. The key is resolved against a fixed allow-list, so a
hand-edited URL can never reach the SQL ORDER BY.
There is no combined score. Metrics stay separate by design.
Clicking a contributor opens /contributors/{login} with summary metrics,
time-series charts, the repositories they contributed to, their pull requests and
their reviews — all for the selected period.
DevPulse imports the organization's teams (what GitHub calls a group): their members and the repositories each team has access to. Membership is mirrored, not accumulated — when somebody leaves a team on GitHub they stop being a member here on the next synchronization, and a deleted team is removed.
A team filter appears on the overview, contributors and repositories pages:
/?period=3m&team=platform
/contributors?team=platform
/repositories?team=platform&sort=commits&dir=desc
The pages scope differently, because they answer different questions. The rule is the same everywhere: contributor-oriented figures follow the team's people, repository lists follow the team's repositories.
| Page | ?team=x means |
|---|---|
/ (overview) |
Headline figures, charts, the three top-contributor lists and the activity feed cover the members of team x. The repositories panel lists the repositories team x has access to. |
/contributors |
Only members of team x, counting their activity across the whole organization. |
/repositories |
Only repositories team x has access to, with figures covering every contributor, not only team members. |
Each page states which rule it is applying, so the numbers are never ambiguous. On the overview this matters for one card in particular: Active repositories counts the repositories the team's members worked in, which is not necessarily the same as the number of repositories the team owns. The filter is carried through sorting, pagination and date changes, and an unknown or deleted slug quietly falls back to the unfiltered view rather than breaking a shared link.
Teams need the read:org scope. If your token does not have it, team
synchronization fails while commits, pull requests and reviews are still
collected; the run is marked partial and the status page explains it. To turn
the feature off entirely:
SYNC_TEAMS=false- The PDF report is organization-wide; the team filter is not applied to it, and the report link deliberately does not carry a team rather than silently ignoring one.
- Only teams visible to the token are imported. Secret teams the token cannot read are invisible to DevPulse.
- A repository a team can reach that is outside
GITHUB_ORGis ignored, like every other foreign repository.
Every page carries a Report button in the header. It offers the period the page is currently showing, plus each preset:
- Last 7 days
- Last 30 days
- Last 3 months
- Last 6 months
- Last 12 months
For a custom range, set it in the date selector and then choose Current filter — the report follows whatever the page is showing.
The download is served by:
GET /report.pdf?period=3m
GET /report.pdf?period=custom&from=2026-01-01&to=2026-03-31
GET /report.pdf?period=12m&sort=reviews_submitted
sort accepts any contributor ranking metric and
orders the contributor table in the document; it defaults to commits. Opening
the link from /contributors keeps whatever sort is on screen.
The document is landscape A4 and contains:
- A header with the organization, the period (and its UTC bounds), when the report was generated, the sort key, the archived-repository policy, the excluded logins and the last successful synchronization.
- The six overview numbers, each compared with the immediately preceding period of equal length.
- A bar chart of commits, pull requests opened and reviews per bucket, at the same granularity the dashboard uses for that period.
- The contributor table with all fourteen metrics, up to 100 rows.
- The repository table, including archived and private repositories marked as such, up to 100 rows.
- The metric glossary, with the same wording as the UI.
Tables that hit the row cap say so on the page rather than cutting off silently.
Reports are built from the same PostgreSQL aggregates as the dashboard, honour
INCLUDE_ARCHIVED and EXCLUDED_USERS, and are scoped to GITHUB_ORG.
Generating a report never calls GitHub.
The file is named devpulse-<org>-<from>-to-<to>.pdf.
Before the first successful synchronization the report is still a valid document; it renders empty and says so, rather than returning an error.
The PDF uses the standard built-in fonts, which cover Latin-1. A display name in
another script (CJK, for example) or an emoji is replaced with ? in the
document; the ASCII @login shown next to it is always intact. Embedding a
Unicode font would fix this at the cost of a larger binary.
Two mechanisms, both applied at query time:
- Automatic detection. Accounts GitHub reports as
BotorMannequin, and logins ending in[bot](plusgithub-actions), are stored withis_bot = trueand never appear in rankings, totals or charts. - Manual exclusion.
EXCLUDED_USERS=dependabot[bot],renovate[bot]removes those logins (case-insensitive) from rankings and totals.
Because exclusion happens at query time, editing EXCLUDED_USERS and restarting
takes effect immediately — no re-synchronization needed. Excluded activity stays
in the database and can be restored by removing the login from the list.
Archived repositories are always stored, so nothing disappears from history.
With INCLUDE_ARCHIVED=false (the default):
- They are excluded from the dashboard overview, contributor rankings, repository rankings and charts.
- Their activity is not collected on subsequent runs (no point spending GraphQL budget on data that is filtered out).
- They still appear in
/repositories, dimmed and badged archived. - Their detail page still shows their own numbers, with a note explaining they are excluded from organization-wide rankings.
Set INCLUDE_ARCHIVED=true to fold them back into every statistic.
GitHub's GraphQL API uses a point budget (5,000 points/hour for most accounts).
Every DevPulse query requests the rateLimit block, and every call logs:
{"level":"DEBUG","msg":"github graphql call","operation":"commits",
"duration_ms":210,"cost":1,"remaining":4874,"limit":5000,
"reset_at":"2026-09-20T11:00:00Z"}Each synchronization run logs its total cost, and /status shows the live budget
with a progress bar and reset time.
Protections:
- Before each run and before each repository, the remaining budget is checked
against
MIN_RATE_LIMIT_REMAINING(default 200). Below the floor the run stops cleanly, logs the reason, and is recorded aspartial. It does not retry in a tight loop — it waits for the next scheduled tick. - Transient
5xx,429and secondary-rate-limit responses are retried with exponential backoff, at most four attempts. NOT_FOUNDis never retried.- Page sizes are tuned to stay inside GitHub's node limits, and incremental syncs avoid re-reading data that has not changed.
The token. GITHUB_TOKEN is read from the environment into memory. It is
never stored in PostgreSQL, never rendered in HTML, never included in API
responses or error pages, and never logged — the Authorization header is set on
the request and never echoed. Startup logs the configuration with the token
replaced by [redacted]. Automated tests assert that the token appears in no log
line and in no HTTP response.
Least privilege. DevPulse needs only read permissions and never writes to GitHub. See PAT permissions.
SQL. Every statement is parameterized. The only piece of SQL assembled at
runtime is the ranking ORDER BY, and it is chosen from a fixed table of known
column aliases; an unknown sort key falls back to the default.
Output escaping. Pages are rendered with html/template, which escapes
contextually — repository names, logins, pull request titles and descriptions
from GitHub are treated as untrusted data, never as safe HTML. Chart payloads are
JSON with <, > and & escaped so a crafted string cannot break out of the
<script> element.
HTTP headers. A Content-Security-Policy, X-Content-Type-Options: nosniff,
X-Frame-Options: DENY and Referrer-Policy: no-referrer are set on every
response.
Errors. Database connection errors are stripped of embedded credentials before they are logged. Internal errors render a generic page; details go to the application log.
Note on access control. DevPulse has no authentication of its own. Anyone who can reach the port sees the dashboard. Do not publish it to the internet without putting an authenticating proxy in front of it.
GITHUB_TOKEN must be set at startup
Configuration validation failed. The message lists every missing or malformed
variable. Check that .env exists and that docker compose picked it up
(docker compose config).
github rejected the token (status 401)
The token is invalid, expired, or not authorized for the organization. If your
organization enforces SAML SSO, open the token settings and click Authorize for
that organization.
github resource not found ... organization "x"
GITHUB_ORG is misspelled, or the token cannot see the organization. The value
must be the organization login (as in github.com/<login>), not its display
name.
The dashboard says "No data for … yet"
The first synchronization has not finished. Check /status and
docker compose logs -f app. For a large organization the initial import takes a
while.
Some commits have no contributor
A commit whose author e-mail is not linked to a GitHub account (deleted user,
unconfigured local git e-mail) cannot be attributed. DevPulse stores it with the
raw author name but leaves contributor_id empty, so it counts in repository
totals but in nobody's ranking. Ask contributors to add their commit e-mail to
their GitHub account.
Changed files are zero for some commits
GitHub does not report changedFilesIfAvailable for every commit (very large
commits in particular). Unavailable values count as zero.
A repository failed but others succeeded
That is intended. Repository errors are logged with the repository name, the run
continues, and its status becomes partial. The errors are visible on /status.
Team synchronization fails but everything else works
The token lacks the read:org scope, or the organization restricts team
visibility. The run is marked partial and contribution data is unaffected.
Grant the scope, or set SYNC_TEAMS=false to stop trying.
synchronization postponed until the GitHub rate limit recovers
The point budget fell below MIN_RATE_LIMIT_REMAINING. Wait for resetAt (shown
on /status), lengthen SYNC_INTERVAL, or reduce HISTORY_MONTHS.
Database not reachable
/health returns 503 with "database":"unavailable". Check
docker compose ps and the postgres logs. The app retries for 60 s at startup.
Starting over
docker compose down -v deletes the database volume; the next start
re-synchronizes from scratch.
Also available in the UI under "What each metric means".
| Metric | Definition |
|---|---|
| Commits | Commits on the default branch attributed to the contributor's GitHub account during the period. |
| Additions | Lines added, as reported by GitHub for those commits. |
| Deletions | Lines deleted, as reported by GitHub for those commits. |
| Changed files | Sum of files touched by those commits. Unavailable values count as zero. |
| PRs opened | Pull requests authored by the contributor and created during the period. |
| PRs merged | Pull requests authored by the contributor that were merged during the period. |
| PRs closed | Pull requests authored by the contributor that were closed without being merged. |
| Reviews submitted | Review submissions made during the period, including approvals, change requests and comment-only reviews. Pending (unsubmitted) reviews are not counted. |
| Approvals | Review submissions whose state was APPROVED. |
| Changes requested | Review submissions whose state was CHANGES_REQUESTED. |
| Review comments | Inline comments attached to those review submissions. |
| Unique PRs reviewed | Distinct pull requests reviewed at least once. Two reviews on one pull request count once. |
| Active days | Distinct UTC calendar days with any tracked activity (commit, pull request or review). |
| Repositories | Distinct repositories of the configured organization the contributor was active in. |
Attribution rules: commit metrics belong to the commit author; PR metrics to the pull request author; review metrics to the reviewer. Days are UTC.
The repository is plain Go with no code generation.
gofmt -l .
go vet ./...
go test ./...go test ./... passes without a database: the PostgreSQL integration tests skip
themselves unless TEST_DATABASE_URL is set. To run everything:
docker run -d --name devpulse-test-pg \
-e POSTGRES_USER=devpulse -e POSTGRES_PASSWORD=devpulse -e POSTGRES_DB=devpulse_test \
-p 55432:5432 postgres:17-alpine
TEST_DATABASE_URL='postgres://devpulse:devpulse@127.0.0.1:55432/devpulse_test?sslmode=disable' \
go test ./...Test coverage includes date-range calculation, organization filtering and isolation, contributor aggregation, ranking and sorting, bot exclusion, archived repository exclusion, duplicate-synchronization prevention, upsert idempotency, configuration validation, GraphQL pagination (including nested review pages), team import (membership replacement, deleted teams, nested pagination) and team filtering including cross-organization isolation, rate-limit handling, template escaping of GitHub-provided strings, PDF report generation for every period (including table widths that must fit the page), and an end-to-end check that neither rendering a page nor generating a report ever calls GitHub. Collector tests run against recorded GraphQL responses — no network access required.
.github/workflows/ci.yml runs on every pull request against main, and can be
started by hand from the Actions tab. It needs no secrets, so pull requests from
forks work.
Go tests — gofmt check, go mod tidy check (fails if go.mod/go.sum
would change), go build, go vet, then go test -race against a PostgreSQL
service container, so the integration tests actually run instead of skipping.
The coverage profile is uploaded as a build artifact.
Docker image and stack — validates docker compose config, builds the image,
starts the full stack with a placeholder token and checks that /health and
/ready answer, that /ready reports an applied schema version, that every page
and a PDF report return 200, and that the token never appears in the container
logs. The placeholder token makes synchronization fail on purpose: the web
application must keep serving regardless. Container logs are dumped if anything
fails.
Building the image:
docker build -t devpulse:local --build-arg VERSION=1.0.0 .Drop a new migrations/NNNN_name.sql file in place. It is embedded at build time
and applied automatically, in lexical order, on the next start.