Coolify passes predefined vars to Docker Compose via --env-file for
interpolation only; a var reaches the container solely if the compose file
references it. Removing the reference left SOURCE_COMMIT unset, so deploynotify
reported "unknown". Re-add SOURCE_COMMIT: ${SOURCE_COMMIT:-} per Coolify docs.
Match the store-scraper-bot behaviour: DM the owner once per boot regardless of
commit, and report an unknown SHA rather than staying silent. Drop the store-
based dedup and the empty-SHA skip; resolveCommitSHA defaults to "unknown".
Declaring SOURCE_COMMIT: ${SOURCE_COMMIT:-} in compose resolved to an empty
string at parse time and overrode the value Coolify auto-injects into the
container at runtime, so gitSHA stayed empty and the owner DM was skipped.
Remove the declaration; the bot already reads SOURCE_COMMIT from the container
env at startup.
Baking SOURCE_COMMIT as a Docker build arg never worked: Coolify exposes it
as a runtime env var, not a build arg, so gitSHA was always empty and the owner
DM was skipped. Read SOURCE_COMMIT from the container env at startup (falling
back to the ldflags-baked gitSHA for local builds), forward it via the compose
environment, and drop the dead build-arg baking.
Emit one structured line per authorized command: input text, sender id +
username, chat type/id/title (DM vs group), and outcome (INFO ok / ERROR with
err). Replaces the prior error-only log; denied commands stay silent.
The Bot API GET reliably clears the webhook, so the retry loop added for the
empty-body POST is no longer needed. One best-effort call; log on failure.
go-telegram's DeleteWebhook posts an empty multipart form (DropPendingUpdates
is omitempty); some networks answer that bodyless POST with an empty response,
failing to parse and leaving the webhook active so getUpdates 409s forever.
Replace it with a plain GET /deleteWebhook, which sends no body and parses the
JSON reliably. Keeps pending updates for a lossless cutover.
Telegram's edge can return an empty body on the first request after a cold
container start, failing the single best-effort DeleteWebhook with 'unexpected
end of JSON input' and leaving a stale webhook that 409s getUpdates. Retry up
to 5 times with a 1s backoff so long polling is reliably unblocked.
Move DeleteWebhook ahead of the deploynotify owner DM and polling so the
webhook is cleared first thing once the bot is ready, instead of just before
getUpdates. DropPendingUpdates=false keeps the lossless cutover.
The in-process scheduler is the sole cron trigger on self-host, so the
/cron/{name} HTTP endpoint and its CRON_SHARED_SECRET are dead surface.
Reduce the HTTP server to GET / (health) and remove the cron secret config,
SSM binding, and now-unused timeout.
Delete the byte-oriented KVStore/VersionedStore abstraction and the
DynamoDB/memory KV backends. Add a generic typed store (DocStore[T] with
Provider/Collection/Typed) persisting each value as a flattened native
Mongo document (storedDoc[T] via bson inline) — no value envelope.
- MongoDB is the only runtime backend; memory kept for tests/local.
- All modules + deploynotify use typed stores; persisted structs carry
bson tags == json names (incl. nested lolschedule/wordle types).
- lolschedule wraps its array/scalar values in named structs.
- migrate-dynamo-to-mongo writes the flattened shape via Typed[bson.M]
with wrap rules; Scan/--dry-run/--verify retained.
Verified: go vet/build clean; full go test green hermetically and
in-container vs real Mongo 7 + DynamoDB Local (storage integration +
migrator e2e).
Delete Firestore provider implementation and tests (5 files). Extract shared
key helpers to new internal/storage/keys.go module. Remove Firestore case
from server bootstrap and storage factory. Drop cloud.google.com/go/firestore
and related dependencies from go.mod. Remove firestore-emulator test targets
from Makefile. Update README storage section to reflect MongoDB-only backend.
Replace content-addressed storage versioning (CAS) with explicit version fields
on documents. Versions are managed via optimistic locking pattern: Get returns
version, Put increments atomically and fails if version mismatch detected.
Callers (portfolio, lolschedule) updated to use GetVersioned/PutVersioned,
simplifying concurrency handling and eliminating CAS overhead.
Persist the KV value field as a BSON string so it is directly readable in
the Atlas/Compass UI, matching DynamoDB's String storage. Reads still accept
BSON binary as a backward-compat fallback. Non-UTF-8 callers must encode
upstream, same constraint as DynamoDB; all current callers write JSON.
Mark phases 1–5 (MongoDB, cron, polling, migration, decommission) as completed.
Update phase files with finalized implementation notes and test coverage summaries.
Add standalone migrator tool with scan-and-transform logic for moving persisted
state from DynamoDB to MongoDB. Includes tests and README with usage instructions.
Wire MongoDB client and cron scheduler into server initialization. Rename
TELEGRAM_OWNER to TELEGRAM_ADMIN_ID and SLACK_OWNER to SLACK_ADMIN_ID
for clarity. Enable long-polling as default with cron scheduler for self-hosted.
Add 5-phase plan to self-host on Coolify (docker-compose) with a MongoDB
Atlas backend, an in-process cron scheduler, DynamoDB->Atlas migration,
and full AWS teardown. Red-teamed and validated; decommission scope
verified against live AWS/Cloudflare accounts. Include free-tier audit
and S3-elimination research reports. Ignore wrangler local cache.
Parallelizing the per-symbol fetches opened N simultaneous TLS handshakes into
an empty connection pool. On the memory-constrained Lambda (256MB ~0.15 vCPU)
those CPU-bound handshakes thrashed and each exceeded the per-fetch timeout, so
every ticker rendered "(no price)". Sequential fetches reuse the price client's
keep-alive connection (one handshake), which is why the code was sequential by
design.
- revert stock and coin stats loops to sequential (keep the reply-budget
sub-context and 3s per-fetch timeout)
- log per-symbol fetch failures instead of silently swallowing them
Stats handlers reused the single bounded update context for both upstream
price fetches and the final Telegram reply, while per-upstream HTTP timeouts
equalled the whole-handler budget. One slow upstream drained the deadline and
the reply failed with "context deadline exceeded" (observed on /stock_stats).
- add chathelper.FetchContext: fetches run under a child context that reserves
a tail of the deadline for the reply, which is sent on the original context
- cap kbs/coin/gold/vnappmob HTTP timeouts at 3s so one hung upstream cannot
consume the reply budget
- fetch held stock/coin prices concurrently (bounded) so total latency tracks
the slowest single fetch instead of their sum; per-symbol failures degrade to
a "no price" line and the summary still sends
- Change /coin_sell semantics from quantity to USD amount
- Add price-validity and dust-quantity guards to buy/sell
- Improve usage text and error messages
- Add tests for invalid amount, dust, zero price, and insufficient holdings
Migration complete and verified in production; the trading->stock
DynamoDB partition copy is no longer needed and the old trading rows
have been deleted.
tiennm99