Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@CHANGELOG.md`:
- Line 37: The changelog entry in CHANGELOG.md has a spelling error in the
commit description. Update the text in that entry to change “privilages” to
“privileges” while keeping the rest of the commit message and link intact.
- Line 30: Update the changelog entry text in CHANGELOG.md to fix the spelling
error in the commit description by changing “grammer” to “grammar”; adjust the
specific changelog item that mentions the quick start section so the user-facing
release note is spelled correctly.
- Line 24: Fix the spelling mistake in the changelog entry from “reflaction” to
“reflection” in the affected commit description. Update the user-facing text in
CHANGELOG.md so the summary line reads correctly and remains consistent with the
existing entry format.

In `@docs/requirements.md`:
- Around line 25-30: Fix the small wording typos in the storage requirements
paragraph by editing the prose in the relevant documentation section: change
“you might needs SSD” to “you might need an SSD,” and correct “arround” to
“around.” Keep the rest of the paragraph intact and make the wording read
naturally in the surrounding requirements text.

In `@docs/setup.md`:
- Around line 76-80: The privilege description in the setup docs is incomplete
and the `keymgr` role is introduced without explanation. Update the prose near
the privilege list to complete the writer privileges description in the setup
documentation, and add a short sentence explaining what `keymgr` grants and when
it should be used. Refer to the existing privilege wording around the
reader/writer/keymgr section so the documentation stays consistent.
- Line 24: Remove the duplicated word in the documentation sentence so the
phrase reads naturally; update the text in the setup guide where it mentions the
GitHub project packages, and ensure the wording around the linked packages URL
is corrected without changing the meaning.

In `@indexer/cli/setup.go`:
- Around line 83-85: The help text for the database creation command is using a
raw string literal with escaped newline text, so Cobra will display the
characters literally instead of a line break. Update the help text in the setup
command definition to use an actual newline inside the backtick string for the
Long description, keeping the command metadata in the same place so the help
output renders correctly.
- Line 70: The create-user command still exposes a stale --user flag even though
user_name is now passed as a positional argument, so remove the unused flag from
the setup in createUserCmd to keep the help text and behavior aligned. Update
the command wiring in setup so only the positional user_name path is supported,
and verify any related parsing or validation around createUserCmd no longer
references the old flag.

In `@indexer/data_processor/operator.go`:
- Around line 443-448: Message conversion failures are currently being dropped
because nil entries from ConvertToDbMessages are skipped when building
aggregatedDbMessages, allowing ProcessMessages to succeed with missing rows.
Update the message conversion flow in ProcessMessages and the affected batch
handling around msgResults so ConvertToDbMessages errors are collected in a
separate error channel/slice and returned immediately instead of continuing to
Phase 2. Make sure the aggregation step that calls aggregatedDbMessages.Merge
only runs when all conversions succeeded, and abort before any insert if any
conversion error was encountered.

In `@indexer/main_operator/operator.go`:
- Around line 107-108: `HistoricProcess` is not observing the shutdown context
even though `signalHandler.RegisterOperation()` is used before it starts. Update
`orch.HistoricProcess` (and any chunk-processing path it calls) to accept
`signalHandler.Context()` and check `ctx.Done()` inside the historic chunk loop
so a SIGTERM can stop processing early. Use the existing
`signalHandler.Context()` and `HistoricProcess` symbols to wire the context
through the historic range processing flow.

In `@pkgs/database/timescaledb/insert.go`:
- Around line 43-47: InsertRows should validate that every element in the rows
batch targets the same table and column layout before building the COPY call.
Add a homogeneity check in InsertRows around the first-row metadata usage so
rows[0].TableName() and rows[0].TableColumns() are only used after confirming
all Insertable items match; otherwise reject the batch early with a clear error.
Keep the fix localized to InsertRows and the CopyFromSlice/CopyFrom path so
mixed batches cannot be copied with the wrong table metadata.

In `@pkgs/database/timescaledb/pool.go`:
- Around line 72-91: The buildDSN logic in pool.go is concatenating libpq
keyword/value pairs directly, so fields like password, SslRootCert, SslCert, and
SslKey can break parsing when they contain spaces, quotes, or backslashes.
Update buildDSN to either construct a URL-style DSN or quote/escape each value
before appending it to the string, keeping the existing config fields and
pgxpool.ParseConfig flow intact.

In `@pkgs/database/timescaledb/query_block.go`:
- Around line 313-318: The error handling in the block-time query is too broad:
the row.Scan path in the query block collapses every database failure into the
same “not enough block data” message. Update the logic around row.Scan so only
pgx.ErrNoRows maps to the friendly missing-data error, and any other scan/query
error is wrapped and returned with its original details. Keep the heightDiff ==
0 check as the data-shortage case, and ensure the query block uses errors and
github.com/jackc/pgx/v5 to distinguish these paths cleanly.

---

Outside diff comments:
In `@indexer/data_processor/operator.go`:
- Around line 153-167: ProcessBlocks is inserting every entry from blocksData
even when processBlock exits early on decode/parse errors, leaving zero-value
rows behind. Update ProcessBlocks to track only successfully processed block
indexes, similar to the transaction/validator paths, and build the insert
payload from that valid set before calling dbPool.InsertRows. Make the change in
the code path that fills blocksData and the corresponding insertion logic so
failed blocks are skipped entirely.

In `@indexer/db_init/tables.go`:
- Around line 579-588: The CREATE USER statement in tables.go is built by
concatenating userName and the prompted password directly into SQL, so update
the user creation flow to use PostgreSQL-safe quoting or parameterization
instead of fmt.Sprintf. Locate the SQL assembly around the password prompt in
the user creation path and ensure both the role name and password are safely
escaped before db.pool.Exec is called.

---

Nitpick comments:
In `@cliff.toml`:
- Line 19: The generated header text in cliff.toml includes an unnecessary
literal newline escape after the Semantic Versioning sentence, which can
introduce an extra blank line in the output. Update the header string so the
Semantic Versioning line in the TOML block ends naturally without the appended
newline escape, keeping the template aligned with the surrounding header text.

In `@docs/requirements.md`:
- Line 64: The Go version noted in the docs is out of sync with the toolchain
version used in CI, which can confuse readers about what is actually validated.
Update the version referenced in the requirements documentation to match the
pinned Go version used by the govulncheck workflow, and keep the `go.mod`-based
minimum separate if needed. Use the existing Go version reference in
`requirements.md` and the CI workflow’s version pin as the source of truth when
aligning the wording.

In `@indexer/cli/setup.go`:
- Around line 241-245: The repeated "1 month" aggregate interval in the setup
list should be extracted into a local constant to keep the entries aligned and
satisfy SonarCloud. Update the slice in setup.go where the metric definitions
are assembled in the setup logic to use a shared constant instead of repeating
the literal, and apply that constant consistently to all remaining aggregate
entries in the same list.

In `@indexer/decoder/registry.go`:
- Around line 162-164: The `strings.Join(m.Args, ",")` assignment in
`registry.go` makes `Args` ambiguous when individual arguments can contain
commas. Update the `Args` serialization in the registry path that builds the
record from `m.PkgPath`, `m.Func`, and `m.Args` to use an unambiguous encoding
such as a JSON array, or otherwise escape separators so the original arguments
can be reconstructed exactly. If the downstream column is intentionally a legacy
comma-separated field, keep it unchanged only if that semantics is required.

In `@indexer/orchestrator/operator.go`:
- Line 270: Fix the typo in the doc comment for collectTransactionsFromBlocks by
changing “smae” to “same”; update the comment text only and leave the function
logic unchanged.

In `@README.md`:
- Line 63: The README setup command still uses the deprecated docker-compose v1
syntax. Update the documented command to the modern docker compose form so new
users can start the timescaledb service without setup issues. Refer to the
README setup step containing docker-compose up -d timescaledb and replace it
with the current Docker Compose plugin command.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@api/handlers/transactions.go`:
- Around line 67-77: The dispatch map in GetTransactionMessage is rebuilt on
every request, causing avoidable allocation overhead. Move the map creation out
of the hot path by storing it on TransactionsHandler or initializing it once
with sync.Once, and keep the receiver-bound handlers like getBankSendResponse
and getMsgRunResponse wired through that cached map so the per-request call only
performs lookup.