Dump transactions for a program

Find the transactions that call your program, at top level or through a recorded inner call (CPI).

Need	Tool
Print new transactions for a program	Live stream example with a program address
Save historical matches in SQLite	`blockzilla-dump program` (this page)
Process matches in your own Rust code	Local filters

Print live program transactions

Private beta. You need a consumer key. See Live stream.

From the repository root:

BLOCKZILLA_STREAM_MODE=sse node hivezilla/stream-relay/client/example.mjs \
  https://stream.blockzilla.dev ~/.config/blockzilla/stream.key ./pumpfun.cursor \
  6EF8rrecthR5Dkzon8Nwu78hRvfCKubJ14M5uBEwF6P > pumpfun-transactions.jsonl

Replace the last address with your program. Transaction lines have slot, bankId, commitment, blockhash, signature, success, topLevelPrograms, and innerPrograms. Failed transactions are included. The output also contains small root and invalidate records. A processed bank can be invalidated.

This is a console example. If it stops, some printed lines can repeat on the next run. For a crash-safe application, use the SDK as shown in Live stream.

Save historical matches in SQLite

blockzilla-dump is a separate program. It reads archive files from a local folder and writes the matching transactions to SQLite. It can resume after a stop.

Requirement: a local copy of a Compact V2 epoch, with all files from the same build. It cannot read the public sample URL directly.

Install it from the repository root:

cargo install --locked --path indexer/blockzilla-dump --bin blockzilla-dump --debug

Run it:

blockzilla-dump program \
  --archive /data/blockzilla/compact-v2/900 \
  --source-generation-prefix my-copy-v1 \
  --message-schema current \
  --metadata-schema current-typed-error \
  --epoch 900 \
  --program 6EF8rrecthR5Dkzon8Nwu78hRvfCKubJ14M5uBEwF6P \
  --threads 4 \
  --output pumpfun-900.sqlite

--source-generation-prefix is your name for this copy of the files. Keep it the same when you resume. Use a new name and a new output file when the files change.
--message-schema current and --metadata-schema current-typed-error are correct for the published sample epochs. For other files, use the schema that was recorded when they were built.
--archive can also be a folder with epoch-N subfolders. Repeat --epoch to scan several epochs.
The command scans the full epoch. It has no --max-blocks option.

To resume after a stop, run the same command again.

Query the result

Check progress:

blockzilla-dump status --output pumpfun-900.sqlite --json

List the newest matches:

SELECT t.epoch, t.slot, t.tx_index, t.primary_signature_base58,
       p.direct_count, p.cpi_count
FROM transactions AS t
JOIN program_matches AS p USING (epoch, slot, tx_index)
ORDER BY t.epoch DESC, t.slot DESC, t.tx_index DESC
LIMIT 20;

direct_count counts top-level calls. cpi_count counts recorded inner calls.
A transaction that only lists the program as an account does not match.
Failed transactions are included. This schema has no success column: the execution result is inside the compact metadata blob.
The database also stores all signatures, the transaction accounts, and the compact message and metadata bytes. These use Blockzilla’s compact schema, not Solana wire format or RPC JSON.

Exit status

Exit	Meaning
`0`	The scan finished and every match was decided.
`1`	An error stopped the scan. Run the same command to resume.
`2`	Only with `--on-indeterminate record`: the scan finished, but some transactions could not be decided. They are listed in the `coverage_issues` table.

By default (--on-indeterminate fail) the scan stops when it cannot decide a match. It never drops a possible match silently.

Reference: blockzilla-dump README, selector rules, and database schema.