Skype's queue engine returns — no C extension, no daemon, no provider approval required
PgQue v0.1.0 rebuilds PgQ in pure PL/pgSQL so the zero-bloat pattern works on RDS, Aurora, and everywhere else that blocks extensions.
PgQ handled messaging for hundreds of millions of Skype users, but its C extension and pgqd daemon made it DOA on managed Postgres. #6 brought the original PgQ v3.5.1 in as a reference, and @NikolayS rebuilt the engine in pure SQL from there. The trick that makes it bloat-free — TRUNCATE-based three-table rotation instead of per-row DELETE — survives intact. #64 matches the original 1-second tick cadence, #60 locks down uninstall() to superusers only, and #55 adds text overloads so you're not forced into jsonb. One SQL file, one transaction, zero restart.
rpg hits 222/232 psql compatibility — and now it runs in your browser
v0.11.0 ships WASM builds, eight new describe commands, and a regression suite that gates every merge.
PostgreSQL's own regression tests now run against rpg on every PR — unmodified .sql files, byte-identical output comparison. #818 marks the 0.11.0 release with 222 passing tests. The WASM build in #759 compiles to wasm32-unknown-unknown and runs in browsers via WebSocket proxy, which means your terminal client can now be a URL. #806 adds eight describe commands (partitioned relations, access methods, collations, text search objects), and #808 finally tracks standard_conforming_strings so backslashes in E'' strings parse correctly.
that time rpg logged your password in cleartext
A taint-chain security fix ensures credentials never reach the log path.
Connection strings contain passwords. Log statements accept strings. Without careful taint tracking, the two eventually meet. #817 breaks that chain by ensuring credential-bearing values can't flow into logging calls. The fix is small; the audit that found it probably wasn't.
when your sampler times out, now you'll know
pg_ash catches statement_timeout in take_sample() and increments a counter instead of silently skipping.
Active session sampling runs on a tight loop — if a sample query hits statement_timeout, the old behavior was to fail silently and move on. #29 catches that exception and increments a missed_samples counter, so you can tell the difference between 'nothing was happening' and 'the sampler couldn't keep up'. Meanwhile, #31 is pending with configurable N-partitions and rollup tables for v1.4.
EXPLAIN highlighting lands in the REPL — /plan finally prepends EXPLAIN
Plain EXPLAIN queries get syntax highlighting, and the /plan command stops expecting you to type EXPLAIN yourself.
The /plan command was supposed to show query plans, but it expected you to already have EXPLAIN in your query — defeating the point. #811 makes it prepend EXPLAIN automatically. #812 adds syntax highlighting for plain EXPLAIN output in the REPL, so you're not staring at monochrome plan nodes anymore.
pg_ash v1.4 brings configurable partitions and rollup tables
Pending work lets you tune retention granularity without touching the core sampling logic.
The current pg_ash partition scheme is fixed — fine for most workloads, awkward when you need longer retention at coarser granularity. #31 introduces configurable N-partitions and rollup tables, decoupling sample frequency from storage policy. Still in review, but the schema changes are substantial enough to watch.
pgque might get a new name before v0.2
An open PR clarifies the log-vs-task-queue distinction and considers renaming the whole project.
Is it a log? A task queue? Both? #76 proposes a rename and sharper positioning before the API solidifies. The PR also cleans up dead code via #75. If you're building against pgque, this is the moment to voice opinions.
17 commits, zero PRs — samospec is being written in the dark
The multi-AI spec tool saw heavy commit activity but nothing merged through the PR process.
samospec — the Git-native CLI for multi-AI spec creation — accumulated 17 commits this week with no corresponding pull requests. Either the feature is too early for review, or someone's speedrunning to a demo. Either way, the repo is active and the commit log suggests iterative refinement is happening, just not in public.
| postgres_dba | ★★★★★★★★★★ | 1,262 |
| pgque | ★★★★★★★★★☆ | 1,114 |
| rpg | ★★☆☆☆☆☆☆☆☆ | 230 |
| pg_ash | ★★☆☆☆☆☆☆☆☆ | 219 |
| gitzette | ★☆☆☆☆☆☆☆☆☆ | 6 |
| sqlever | ★☆☆☆☆☆☆☆☆☆ | 5 |
| leandex | ★☆☆☆☆☆☆☆☆☆ | 2 |
| samospec | ★☆☆☆☆☆☆☆☆☆ | 1 |
| NikolayS | ★☆☆☆☆☆☆☆☆☆ | 1 |
| pg_plumbing | ★☆☆☆☆☆☆☆☆☆ | 1 |
| gitzette-dispatch | ★☆☆☆☆☆☆☆☆☆ | 1 |
EXPLAIN highlighting lands in the REPL — /plan finally prepends EXPLAIN
Plain EXPLAIN queries get syntax highlighting, and the /plan command stops expecting you to type EXPLAIN yourself.
The /plan command was supposed to show query plans, but it expected you to already have EXPLAIN in your query — defeating the point. #811 makes it prepend EXPLAIN automatically. #812 adds syntax highlighting for plain EXPLAIN output in the REPL, so you're not staring at monochrome plan nodes anymore.
pg_ash v1.4 brings configurable partitions and rollup tables
Pending work lets you tune retention granularity without touching the core sampling logic.
The current pg_ash partition scheme is fixed — fine for most workloads, awkward when you need longer retention at coarser granularity. #31 introduces configurable N-partitions and rollup tables, decoupling sample frequency from storage policy. Still in review, but the schema changes are substantial enough to watch.
pgque might get a new name before v0.2
An open PR clarifies the log-vs-task-queue distinction and considers renaming the whole project.
Is it a log? A task queue? Both? #76 proposes a rename and sharper positioning before the API solidifies. The PR also cleans up dead code via #75. If you're building against pgque, this is the moment to voice opinions.
17 commits, zero PRs — samospec is being written in the dark
The multi-AI spec tool saw heavy commit activity but nothing merged through the PR process.
samospec — the Git-native CLI for multi-AI spec creation — accumulated 17 commits this week with no corresponding pull requests. Either the feature is too early for review, or someone's speedrunning to a demo. Either way, the repo is active and the commit log suggests iterative refinement is happening, just not in public.
| postgres_dba | ★★★★★★★★★★ | 1,262 |
| pgque | ★★★★★★★★★☆ | 1,114 |
| rpg | ★★☆☆☆☆☆☆☆☆ | 230 |
| pg_ash | ★★☆☆☆☆☆☆☆☆ | 219 |
| gitzette | ★☆☆☆☆☆☆☆☆☆ | 6 |
| sqlever | ★☆☆☆☆☆☆☆☆☆ | 5 |
| leandex | ★☆☆☆☆☆☆☆☆☆ | 2 |
| samospec | ★☆☆☆☆☆☆☆☆☆ | 1 |
| NikolayS | ★☆☆☆☆☆☆☆☆☆ | 1 |
| pg_plumbing | ★☆☆☆☆☆☆☆☆☆ | 1 |
| gitzette-dispatch | ★☆☆☆☆☆☆☆☆☆ | 1 |