[Ideas] Wanna PostgresML in CBDB? An AI application database. LLM embeddings, generate text, or make online predictions using only SQL.

[avamingli]

Description

https://github.com/postgresml/postgresml

PostgresML is opened source since April 2022.

PostgresML is an AI application database. Download open source models from Huggingface, or train your own, to create and index LLM embeddings, generate text, or make online predictions using only SQL.

Use case/motivation

No response

Related issues

No response

Are you willing to submit a PR?

• Yes I am willing to submit a PR!

[edespino]

If PostgresML were to come to Cloudberry DB, I think this would be a huge win for Cloudberry. I have heard from a few folks requesting support for it. Any chance this will be picked up soon? I can assist if needed. It is relatively easy to follow the Building from Source instructions with PostgreML and PostgreSQL. The examples are amazing what AI is bringing to the tech space.

When attempting to build against Cloudberry DB (HEAD), I received the following compilation issues:

   Compiling pgrx v0.11.3
error[E0425]: cannot find function `heap_copytuple` in module `pg_sys`
     --> /home/eespino/.cargo/registry/src/index.crates.io-6f17d22bba15001f/pgrx-0.11.3/src/heap_tuple.rs:182:37
      |
182   |         let copy = unsafe { pg_sys::heap_copytuple(self.tuple.into_pg()) };
      |                                     ^^^^^^^^^^^^^^ help: a function with a similar name exists: `SPI_copytuple`
      |
     ::: /home/eespino/postgresml/pgml-extension/target/release/build/pgrx-pg-sys-4198585f221daa95/out/pg14.rs:45629:1
      |
45629 | #[pgrx_macros::pg_guard]
      | ------------------------ similarly named function `SPI_copytuple` defined here

error[E0425]: cannot find function `heap_form_tuple` in module `pg_sys`
     --> /home/eespino/.cargo/registry/src/index.crates.io-6f17d22bba15001f/pgrx-0.11.3/src/heap_tuple.rs:237:29
      |
237   |                     pg_sys::heap_form_tuple(tuple_desc.as_ptr(), datums, is_null.as_mut_ptr());
      |                             ^^^^^^^^^^^^^^^ help: a function with a similar name exists: `heap_deform_tuple`
      |
     ::: /home/eespino/postgresml/pgml-extension/target/release/build/pgrx-pg-sys-4198585f221daa95/out/pg14.rs:45629:1
      |
45629 | #[pgrx_macros::pg_guard]
      | ------------------------ similarly named function `heap_deform_tuple` defined here

error[E0425]: cannot find function `heap_form_tuple` in module `pg_sys`
     --> /home/eespino/.cargo/registry/src/index.crates.io-6f17d22bba15001f/pgrx-0.11.3/src/heap_tuple.rs:281:25
      |
281   |                 pg_sys::heap_form_tuple(tupdesc.as_ptr(), datums.as_mut_ptr(), nulls.as_mut_ptr());
      |                         ^^^^^^^^^^^^^^^ help: a function with a similar name exists: `heap_deform_tuple`
      |
     ::: /home/eespino/postgresml/pgml-extension/target/release/build/pgrx-pg-sys-4198585f221daa95/out/pg14.rs:45629:1
      |
45629 | #[pgrx_macros::pg_guard]
      | ------------------------ similarly named function `heap_deform_tuple` defined here

error[E0425]: cannot find function `heap_form_tuple` in module `pg_sys`
     --> /home/eespino/.cargo/registry/src/index.crates.io-6f17d22bba15001f/pgrx-0.11.3/src/htup.rs:210:33
      |
210   |                         pg_sys::heap_form_tuple(tupdesc, datums.as_mut_ptr(), nulls.as_mut_ptr())
      |                                 ^^^^^^^^^^^^^^^ help: a function with a similar name exists: `heap_deform_tuple`
      |
     ::: /home/eespino/postgresml/pgml-extension/target/release/build/pgrx-pg-sys-4198585f221daa95/out/pg14.rs:45629:1
      |
45629 | #[pgrx_macros::pg_guard]
      | ------------------------ similarly named function `heap_deform_tuple` defined here

error[E0425]: cannot find function, tuple struct or tuple variant `MemoryContextDelete` in module `pg_sys`
     --> /home/eespino/.cargo/registry/src/index.crates.io-6f17d22bba15001f/pgrx-0.11.3/src/memcxt.rs:189:21
      |
189   |             pg_sys::MemoryContextDelete(self.owned);
      |                     ^^^^^^^^^^^^^^^^^^^ help: a function with a similar name exists: `MemoryContextReset`
      |
     ::: /home/eespino/postgresml/pgml-extension/target/release/build/pgrx-pg-sys-4198585f221daa95/out/pg14.rs:45629:1
      |
45629 | #[pgrx_macros::pg_guard]
      | ------------------------ similarly named function `MemoryContextReset` defined here

error[E0425]: cannot find function, tuple struct or tuple variant `MemoryContextDelete` in module `pg_sys`
     --> /home/eespino/.cargo/registry/src/index.crates.io-6f17d22bba15001f/pgrx-0.11.3/src/memcxt.rs:400:29
      |
400   |                     pg_sys::MemoryContextDelete(context);
      |                             ^^^^^^^^^^^^^^^^^^^ help: a function with a similar name exists: `MemoryContextReset`
      |
     ::: /home/eespino/postgresml/pgml-extension/target/release/build/pgrx-pg-sys-4198585f221daa95/out/pg14.rs:45629:1
      |
45629 | #[pgrx_macros::pg_guard]
      | ------------------------ similarly named function `MemoryContextReset` defined here

error[E0063]: missing field `bgw_start_rule` in initializer of `pgrx_pg_sys::BackgroundWorker`
   --> /home/eespino/.cargo/registry/src/index.crates.io-6f17d22bba15001f/pgrx-0.11.3/src/bgworkers.rs:627:19
    |
627 |         let bgw = pg_sys::BackgroundWorker {
    |                   ^^^^^^^^^^^^^^^^^^^^^^^^ missing `bgw_start_rule`

error[E0061]: this function takes 2 arguments but 0 arguments were supplied
     --> /home/eespino/.cargo/registry/src/index.crates.io-6f17d22bba15001f/pgrx-0.11.3/src/bgworkers.rs:656:9
      |
656   |         pg_sys::check_for_interrupts!();
      |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ two arguments of type `*const i8` and `i32` are missing
      |
note: function defined here
     --> /home/eespino/postgresml/pgml-extension/target/release/build/pgrx-pg-sys-4198585f221daa95/out/pg14.rs:46064:12
      |
46064 |     pub fn ProcessInterrupts(filename: *const ::std::os::raw::c_char, lineno: ::std::os::raw::c_int);
      |            ^^^^^^^^^^^^^^^^^
      = note: this error originates in the macro `pg_sys::check_for_interrupts` (in Nightly builds, run with -Z macro-backtrace for more info)
help: provide the arguments
     -->   /home/eespino/.cargo/registry/src/index.crates.io-6f17d22bba15001f/pgrx-pg-sys-0.11.3/src/submodules/elog.rs:464:42
      |
464   |                 $crate::ProcessInterrupts(/* *const i8 */, /* i32 */);
      |                                          ~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Some errors have detailed explanations: E0061, E0063, E0425.
For more information about an error, try `rustc --explain E0061`.
error: could not compile `pgrx` (lib) due to 8 previous errors
warning: build failed, waiting for other jobs to finish...
rustc-LLVM ERROR: out of memory
Allocation failed
error: could not compile `pgrx-pg-sys` (lib)

Caused by:
  process didn't exit successfully: `/home/eespino/.rustup/toolchains/stable-x86_64-unknown-linux-gnu/bin/rustc --crate-name pgrx_pg_sys --edition=2021 /home/eespino/.cargo/registry/src/index.crates.io-6f17d22bba15001f/pgrx-pg-sys-0.11.3/src/lib.rs --error-format=json --json=diagnostic-rendered-ansi,artifacts,future-incompat --diagnostic-width=183 --crate-type lib --emit=dep-info,metadata,link -C opt-level=3 -C codegen-units=1 --cfg 'feature="cshim"' --cfg 'feature="default"' --cfg 'feature="pg14"' -C metadata=b01a8b1088281dab -C extra-filename=-b01a8b1088281dab --out-dir /home/eespino/postgresml/pgml-extension/target/release/deps -C strip=debuginfo -L dependency=/home/eespino/postgresml/pgml-extension/target/release/deps --extern libc=/home/eespino/postgresml/pgml-extension/target/release/deps/liblibc-20172ff1c7b647c3.rmeta --extern memoffset=/home/eespino/postgresml/pgml-extension/target/release/deps/libmemoffset-9b203808f2d517d1.rmeta --extern pgrx_macros=/home/eespino/postgresml/pgml-extension/target/release/deps/libpgrx_macros-8412f7fb42f8f4bc.so --extern pgrx_sql_entity_graph=/home/eespino/postgresml/pgml-extension/target/release/deps/libpgrx_sql_entity_graph-0848c1a129d6d5ef.rmeta --extern serde=/home/eespino/postgresml/pgml-extension/target/release/deps/libserde-d6a815f9ac706e34.rmeta --extern sptr=/home/eespino/postgresml/pgml-extension/target/release/deps/libsptr-cdc0376797587724.rmeta --cap-lints allow -C link-args=-Wl,-undefined,dynamic_lookup,-fuse-ld=lld -L /home/eespino/postgresml/pgml-extension/target/release/build/pgrx-pg-sys-4198585f221daa95/out/cshim -l static=pgrx-cshim-14` (signal: 6, SIGABRT: process abort signal)

[avamingli]

Hi, @edespino Thanks for your interest and help.

When attempting to build against Cloudberry DB (HEAD), I received the following compilation issues:

I don't know about rust, so I can't tell about that.
It will be cool if someone have interest about this topic and make it work well with CBDB and verify cases and etc.

[edespino]

I believe Greenplum supports PostgresML. It is an important product differentiator. Maybe this can get on the Cloudberry 2024 Roadmap?

[jianlirong]

Sure. We will do some investigation first, and then update the status of this work.

[my-ship-it]

As @jianlirong mentioned:

Furthermore, a concrete example of PostgresML in action would be invaluable in guiding our research efforts.

Our initial plan is to maintain a consistent user interface between the frontend and PostgresML, but when it comes to the backend execution of AI/ML algorithms, we'll leverage a more modern, distributed computing framework specifically designed for AI/ML workloads, such as Apache Ray

[avamingli]

We have conclusions as answer, close discussion.