Fixes the bug 1196

[diego-minguzzi]

Description

This PR fixes the bug 1196. Previously Llama Cpp failed because it did not find gRPC libraries and the protoc compiler.

Now it is possible to build Local AI along with gRPC using the command:

BUILD_GRPC=ON make build

It was modified the Makefile in the root directory. In case the variable BUILD_GRPC is defined, it runs the script at the path:

backend/cpp/grpc/script/build_grpc.sh

Which clones the gRPC repo and installs it at the path backend/cpp/grpc/installed_packages

The command:
make clean
removes gRPC and the installed files.

The command:
make rebuild

does not affects the installed gRPC.
Makefile was modified at the lines 405-424.

Moreover, under Linux Ubuntu, it gRPC requires the packages:
build-essential autoconf libtool pkg-config

[mudler]

Hi @diego-minguzzi thanks for having a look at this!

[mudler]

I'm wondering - here probably the best approach would be to use cmake directly.

I was reading initially https://github.com/grpc/grpc/tree/master/src/cpp#fetchcontent but leaving this for a follow-up, maybe you'd like to take a look first?

[diego-minguzzi]

Yes, I think that at least the bash-based solution allows people to build without docker, when the PR is approved.
Then, I can try to modify the gRPC installation using the CMake.
Does it sound good?

[mudler]

Sounds good!

[diego-minguzzi]

@lunamidori5 : could you please review the PR #1232 ? Thanks

[lunamidori5]

@mudler is already checking it out

[mudler]

As a first shot at it looks good! Thanks @diego-minguzzi

[lunamidori5]

@mudler is it going to auto merge?

[diego-minguzzi]

@mudler , @lunamidori5 : I have found that my PR fails to build because of the following error:

38 3370.4 In file included from /build/backend/cpp/llama/llama.cpp/build/examples/grpc-server/backend.grpc.pb.cc:5:
#38 3370.4 /build/backend/cpp/llama/llama.cpp/build/examples/grpc-server/backend.pb.h:19:2: error: #error "This file was generated by an older version of protoc which is"
#38 3370.4    19 | #error "This file was generated by an older version of protoc which is"
#38 3370.4       |  ^~~~~
#38 3370.4 /build/backend/cpp/llama/llama.cpp/build/examples/grpc-server/backend.pb.h:20:2: error: #error "incompatible with your Protocol Buffer headers. Please"
#38 3370.4    20 | #error "incompatible with your Protocol Buffer headers. Please"

I have found that the Dockerfile uses gRPC 1.58.0, as I do.

In the file LocalAI/pkg/grpc/proto/backend_grpc.pb.go there is: Requires gRPC-Go v1.32.0 or later.

Maybe gRPC must be downgraded , e.g. to 1.32?

Moreover, I think it is not the content of my PR that causes the build to fail, since the docker files do not use the build option that runs the gRPC build script, it is something else.
Let me know your thoughts on this.

[lunamidori5]

@diego-minguzzi so due to something odd with github, every time I add @mudler for review it adds me too (I do not know why) just act as if I am not here <3

[mudler]

@mudler , @lunamidori5 : I have found that my PR fails to build because of the following error:

38 3370.4 In file included from /build/backend/cpp/llama/llama.cpp/build/examples/grpc-server/backend.grpc.pb.cc:5:
#38 3370.4 /build/backend/cpp/llama/llama.cpp/build/examples/grpc-server/backend.pb.h:19:2: error: #error "This file was generated by an older version of protoc which is"
#38 3370.4    19 | #error "This file was generated by an older version of protoc which is"
#38 3370.4       |  ^~~~~
#38 3370.4 /build/backend/cpp/llama/llama.cpp/build/examples/grpc-server/backend.pb.h:20:2: error: #error "incompatible with your Protocol Buffer headers. Please"
#38 3370.4    20 | #error "incompatible with your Protocol Buffer headers. Please"

I have found that the Dockerfile uses gRPC 1.58.0, as I do.

In the file LocalAI/pkg/grpc/proto/backend_grpc.pb.go there is: Requires gRPC-Go v1.32.0 or later.

Maybe gRPC must be downgraded , e.g. to 1.32?

Moreover, I think it is not the content of my PR that causes the build to fail, since the docker files do not use the build option that runs the gRPC build script, it is something else. Let me know your thoughts on this.

In the logs I clearly see the script kicking in:

#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/impl/server_callback_handlers.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/impl/server_initializer.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/impl/service_type.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/impl/status.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/impl/sync.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/resource_quota.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/security/audit_logging.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/security/auth_context.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/security/auth_metadata_processor.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/security/authorization_policy_provider.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/security/credentials.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/security/server_credentials.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/security/tls_certificate_provider.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/security/tls_certificate_verifier.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/security/tls_credentials_options.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/server.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/server_builder.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/server_context.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/server_interface.h
#38 3244.7 -- Up-to-date: /build/backend/cpp/grpc/installed_packages/include/grpcpp/server_posix.h

See the logs here: https://github.com/mudler/LocalAI/actions/runs/6731857467/job/18297419421 - it looks like that BUILD_GRPC is always on @diego-minguzzi can you make it look for an explicit ON? better make it disabled by default

[diego-minguzzi]

@mudler : yes, the BUILD_GRPC was on because the Dockerfile uses the same variable for the same purpose.
Now I changed the variable that causes the script to run into BUILD_GRPC_FOR_BACKEND_LLAMA.
The command is thus:
BUILD_GRPC_FOR_BACKEND_LLAMA=ON make build

Should work now, I have no way to use it with Docker.
Can I trigger the rerun of the pipelines? I am not very familiar with the github site, I suppose not.

[diego-minguzzi]

Merged the master and the remote into my branch.
In my branch the go.mod was aligned with what requested by @Aisuko .

[mudler]

Looking good here, thanks!

changes addressed