Mathis/llamactl
1
0
Fork 0
mirror of https://github.com/lordmathis/llamactl.git synced 2025-11-07 01:24:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

Refactor command parsing and building

Browse Source
This commit is contained in:
Matúš Námešný
2025-09-19 20:23:25 +02:00
parent 34a949d22e
commit 64842e74b0
14 changed files with 434 additions and 539 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
pkg/backends/vllm/parser.go
View File

@@ -1,34 +0,0 @@
package vllm
import (
"llamactl/pkg/backends"
)
// ParseVllmCommand parses a vLLM serve command string into VllmServerOptions
// Supports multiple formats:
// 1. Full command: "vllm serve --model MODEL_NAME --other-args"
// 2. Full path: "/usr/local/bin/vllm serve --model MODEL_NAME"
// 3. Serve only: "serve --model MODEL_NAME --other-args"
// 4. Args only: "--model MODEL_NAME --other-args"
// 5. Multiline commands with backslashes
func ParseVllmCommand(command string) (*VllmServerOptions, error) {
executableNames := []string{"vllm"}
subcommandNames := []string{"serve"}
multiValuedFlags := map[string]bool{
"middleware": true,
"api_key": true,
"allowed_origins": true,
"allowed_methods": true,
"allowed_headers": true,
"lora_modules": true,
"prompt_adapters": true,
}
var vllmOptions VllmServerOptions
if err := backends.ParseCommand(command, executableNames, subcommandNames, multiValuedFlags, &vllmOptions); err != nil {
return nil, err
}
return &vllmOptions, nil
}

84
pkg/backends/vllm/parser_test.go
View File

@@ -1,84 +0,0 @@
package vllm_test
import (
"llamactl/pkg/backends/vllm"
"testing"
)
func TestParseVllmCommand(t *testing.T) {
tests := []struct {
name string
command string
expectErr bool
}{
{
name: "basic vllm serve command",
command: "vllm serve --model microsoft/DialoGPT-medium",
expectErr: false,
},
{
name: "serve only command",
command: "serve --model microsoft/DialoGPT-medium",
expectErr: false,
},
{
name: "args only",
command: "--model microsoft/DialoGPT-medium --tensor-parallel-size 2",
expectErr: false,
},
{
name: "empty command",
command: "",
expectErr: true,
},
{
name: "unterminated quote",
command: `vllm serve --model "unterminated`,
expectErr: true,
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
result, err := vllm.ParseVllmCommand(tt.command)
if tt.expectErr {
if err == nil {
t.Errorf("expected error but got none")
}
return
}
if err != nil {
t.Errorf("unexpected error: %v", err)
return
}
if result == nil {
t.Errorf("expected result but got nil")
}
})
}
}
func TestParseVllmCommandValues(t *testing.T) {
command := "vllm serve --model test-model --tensor-parallel-size 4 --gpu-memory-utilization 0.8 --enable-log-outputs"
result, err := vllm.ParseVllmCommand(command)
if err != nil {
t.Fatalf("unexpected error: %v", err)
}
if result.Model != "test-model" {
t.Errorf("expected model 'test-model', got '%s'", result.Model)
}
if result.TensorParallelSize != 4 {
t.Errorf("expected tensor_parallel_size 4, got %d", result.TensorParallelSize)
}
if result.GPUMemoryUtilization != 0.8 {
t.Errorf("expected gpu_memory_utilization 0.8, got %f", result.GPUMemoryUtilization)
}
if !result.EnableLogOutputs {
t.Errorf("expected enable_log_outputs true, got %v", result.EnableLogOutputs)
}
}

28
pkg/backends/vllm/vllm.go
View File

@@ -142,3 +142,31 @@ func (o *VllmServerOptions) BuildCommandArgs() []string {
}
return backends.BuildCommandArgs(o, multipleFlags)
}
// ParseVllmCommand parses a vLLM serve command string into VllmServerOptions
// Supports multiple formats:
// 1. Full command: "vllm serve --model MODEL_NAME --other-args"
// 2. Full path: "/usr/local/bin/vllm serve --model MODEL_NAME"
// 3. Serve only: "serve --model MODEL_NAME --other-args"
// 4. Args only: "--model MODEL_NAME --other-args"
// 5. Multiline commands with backslashes
func ParseVllmCommand(command string) (*VllmServerOptions, error) {
executableNames := []string{"vllm"}
subcommandNames := []string{"serve"}
multiValuedFlags := map[string]bool{
"middleware": true,
"api_key": true,
"allowed_origins": true,
"allowed_methods": true,
"allowed_headers": true,
"lora_modules": true,
"prompt_adapters": true,
}
var vllmOptions VllmServerOptions
if err := backends.ParseCommand(command, executableNames, subcommandNames, multiValuedFlags, &vllmOptions); err != nil {
return nil, err
}
return &vllmOptions, nil
}

78
pkg/backends/vllm/vllm_test.go
View File

@@ -7,6 +7,84 @@ import (
"testing"
)
func TestParseVllmCommand(t *testing.T) {
tests := []struct {
name string
command string
expectErr bool
}{
{
name: "basic vllm serve command",
command: "vllm serve --model microsoft/DialoGPT-medium",
expectErr: false,
},
{
name: "serve only command",
command: "serve --model microsoft/DialoGPT-medium",
expectErr: false,
},
{
name: "args only",
command: "--model microsoft/DialoGPT-medium --tensor-parallel-size 2",
expectErr: false,
},
{
name: "empty command",
command: "",
expectErr: true,
},
{
name: "unterminated quote",
command: `vllm serve --model "unterminated`,
expectErr: true,
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
result, err := vllm.ParseVllmCommand(tt.command)
if tt.expectErr {
if err == nil {
t.Errorf("expected error but got none")
}
return
}
if err != nil {
t.Errorf("unexpected error: %v", err)
return
}
if result == nil {
t.Errorf("expected result but got nil")
}
})
}
}
func TestParseVllmCommandValues(t *testing.T) {
command := "vllm serve --model test-model --tensor-parallel-size 4 --gpu-memory-utilization 0.8 --enable-log-outputs"
result, err := vllm.ParseVllmCommand(command)
if err != nil {
t.Fatalf("unexpected error: %v", err)
}
if result.Model != "test-model" {
t.Errorf("expected model 'test-model', got '%s'", result.Model)
}
if result.TensorParallelSize != 4 {
t.Errorf("expected tensor_parallel_size 4, got %d", result.TensorParallelSize)
}
if result.GPUMemoryUtilization != 0.8 {
t.Errorf("expected gpu_memory_utilization 0.8, got %f", result.GPUMemoryUtilization)
}
if !result.EnableLogOutputs {
t.Errorf("expected enable_log_outputs true, got %v", result.EnableLogOutputs)
}
}
func TestBuildCommandArgs(t *testing.T) {
options := vllm.VllmServerOptions{
Model: "microsoft/DialoGPT-medium",