add llm adapter interface and implementations

internal/llm/adapter_groq.go

@@ -0,0 +1,73 @@
+package llm
+
+import (
+	"context"
+	"fmt"
+	"log"
+	"time"
+
+	"github.com/sashabaranov/go-openai"
+)
+
+// GroqAdapter implements Adapter using Groq's OpenAI-compatible API
+type GroqAdapter struct {
+	client *openai.Client
+	config Config
+}
+
+// NewGroqAdapter creates a new Groq LLM adapter
+func NewGroqAdapter(cfg Config) *GroqAdapter {
+	clientConfig := openai.DefaultConfig(cfg.APIKey)
+	clientConfig.BaseURL = "https://api.groq.com/openai/v1"
+	return &GroqAdapter{
+		client: openai.NewClientWithConfig(clientConfig),
+		config: cfg,
+	}
+}
+
+func (a *GroqAdapter) Process(ctx context.Context, text string) (string, error) {
+	if text == "" {
+		return "", nil
+	}
+
+	opts := PostProcessingOptions{
+		RemoveStutters:    a.config.RemoveStutters,
+		AddPunctuation:    a.config.AddPunctuation,
+		FixGrammar:        a.config.FixGrammar,
+		RemoveFillerWords: a.config.RemoveFillerWords,
+	}
+
+	systemPrompt := BuildSystemPrompt(opts, a.config.Keywords)
+	userPrompt := BuildUserPrompt(text, a.config.CustomPrompt)
+
+	model := a.config.Model
+	if model == "" {
+		model = "llama-3.3-70b-versatile"
+	}
+
+	req := openai.ChatCompletionRequest{
+		Model: model,
+		Messages: []openai.ChatCompletionMessage{
+			{Role: openai.ChatMessageRoleSystem, Content: systemPrompt},
+			{Role: openai.ChatMessageRoleUser, Content: userPrompt},
+		},
+		Temperature: 0.3, // Low temperature for consistent cleanup
+	}
+
+	start := time.Now()
+	resp, err := a.client.CreateChatCompletion(ctx, req)
+	duration := time.Since(start)
+
+	if err != nil {
+		log.Printf("groq-llm-adapter: API call failed after %v: %v", duration, err)
+		return "", fmt.Errorf("groq chat completion: %w", err)
+	}
+
+	if len(resp.Choices) == 0 {
+		return "", fmt.Errorf("groq chat completion: no response choices")
+	}
+
+	result := resp.Choices[0].Message.Content
+	log.Printf("groq-llm-adapter: processed in %v: %q -> %q", duration, text, result)
+	return result, nil
+}

internal/llm/adapter_openai.go

@@ -0,0 +1,71 @@
+package llm
+
+import (
+	"context"
+	"fmt"
+	"log"
+	"time"
+
+	"github.com/sashabaranov/go-openai"
+)
+
+// OpenAIAdapter implements Adapter using OpenAI's chat completions API
+type OpenAIAdapter struct {
+	client *openai.Client
+	config Config
+}
+
+// NewOpenAIAdapter creates a new OpenAI LLM adapter
+func NewOpenAIAdapter(cfg Config) *OpenAIAdapter {
+	return &OpenAIAdapter{
+		client: openai.NewClient(cfg.APIKey),
+		config: cfg,
+	}
+}
+
+func (a *OpenAIAdapter) Process(ctx context.Context, text string) (string, error) {
+	if text == "" {
+		return "", nil
+	}
+
+	opts := PostProcessingOptions{
+		RemoveStutters:    a.config.RemoveStutters,
+		AddPunctuation:    a.config.AddPunctuation,
+		FixGrammar:        a.config.FixGrammar,
+		RemoveFillerWords: a.config.RemoveFillerWords,
+	}
+
+	systemPrompt := BuildSystemPrompt(opts, a.config.Keywords)
+	userPrompt := BuildUserPrompt(text, a.config.CustomPrompt)
+
+	model := a.config.Model
+	if model == "" {
+		model = "gpt-4o-mini"
+	}
+
+	req := openai.ChatCompletionRequest{
+		Model: model,
+		Messages: []openai.ChatCompletionMessage{
+			{Role: openai.ChatMessageRoleSystem, Content: systemPrompt},
+			{Role: openai.ChatMessageRoleUser, Content: userPrompt},
+		},
+		Temperature: 0.3, // Low temperature for consistent cleanup
+	}
+
+	start := time.Now()
+	resp, err := a.client.CreateChatCompletion(ctx, req)
+	duration := time.Since(start)
+
+	if err != nil {
+		log.Printf("openai-llm-adapter: API call failed after %v: %v", duration, err)
+		return "", fmt.Errorf("openai chat completion: %w", err)
+	}
+
+	if len(resp.Choices) == 0 {
+		return "", fmt.Errorf("openai chat completion: no response choices")
+	}
+
+	result := resp.Choices[0].Message.Content
+	log.Printf("openai-llm-adapter: processed in %v: %q -> %q", duration, text, result)
+	return result, nil
+}

internal/llm/llm.go

@@ -0,0 +1,42 @@
+package llm
+
+import (
+	"context"
+	"fmt"
+)
+
+// Adapter interface for LLM text processing
+type Adapter interface {
+	Process(ctx context.Context, text string) (string, error)
+}
+
+// Config holds LLM adapter configuration
+type Config struct {
+	Provider          string
+	APIKey            string
+	Model             string
+	RemoveStutters    bool
+	AddPunctuation    bool
+	FixGrammar        bool
+	RemoveFillerWords bool
+	CustomPrompt      string
+	Keywords          []string
+}
+
+// NewAdapter creates an LLM adapter based on the provider
+func NewAdapter(cfg Config) (Adapter, error) {
+	switch cfg.Provider {
+	case "openai":
+		if cfg.APIKey == "" {
+			return nil, fmt.Errorf("OpenAI API key required")
+		}
+		return NewOpenAIAdapter(cfg), nil
+	case "groq":
+		if cfg.APIKey == "" {
+			return nil, fmt.Errorf("Groq API key required")
+		}
+		return NewGroqAdapter(cfg), nil
+	default:
+		return nil, fmt.Errorf("unsupported LLM provider: %s", cfg.Provider)
+	}
+}

internal/llm/llm_test.go

@@ -0,0 +1,155 @@
+package llm
+
+import (
+	"strings"
+	"testing"
+)
+
+func TestBuildSystemPrompt(t *testing.T) {
+	tests := []struct {
+		name     string
+		opts     PostProcessingOptions
+		keywords []string
+		contains []string
+	}{
+		{
+			name: "all options enabled",
+			opts: PostProcessingOptions{
+				RemoveStutters:    true,
+				AddPunctuation:    true,
+				FixGrammar:        true,
+				RemoveFillerWords: true,
+			},
+			keywords: nil,
+			contains: []string{
+				"Remove stutters",
+				"Add proper punctuation",
+				"Fix grammar",
+				"Remove filler words",
+			},
+		},
+		{
+			name: "only grammar",
+			opts: PostProcessingOptions{
+				FixGrammar: true,
+			},
+			keywords: nil,
+			contains: []string{
+				"Fix grammar",
+			},
+		},
+		{
+			name: "with keywords",
+			opts: PostProcessingOptions{
+				RemoveStutters: true,
+			},
+			keywords: []string{"Kubernetes", "TypeScript", "hyprvoice"},
+			contains: []string{
+				"Kubernetes",
+				"TypeScript",
+				"hyprvoice",
+				"Context keywords",
+			},
+		},
+		{
+			name:     "no options - should have default",
+			opts:     PostProcessingOptions{},
+			keywords: nil,
+			contains: []string{
+				"Clean up the text",
+			},
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			result := BuildSystemPrompt(tc.opts, tc.keywords)
+			for _, expected := range tc.contains {
+				if !strings.Contains(result, expected) {
+					t.Errorf("expected prompt to contain %q, got: %s", expected, result)
+				}
+			}
+		})
+	}
+}
+
+func TestBuildUserPrompt(t *testing.T) {
+	tests := []struct {
+		name         string
+		text         string
+		customPrompt string
+		expected     string
+	}{
+		{
+			name:         "no custom prompt",
+			text:         "hello world",
+			customPrompt: "",
+			expected:     "hello world",
+		},
+		{
+			name:         "with custom prompt",
+			text:         "hello world",
+			customPrompt: "Format as a haiku",
+			expected:     "Format as a haiku\n\nText to process:\nhello world",
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			result := BuildUserPrompt(tc.text, tc.customPrompt)
+			if result != tc.expected {
+				t.Errorf("expected %q, got %q", tc.expected, result)
+			}
+		})
+	}
+}
+
+func TestNewAdapter(t *testing.T) {
+	// Test OpenAI adapter creation
+	openaiCfg := Config{
+		Provider: "openai",
+		APIKey:   "sk-test-key",
+		Model:    "gpt-4o-mini",
+	}
+	adapter, err := NewAdapter(openaiCfg)
+	if err != nil {
+		t.Fatalf("failed to create openai adapter: %v", err)
+	}
+	if _, ok := adapter.(*OpenAIAdapter); !ok {
+		t.Error("expected OpenAIAdapter type")
+	}
+
+	// Test Groq adapter creation
+	groqCfg := Config{
+		Provider: "groq",
+		APIKey:   "gsk_test-key",
+		Model:    "llama-3.3-70b-versatile",
+	}
+	adapter, err = NewAdapter(groqCfg)
+	if err != nil {
+		t.Fatalf("failed to create groq adapter: %v", err)
+	}
+	if _, ok := adapter.(*GroqAdapter); !ok {
+		t.Error("expected GroqAdapter type")
+	}
+
+	// Test missing API key
+	noKeyCfg := Config{
+		Provider: "openai",
+		APIKey:   "",
+	}
+	_, err = NewAdapter(noKeyCfg)
+	if err == nil {
+		t.Error("expected error for missing API key")
+	}
+
+	// Test unsupported provider
+	badCfg := Config{
+		Provider: "unsupported",
+		APIKey:   "key",
+	}
+	_, err = NewAdapter(badCfg)
+	if err == nil {
+		t.Error("expected error for unsupported provider")
+	}
+}

internal/llm/prompt.go

@@ -0,0 +1,65 @@
+package llm
+
+import (
+	"fmt"
+	"strings"
+)
+
+// PostProcessingOptions controls which cleanup operations to request
+type PostProcessingOptions struct {
+	RemoveStutters    bool
+	AddPunctuation    bool
+	FixGrammar        bool
+	RemoveFillerWords bool
+}
+
+// BuildSystemPrompt generates the system prompt for text cleanup
+func BuildSystemPrompt(opts PostProcessingOptions, keywords []string) string {
+	var tasks []string
+
+	if opts.RemoveStutters {
+		tasks = append(tasks, "Remove stutters and repeated words/phrases")
+	}
+	if opts.AddPunctuation {
+		tasks = append(tasks, "Add proper punctuation")
+	}
+	if opts.FixGrammar {
+		tasks = append(tasks, "Fix grammar errors")
+	}
+	if opts.RemoveFillerWords {
+		tasks = append(tasks, "Remove filler words (um, uh, like, you know, etc.)")
+	}
+
+	// If no tasks, just clean up generally
+	if len(tasks) == 0 {
+		tasks = append(tasks, "Clean up the text while preserving meaning")
+	}
+
+	prompt := "You are a text cleanup assistant. Your job is to clean up speech-to-text transcriptions.\n\n"
+	prompt += "Tasks:\n"
+	for _, task := range tasks {
+		prompt += fmt.Sprintf("- %s\n", task)
+	}
+
+	prompt += "\nRules:\n"
+	prompt += "- Preserve the original meaning and intent\n"
+	prompt += "- Keep the same language as the input\n"
+	prompt += "- Do not add any new information\n"
+	prompt += "- Do not remove meaningful content\n"
+	prompt += "- Output ONLY the cleaned text, nothing else\n"
+	prompt += "- If the input is empty or nonsensical, return it as-is\n"
+
+	if len(keywords) > 0 {
+		prompt += fmt.Sprintf("\nContext keywords (use correct spelling for these terms): %s\n", strings.Join(keywords, ", "))
+	}
+
+	return prompt
+}
+
+// BuildUserPrompt generates the user prompt with the text to process
+func BuildUserPrompt(text string, customPrompt string) string {
+	if customPrompt != "" {
+		return fmt.Sprintf("%s\n\nText to process:\n%s", customPrompt, text)
+	}
+	return text
+}

progress.txt

@@ -47,3 +47,20 @@ Key decisions:
 - LLM provider names are "openai" and "groq" (not "groq-transcription")
 - PostProcessing defaults to all true only if ALL options are false (zero values)
 - Keywords at root level (global), used by both transcription and LLM
+
+## Task 3: Create LLM adapter interface and implementations - COMPLETE
+
+Created internal/llm package with:
+- `Adapter` interface: `Process(ctx, text) (string, error)`
+- `Config` struct mirroring config.LLMAdapterConfig
+- `prompt.go` with `BuildSystemPrompt(opts, keywords)` and `BuildUserPrompt(text, customPrompt)`
+- `OpenAIAdapter` using go-openai chat completions API
+- `GroqAdapter` using Groq's OpenAI-compatible API (baseURL override)
+- `NewAdapter(config)` factory function
+
+Key decisions:
+- Low temperature (0.3) for consistent text cleanup
+- Default models: gpt-4o-mini (OpenAI), llama-3.3-70b-versatile (Groq)
+- System prompt builds dynamically based on enabled options
+- Keywords included in system prompt for correct spelling hints
+- Custom prompt prepended to user prompt if enabled

tasks/prd.jsonc

@@ -66,7 +66,7 @@
         "Factory returns correct adapter",
         "Typecheck passes"
       ],
-      "passes": false
+      "passes": true
     },
     {
       "title": "Integrate LLM phase into pipeline",