refactor: restructure project layout — main.go to root, llm+config to pkg/

- Move cmd/agentu/main.go → main.go (project root) - Move internal/llm/ → pkg/llm/ (leaf package, no internal deps) - Move internal/config/ → pkg/config/ (leaf package, no internal deps) - Update all import paths: agentu/internal/llm → agentu/pkg/llm (14 files) - Update all import paths: agentu/internal/config → agentu/pkg/config (3 files) - Update README: build/run commands, CLI flags section - agent/session/tools/tui stay in internal/ (app-specific business logic) Build: go build . | Test: go test ./... | Vet: go vet ./...
2026-06-24 07:06:26 -07:00
parent 950c63adaa
commit f235b8ce90
22 changed files with 21 additions and 21 deletions
@@ -0,0 +1,202 @@
+package llm
+
+import (
+	"bufio"
+	"bytes"
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+	"net/http"
+	"strings"
+	"time"
+)
+
+const maxChatAttempts = 3
+
+const retryDelay = 100 * time.Millisecond
+
+type OpenAICompatibleClient struct {
+	baseURL    string
+	apiKey     string
+	httpClient *http.Client
+}
+
+func NewOpenAICompatibleClient(baseURL, apiKey string, httpClient *http.Client) *OpenAICompatibleClient {
+	if httpClient == nil {
+		httpClient = http.DefaultClient
+	}
+	return &OpenAICompatibleClient{
+		baseURL:    strings.TrimRight(baseURL, "/"),
+		apiKey:     apiKey,
+		httpClient: httpClient,
+	}
+}
+
+func (c *OpenAICompatibleClient) ChatStream(ctx context.Context, req ChatRequest, emit func(StreamEvent) error) error {
+	req.Stream = true
+	extra := make(map[string]any, len(req.Extra)+1)
+	for key, value := range req.Extra {
+		extra[key] = value
+	}
+	if _, exists := extra["stream_options"]; !exists {
+		extra["stream_options"] = map[string]any{"include_usage": true}
+	}
+	req.Extra = extra
+	body, err := json.Marshal(req)
+	if err != nil {
+		return fmt.Errorf("marshal chat request: %w", err)
+	}
+
+	var lastErr error
+	for attempt := 1; attempt <= maxChatAttempts; attempt++ {
+		httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, c.baseURL+"/v1/chat/completions", bytes.NewReader(body))
+		if err != nil {
+			return fmt.Errorf("create chat request: %w", err)
+		}
+		httpReq.Header.Set("Content-Type", "application/json")
+		httpReq.Header.Set("Accept", "text/event-stream")
+		if c.apiKey != "" {
+			httpReq.Header.Set("Authorization", "Bearer "+c.apiKey)
+		}
+
+		resp, err := c.httpClient.Do(httpReq)
+		if err != nil {
+			lastErr = fmt.Errorf("send chat request: %w", err)
+			if !shouldRetryRequest(ctx, attempt, 0) {
+				return lastErr
+			}
+			if waitErr := waitBeforeRetry(ctx); waitErr != nil {
+				return waitErr
+			}
+			continue
+		}
+
+		if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+			limit := io.LimitReader(resp.Body, 4096)
+			data, _ := io.ReadAll(limit)
+			_ = resp.Body.Close()
+			lastErr = fmt.Errorf("chat request failed: status=%d body=%s", resp.StatusCode, strings.TrimSpace(string(data)))
+			if !shouldRetryRequest(ctx, attempt, resp.StatusCode) {
+				return lastErr
+			}
+			if waitErr := waitBeforeRetry(ctx); waitErr != nil {
+				return waitErr
+			}
+			continue
+		}
+
+		defer resp.Body.Close()
+		return readSSE(resp.Body, emit)
+	}
+	return lastErr
+}
+
+func shouldRetryRequest(ctx context.Context, attempt int, status int) bool {
+	if ctx.Err() != nil || attempt >= maxChatAttempts {
+		return false
+	}
+	if status == 0 {
+		return true
+	}
+	return status == http.StatusRequestTimeout || status == http.StatusTooManyRequests || status >= 500
+}
+
+func waitBeforeRetry(ctx context.Context) error {
+	timer := time.NewTimer(retryDelay)
+	defer timer.Stop()
+	select {
+	case <-ctx.Done():
+		return ctx.Err()
+	case <-timer.C:
+		return nil
+	}
+}
+
+func readSSE(r io.Reader, emit func(StreamEvent) error) error {
+	reader := bufio.NewReader(r)
+	for {
+		line, err := reader.ReadString('\n')
+		if err != nil && !errors.Is(err, io.EOF) {
+			return fmt.Errorf("read stream: %w", err)
+		}
+
+		line = strings.TrimSpace(line)
+		if strings.HasPrefix(line, "data:") {
+			payload := strings.TrimSpace(strings.TrimPrefix(line, "data:"))
+			if payload == "[DONE]" {
+				return nil
+			}
+			event, parseErr := parseStreamPayload(payload)
+			if parseErr != nil {
+				return parseErr
+			}
+			if event.Content != "" || len(event.ToolCalls) > 0 || event.FinishReason != "" || event.Usage != nil {
+				if emitErr := emit(event); emitErr != nil {
+					return emitErr
+				}
+			}
+		}
+
+		if errors.Is(err, io.EOF) {
+			return nil
+		}
+	}
+}
+
+type streamPayload struct {
+	Choices []struct {
+		Delta struct {
+			Content   string              `json:"content"`
+			ToolCalls []toolCallDeltaJSON `json:"tool_calls"`
+		} `json:"delta"`
+		FinishReason string `json:"finish_reason"`
+	} `json:"choices"`
+	Usage *Usage `json:"usage,omitempty"`
+	Error *struct {
+		Message string `json:"message"`
+		Type    string `json:"type"`
+	} `json:"error,omitempty"`
+}
+
+type toolCallDeltaJSON struct {
+	Index    int    `json:"index"`
+	ID       string `json:"id"`
+	Type     string `json:"type"`
+	Function struct {
+		Name      string `json:"name"`
+		Arguments string `json:"arguments"`
+	} `json:"function"`
+}
+
+func parseStreamPayload(payload string) (StreamEvent, error) {
+	var decoded streamPayload
+	if err := json.Unmarshal([]byte(payload), &decoded); err != nil {
+		return StreamEvent{}, fmt.Errorf("parse stream payload: %w", err)
+	}
+	if decoded.Error != nil {
+		if decoded.Error.Type != "" {
+			return StreamEvent{}, fmt.Errorf("provider error: %s: %s", decoded.Error.Type, decoded.Error.Message)
+		}
+		return StreamEvent{}, fmt.Errorf("provider error: %s", decoded.Error.Message)
+	}
+	event := StreamEvent{Usage: decoded.Usage}
+	if len(decoded.Choices) == 0 {
+		return event, nil
+	}
+
+	choice := decoded.Choices[0]
+	event.Content = choice.Delta.Content
+	event.FinishReason = choice.FinishReason
+	for _, tc := range choice.Delta.ToolCalls {
+		event.ToolCalls = append(event.ToolCalls, ToolCallDelta{
+			Index:     tc.Index,
+			ID:        tc.ID,
+			Type:      tc.Type,
+			Name:      tc.Function.Name,
+			Arguments: tc.Function.Arguments,
+		})
+	}
+	return event, nil
+}
@@ -0,0 +1,164 @@
+package llm
+
+import (
+	"context"
+	"encoding/json"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+)
+
+func TestOpenAICompatibleClientStreamsContentAndToolCalls(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path != "/v1/chat/completions" {
+			t.Fatalf("path = %s", r.URL.Path)
+		}
+		if got := r.Header.Get("Authorization"); got != "Bearer sk-test" {
+			t.Fatalf("authorization = %q", got)
+		}
+		body := mustReadBody(t, r)
+		var raw map[string]any
+		if err := json.NewDecoder(strings.NewReader(body)).Decode(&raw); err != nil {
+			t.Fatal(err)
+		}
+		var req ChatRequest
+		if err := json.NewDecoder(strings.NewReader(body)).Decode(&req); err != nil {
+			t.Fatal(err)
+		}
+		if raw["thinking"] != "high" {
+			t.Fatalf("thinking = %#v", raw["thinking"])
+		}
+		streamOptions, ok := raw["stream_options"].(map[string]any)
+		if !ok || streamOptions["include_usage"] != true {
+			t.Fatalf("stream_options = %#v", raw["stream_options"])
+		}
+		if !req.Stream {
+			t.Fatal("request did not enable stream")
+		}
+		if req.ToolChoice != "auto" && len(req.Tools) > 0 {
+			t.Fatalf("tool_choice = %q", req.ToolChoice)
+		}
+		w.Header().Set("Content-Type", "text/event-stream")
+		_, _ = w.Write([]byte(`data: {"choices":[{"delta":{"content":"hi "},"finish_reason":""}]}` + "\n\n"))
+		_, _ = w.Write([]byte(`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","type":"function","function":{"name":"file_read","arguments":"{\"path\""}}]},"finish_reason":""}]}` + "\n\n"))
+		_, _ = w.Write([]byte(`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"function":{"arguments":":\"README.md\"}"}}]},"finish_reason":"tool_calls"}]}` + "\n\n"))
+		_, _ = w.Write([]byte("data: [DONE]\n\n"))
+	}))
+	defer server.Close()
+
+	client := NewOpenAICompatibleClient(server.URL, "sk-test", server.Client())
+	var content strings.Builder
+	var calls []ToolCallDelta
+	err := client.ChatStream(context.Background(), ChatRequest{Model: "test", Extra: map[string]any{"thinking": "high"}}, func(event StreamEvent) error {
+		content.WriteString(event.Content)
+		calls = append(calls, event.ToolCalls...)
+		return nil
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if content.String() != "hi " {
+		t.Fatalf("content = %q", content.String())
+	}
+	if len(calls) != 2 {
+		t.Fatalf("calls = %#v", calls)
+	}
+	if calls[0].Name != "file_read" || calls[1].Arguments != ":\"README.md\"}" {
+		t.Fatalf("unexpected calls: %#v", calls)
+	}
+}
+
+func TestChatStreamCapturesUsage(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		body := mustReadBody(t, r)
+		var raw map[string]any
+		if err := json.NewDecoder(strings.NewReader(body)).Decode(&raw); err != nil {
+			t.Fatal(err)
+		}
+		streamOptions, ok := raw["stream_options"].(map[string]any)
+		if !ok || streamOptions["include_usage"] != true {
+			t.Fatalf("stream_options = %#v", raw["stream_options"])
+		}
+
+		w.Header().Set("Content-Type", "text/event-stream")
+		_, _ = w.Write([]byte(`data: {"choices":[{"delta":{"content":"ok"},"finish_reason":""}]}` + "\n\n"))
+		_, _ = w.Write([]byte(`data: {"choices":[],"usage":{"prompt_tokens":12,"completion_tokens":3,"total_tokens":15}}` + "\n\n"))
+		_, _ = w.Write([]byte("data: [DONE]\n\n"))
+	}))
+	defer server.Close()
+
+	client := NewOpenAICompatibleClient(server.URL, "sk-test", server.Client())
+	var usage *Usage
+	err := client.ChatStream(context.Background(), ChatRequest{Model: "test"}, func(event StreamEvent) error {
+		if event.Usage != nil {
+			usage = event.Usage
+		}
+		return nil
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if usage == nil {
+		t.Fatal("usage was not emitted")
+	}
+	if usage.PromptTokens != 12 || usage.CompletionTokens != 3 || usage.TotalTokens != 15 {
+		t.Fatalf("usage = %#v", usage)
+	}
+}
+
+func TestOpenAICompatibleClientRetriesTransientHTTPError(t *testing.T) {
+	attempts := 0
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		attempts++
+		if attempts == 1 {
+			http.Error(w, "temporary", http.StatusBadGateway)
+			return
+		}
+		w.Header().Set("Content-Type", "text/event-stream")
+		_, _ = w.Write([]byte(`data: {"choices":[{"delta":{"content":"ok"},"finish_reason":""}]}` + "\n\n"))
+		_, _ = w.Write([]byte("data: [DONE]\n\n"))
+	}))
+	defer server.Close()
+
+	client := NewOpenAICompatibleClient(server.URL, "sk-test", server.Client())
+	var content strings.Builder
+	err := client.ChatStream(context.Background(), ChatRequest{Model: "test"}, func(event StreamEvent) error {
+		content.WriteString(event.Content)
+		return nil
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if attempts != 2 {
+		t.Fatalf("attempts = %d, want 2", attempts)
+	}
+	if content.String() != "ok" {
+		t.Fatalf("content = %q", content.String())
+	}
+}
+
+func mustReadBody(t *testing.T, r *http.Request) string {
+	t.Helper()
+	data, err := io.ReadAll(r.Body)
+	if err != nil {
+		t.Fatal(err)
+	}
+	return string(data)
+}
+
+func TestOpenAICompatibleClientHTTPError(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		http.Error(w, "bad key", http.StatusUnauthorized)
+	}))
+	defer server.Close()
+
+	client := NewOpenAICompatibleClient(server.URL, "sk-test", server.Client())
+	err := client.ChatStream(context.Background(), ChatRequest{Model: "test"}, func(StreamEvent) error {
+		return nil
+	})
+	if err == nil || !strings.Contains(err.Error(), "status=401") {
+		t.Fatalf("err = %v", err)
+	}
+}
@@ -0,0 +1,43 @@
+package llm
+
+import "encoding/json"
+
+const tokenEstimateCharRatio = 4
+
+// EstimateTokens estimates the token cost of messages and tool definitions in a
+// chat completion request. It intentionally uses a simple JSON-size heuristic:
+// roughly four UTF-8 bytes per token plus a small per-message framing overhead.
+func EstimateTokens(messages []Message, tools []Tool) int {
+	if len(messages) == 0 && len(tools) == 0 {
+		return 0
+	}
+
+	payload := struct {
+		Messages []Message `json:"messages,omitempty"`
+		Tools    []Tool    `json:"tools,omitempty"`
+	}{
+		Messages: messages,
+		Tools:    tools,
+	}
+	data, err := json.Marshal(payload)
+	if err != nil {
+		return 0
+	}
+
+	return estimateBytesAsTokens(len(data)) + len(messages)*4
+}
+
+func estimateMessageTokens(msg Message) int {
+	data, err := json.Marshal(msg)
+	if err != nil {
+		return 0
+	}
+	return estimateBytesAsTokens(len(data)) + 4
+}
+
+func estimateBytesAsTokens(n int) int {
+	if n <= 0 {
+		return 0
+	}
+	return (n + tokenEstimateCharRatio - 1) / tokenEstimateCharRatio
+}
@@ -0,0 +1,45 @@
+package llm
+
+import "testing"
+
+func TestEstimateTokensEmpty(t *testing.T) {
+	if got := EstimateTokens(nil, nil); got != 0 {
+		t.Fatalf("EstimateTokens(nil, nil) = %d, want 0", got)
+	}
+}
+
+func TestEstimateTokens(t *testing.T) {
+	shortMessages := []Message{{Role: RoleUser, Content: "hello"}}
+	short := EstimateTokens(shortMessages, nil)
+	if short <= 4 {
+		t.Fatalf("short estimate = %d, want message framing plus content", short)
+	}
+	if short > 50 {
+		t.Fatalf("short estimate = %d, looks too high for one short message", short)
+	}
+
+	longMessages := []Message{
+		{Role: RoleUser, Content: "hello"},
+		{Role: RoleAssistant, Content: "This is a longer response with enough content to exceed the short estimate."},
+	}
+	long := EstimateTokens(longMessages, nil)
+	if long <= short {
+		t.Fatalf("long estimate = %d, want > short estimate %d", long, short)
+	}
+
+	withTools := EstimateTokens(shortMessages, []Tool{{
+		Type: "function",
+		Function: ToolFunction{
+			Name:        "file_read",
+			Description: "Read a file",
+			Parameters:  []byte(`{"type":"object"}`),
+		},
+	}})
+	if withTools <= short {
+		t.Fatalf("withTools estimate = %d, want > short estimate %d", withTools, short)
+	}
+
+	if msgTokens := estimateMessageTokens(shortMessages[0]); msgTokens <= 4 {
+		t.Fatalf("estimateMessageTokens = %d, want message framing plus content", msgTokens)
+	}
+}
@@ -0,0 +1,97 @@
+package llm
+
+import (
+	"context"
+	"encoding/json"
+)
+
+const (
+	RoleSystem    = "system"
+	RoleUser      = "user"
+	RoleAssistant = "assistant"
+	RoleTool      = "tool"
+)
+
+type Provider interface {
+	ChatStream(ctx context.Context, req ChatRequest, emit func(StreamEvent) error) error
+}
+
+type ChatRequest struct {
+	Model      string         `json:"model"`
+	Messages   []Message      `json:"messages"`
+	Tools      []Tool         `json:"tools,omitempty"`
+	ToolChoice string         `json:"tool_choice,omitempty"`
+	Stream     bool           `json:"stream"`
+	Extra      map[string]any `json:"-"`
+}
+
+func (r ChatRequest) MarshalJSON() ([]byte, error) {
+	payload := map[string]any{
+		"model":    r.Model,
+		"messages": r.Messages,
+		"stream":   r.Stream,
+	}
+	if len(r.Tools) > 0 {
+		payload["tools"] = r.Tools
+	}
+	if r.ToolChoice != "" {
+		payload["tool_choice"] = r.ToolChoice
+	}
+	for key, value := range r.Extra {
+		if _, exists := payload[key]; exists {
+			continue
+		}
+		payload[key] = value
+	}
+	return json.Marshal(payload)
+}
+
+type Message struct {
+	Role       string     `json:"role"`
+	Content    string     `json:"content,omitempty"`
+	ToolCalls  []ToolCall `json:"tool_calls,omitempty"`
+	ToolCallID string     `json:"tool_call_id,omitempty"`
+}
+
+type ToolCall struct {
+	ID       string       `json:"id,omitempty"`
+	Type     string       `json:"type,omitempty"`
+	Function FunctionCall `json:"function,omitempty"`
+}
+
+type FunctionCall struct {
+	Name      string `json:"name,omitempty"`
+	Arguments string `json:"arguments,omitempty"`
+}
+
+type Tool struct {
+	Type     string       `json:"type"`
+	Function ToolFunction `json:"function"`
+}
+
+type ToolFunction struct {
+	Name        string          `json:"name"`
+	Description string          `json:"description"`
+	Parameters  json.RawMessage `json:"parameters"`
+}
+
+type StreamEvent struct {
+	Content      string
+	ToolCalls    []ToolCallDelta
+	FinishReason string
+	Usage        *Usage
+}
+
+type Usage struct {
+	PromptTokens     int `json:"prompt_tokens"`
+	CompletionTokens int `json:"completion_tokens"`
+	TotalTokens      int `json:"total_tokens"`
+}
+
+type ToolCallDelta struct {
+	Index     int
+	ID        string
+	Type      string
+	Name      string
+	Arguments string
+}