summaryrefslogtreecommitdiff
path: root/internal/llm/openai.go
blob: dbcee4d92aa23a4aa952291cbe7ff1d9121f3e87 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
package llm

import (
	"bytes"
	"context"
	"encoding/json"
	"errors"
	"fmt"
	"net/http"
	"os"
	"time"

	"hexai/internal/logging"
)

// openAIClient implements Client against OpenAI's Chat Completions API.
type openAIClient struct {
    httpClient   *http.Client
    apiKey       string
    baseURL      string
    defaultModel string
}

// Colors and base styling are provided by logging.go

func newOpenAIFromEnv(apiKey string) Client {
	base := os.Getenv("OPENAI_BASE_URL")
	if base == "" {
		base = "https://api.openai.com/v1"
	}
	model := os.Getenv("OPENAI_MODEL")
	if model == "" {
		model = "gpt-4o-mini"
	}
	return &openAIClient{
		httpClient:   &http.Client{Timeout: 30 * time.Second},
		apiKey:       apiKey,
		baseURL:      base,
		defaultModel: model,
	}
}

type oaChatRequest struct {
	Model       string      `json:"model"`
	Messages    []oaMessage `json:"messages"`
	Temperature *float64    `json:"temperature,omitempty"`
	MaxTokens   *int        `json:"max_tokens,omitempty"`
	Stop        []string    `json:"stop,omitempty"`
}

type oaMessage struct {
	Role    string `json:"role"`
	Content string `json:"content"`
}

type oaChatResponse struct {
	Choices []struct {
		Index   int `json:"index"`
		Message struct {
			Role    string `json:"role"`
			Content string `json:"content"`
		} `json:"message"`
		FinishReason string `json:"finish_reason"`
	} `json:"choices"`
	Error *struct {
		Message string `json:"message"`
		Type    string `json:"type"`
		Param   any    `json:"param"`
		Code    any    `json:"code"`
	} `json:"error,omitempty"`
}

func (c *openAIClient) Chat(ctx context.Context, messages []Message, opts ...RequestOption) (string, error) {
	if c.apiKey == "" {
		return nilStringErr("missing OpenAI API key")
	}
	o := Options{Model: c.defaultModel}
	for _, opt := range opts {
		opt(&o)
	}
	if o.Model == "" {
		o.Model = c.defaultModel
	}
	start := time.Now()
    logging.Logf("llm/openai ", "chat start model=%s temp=%.2f max_tokens=%d stop=%d messages=%d", o.Model, o.Temperature, o.MaxTokens, len(o.Stop), len(messages))
	for i, m := range messages {
        // Sending context (cyan)
        logging.Logf("llm/openai ", "msg[%d] role=%s size=%d preview=%s%s%s", i, m.Role, len(m.Content), logging.AnsiCyan, logging.PreviewForLog(m.Content), logging.AnsiBase)
	}
	req := oaChatRequest{Model: o.Model}
	req.Messages = make([]oaMessage, len(messages))
	for i, m := range messages {
		req.Messages[i] = oaMessage{Role: m.Role, Content: m.Content}
	}
	if o.Temperature != 0 {
		req.Temperature = &o.Temperature
	}
	if o.MaxTokens > 0 {
		req.MaxTokens = &o.MaxTokens
	}
	if len(o.Stop) > 0 {
		req.Stop = o.Stop
	}

	body, err := json.Marshal(req)
	if err != nil {
		c.logf("marshal error: %v", err)
		return "", err
	}
	endpoint := c.baseURL + "/chat/completions"
    logging.Logf("llm/openai ", "POST %s", endpoint)
	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, endpoint, bytes.NewReader(body))
	if err != nil {
		c.logf("new request error: %v", err)
		return "", err
	}
	httpReq.Header.Set("Content-Type", "application/json")
	httpReq.Header.Set("Authorization", "Bearer "+c.apiKey)

	resp, err := c.httpClient.Do(httpReq)
    if err != nil {
        logging.Logf("llm/openai ", "%shttp error after %s: %v%s", logging.AnsiRed, time.Since(start), err, logging.AnsiBase)
        return "", err
    }
	defer resp.Body.Close()
	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
		var apiErr oaChatResponse
		_ = json.NewDecoder(resp.Body).Decode(&apiErr)
        if apiErr.Error != nil && apiErr.Error.Message != "" {
            logging.Logf("llm/openai ", "%sapi error status=%d type=%s msg=%s duration=%s%s", logging.AnsiRed, resp.StatusCode, apiErr.Error.Type, apiErr.Error.Message, time.Since(start), logging.AnsiBase)
            return "", fmt.Errorf("openai error: %s (status %d)", apiErr.Error.Message, resp.StatusCode)
        }
        logging.Logf("llm/openai ", "%shttp non-2xx status=%d duration=%s%s", logging.AnsiRed, resp.StatusCode, time.Since(start), logging.AnsiBase)
        return "", fmt.Errorf("openai http error: status %d", resp.StatusCode)
    }
	var out oaChatResponse
    if err := json.NewDecoder(resp.Body).Decode(&out); err != nil {
        logging.Logf("llm/openai ", "%sdecode error after %s: %v%s", logging.AnsiRed, time.Since(start), err, logging.AnsiBase)
        return "", err
    }
    if len(out.Choices) == 0 {
        logging.Logf("llm/openai ", "%sno choices returned duration=%s%s", logging.AnsiRed, time.Since(start), logging.AnsiBase)
        return "", errors.New("openai: no choices returned")
    }
    content := out.Choices[0].Message.Content
    // Received context (green)
    logging.Logf("llm/openai ", "success choice=0 finish=%s size=%d preview=%s%s%s duration=%s", out.Choices[0].FinishReason, len(content), logging.AnsiGreen, logging.PreviewForLog(content), logging.AnsiBase, time.Since(start))
    return content, nil
}

// small helper to keep return type consistent
func nilStringErr(msg string) (string, error) { return "", errors.New(msg) }

func (c *openAIClient) logf(format string, args ...any) { logging.Logf("llm/openai ", format, args...) }

func trimPreview(s string, n int) string {
	if n <= 0 || len(s) <= n {
		return s
	}
	return s[:n] + "…"
}

// Provider metadata
func (c *openAIClient) Name() string          { return "openai" }
func (c *openAIClient) DefaultModel() string  { return c.defaultModel }