242 lines
5.8 KiB
Go
242 lines
5.8 KiB
Go
package controllers
|
|
|
|
import (
|
|
"bufio"
|
|
"bytes"
|
|
"encoding/json"
|
|
"fmt"
|
|
"io"
|
|
"log"
|
|
"net/http"
|
|
"strings"
|
|
"time"
|
|
|
|
"github.com/gofiber/fiber/v2"
|
|
"github.com/sujit-baniya/fiber-boilerplate/pkg/models"
|
|
)
|
|
|
|
type DataIa struct {
|
|
Prompt string `json:"prompt"`
|
|
Model string `json:"model"`
|
|
Stream bool `json:"stream"`
|
|
}
|
|
|
|
type respuestaOllama struct {
|
|
Response string `json:"response"`
|
|
Done bool `json:"done"`
|
|
Error string `json:"error"`
|
|
}
|
|
|
|
type openaiChatMsg struct {
|
|
Role string `json:"role"`
|
|
Content string `json:"content"`
|
|
}
|
|
|
|
type openaiChatReq struct {
|
|
Model string `json:"model"`
|
|
Messages []openaiChatMsg `json:"messages"`
|
|
Stream bool `json:"stream"`
|
|
}
|
|
|
|
type openaiChatChoice struct {
|
|
Delta struct {
|
|
Content string `json:"content"`
|
|
} `json:"delta"`
|
|
FinishReason *string `json:"finish_reason"`
|
|
}
|
|
|
|
type openaiChatResp struct {
|
|
Choices []openaiChatChoice `json:"choices"`
|
|
}
|
|
|
|
type geminiPart struct {
|
|
Text string `json:"text"`
|
|
}
|
|
|
|
type geminiContent struct {
|
|
Parts []geminiPart `json:"parts"`
|
|
}
|
|
|
|
type geminiReq struct {
|
|
Contents []geminiContent `json:"contents"`
|
|
}
|
|
|
|
type geminiCandidate struct {
|
|
Content geminiContent `json:"content"`
|
|
FinishReason *string `json:"finishReason"`
|
|
}
|
|
|
|
type geminiResp struct {
|
|
Candidates []geminiCandidate `json:"candidates"`
|
|
}
|
|
|
|
func GeneraTextoStream(c *fiber.Ctx) error {
|
|
var data DataIa
|
|
if err := c.BodyParser(&data); err != nil {
|
|
return c.Status(fiber.StatusBadRequest).JSON(fiber.Map{"error": "Datos inválidos"})
|
|
}
|
|
data.Stream = true
|
|
|
|
config, err := models.GetAiConfigForService("ia")
|
|
if err != nil {
|
|
return c.Status(fiber.StatusServiceUnavailable).JSON(fiber.Map{
|
|
"error": "No hay configuración de IA activa. Configura una en /app/ai-config",
|
|
})
|
|
}
|
|
|
|
if data.Model == "" {
|
|
data.Model = config.ModelName
|
|
}
|
|
if data.Model == "" {
|
|
data.Model = "gemma3:1b"
|
|
}
|
|
|
|
provider := strings.ToLower(config.Provider)
|
|
baseURL := strings.TrimRight(config.BaseURL, "/")
|
|
hasV1 := strings.Contains(baseURL, "/v1")
|
|
|
|
var endpoint string
|
|
var bodyReader io.Reader
|
|
useOpenAI := provider != "ollama" || hasV1
|
|
useGemini := provider == "gemini"
|
|
|
|
if useGemini {
|
|
model := data.Model
|
|
endpoint = fmt.Sprintf("https://generativelanguage.googleapis.com/v1beta/models/%s:streamGenerateContent?alt=sse&key=%s", model, config.ApiKey)
|
|
gReq := geminiReq{
|
|
Contents: []geminiContent{
|
|
{Parts: []geminiPart{{Text: data.Prompt}}},
|
|
},
|
|
}
|
|
jsonData, _ := json.Marshal(gReq)
|
|
bodyReader = bytes.NewBuffer(jsonData)
|
|
} else if !useOpenAI {
|
|
// Native Ollama API
|
|
endpoint = baseURL + "/api/generate"
|
|
jsonData, _ := json.Marshal(data)
|
|
bodyReader = bytes.NewBuffer(jsonData)
|
|
} else {
|
|
// OpenAI-compatible (/v1/chat/completions)
|
|
baseURL = strings.TrimSuffix(baseURL, "/v1")
|
|
endpoint = baseURL + "/v1/chat/completions"
|
|
chatReq := openaiChatReq{
|
|
Model: data.Model,
|
|
Messages: []openaiChatMsg{
|
|
{Role: "user", Content: data.Prompt},
|
|
},
|
|
Stream: true,
|
|
}
|
|
jsonData, _ := json.Marshal(chatReq)
|
|
bodyReader = bytes.NewBuffer(jsonData)
|
|
}
|
|
|
|
log.Printf("[IA] endpoint=%s model=%s provider=%s", endpoint, data.Model, provider)
|
|
|
|
req, err := http.NewRequest("POST", endpoint, bodyReader)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
req.Header.Set("Content-Type", "application/json")
|
|
|
|
// Auth
|
|
if useGemini {
|
|
// API key ya va en la URL
|
|
} else if provider == "ollama" && config.ApiKey != "" && config.ApiKey != "ollama" {
|
|
req.SetBasicAuth("ollama", config.ApiKey)
|
|
} else if config.ApiKey != "" {
|
|
req.Header.Set("Authorization", "Bearer "+config.ApiKey)
|
|
}
|
|
|
|
client := &http.Client{Timeout: 120 * time.Second}
|
|
resp, err := client.Do(req)
|
|
if err != nil {
|
|
log.Printf("[IA] Error conectando: %v", err)
|
|
return c.Status(fiber.StatusBadGateway).JSON(fiber.Map{
|
|
"error": fmt.Sprintf("Error conectando a %s: %v", provider, err),
|
|
})
|
|
}
|
|
if resp.StatusCode != http.StatusOK {
|
|
bodyBytes, _ := io.ReadAll(resp.Body)
|
|
resp.Body.Close()
|
|
log.Printf("[IA] %s status %d: %s", provider, resp.StatusCode, string(bodyBytes))
|
|
return c.Status(fiber.StatusBadGateway).JSON(fiber.Map{
|
|
"error": fmt.Sprintf("%s respondió con status %d", provider, resp.StatusCode),
|
|
"details": string(bodyBytes),
|
|
})
|
|
}
|
|
|
|
c.Set("Content-Type", "text/plain; charset=utf-8")
|
|
c.Set("Cache-Control", "no-cache")
|
|
c.Set("X-Accel-Buffering", "no")
|
|
|
|
c.Context().SetBodyStreamWriter(func(w *bufio.Writer) {
|
|
defer resp.Body.Close()
|
|
scanner := bufio.NewScanner(resp.Body)
|
|
scanner.Buffer(make([]byte, 0, 64*1024), 1024*1024)
|
|
for scanner.Scan() {
|
|
line := scanner.Text()
|
|
if line == "" {
|
|
continue
|
|
}
|
|
if useGemini {
|
|
if !strings.HasPrefix(line, "data: ") {
|
|
continue
|
|
}
|
|
var msg geminiResp
|
|
if err := json.Unmarshal([]byte(line[6:]), &msg); err != nil {
|
|
continue
|
|
}
|
|
if len(msg.Candidates) > 0 && len(msg.Candidates[0].Content.Parts) > 0 {
|
|
text := msg.Candidates[0].Content.Parts[0].Text
|
|
if text != "" {
|
|
fmt.Fprint(w, text)
|
|
w.Flush()
|
|
}
|
|
if msg.Candidates[0].FinishReason != nil {
|
|
break
|
|
}
|
|
}
|
|
} else if useOpenAI {
|
|
var msg openaiChatResp
|
|
if err := json.Unmarshal([]byte(line), &msg); err != nil {
|
|
if err != io.EOF {
|
|
log.Printf("[IA] Error decodificando respuesta OpenAI: %v", err)
|
|
}
|
|
continue
|
|
}
|
|
if len(msg.Choices) == 0 {
|
|
continue
|
|
}
|
|
content := msg.Choices[0].Delta.Content
|
|
if content != "" {
|
|
fmt.Fprint(w, content)
|
|
w.Flush()
|
|
}
|
|
if msg.Choices[0].FinishReason != nil {
|
|
break
|
|
}
|
|
} else {
|
|
var msg respuestaOllama
|
|
if err := json.Unmarshal([]byte(line), &msg); err != nil {
|
|
continue
|
|
}
|
|
if msg.Error != "" {
|
|
log.Printf("[IA] Ollama error: %s", msg.Error)
|
|
fmt.Fprintf(w, "[Error: %s]", msg.Error)
|
|
w.Flush()
|
|
break
|
|
}
|
|
if msg.Response != "" {
|
|
fmt.Fprint(w, msg.Response)
|
|
w.Flush()
|
|
}
|
|
if msg.Done {
|
|
break
|
|
}
|
|
}
|
|
}
|
|
})
|
|
|
|
return nil
|
|
}
|