2023-04-27 04:18:18 +00:00
|
|
|
package api
|
|
|
|
|
|
|
|
import (
|
|
|
|
"bufio"
|
2023-05-02 18:03:35 +00:00
|
|
|
"bytes"
|
2023-04-27 04:18:18 +00:00
|
|
|
"encoding/json"
|
|
|
|
"fmt"
|
|
|
|
"strings"
|
|
|
|
|
|
|
|
model "github.com/go-skynet/LocalAI/pkg/model"
|
|
|
|
"github.com/gofiber/fiber/v2"
|
|
|
|
"github.com/rs/zerolog/log"
|
|
|
|
"github.com/valyala/fasthttp"
|
|
|
|
)
|
|
|
|
|
|
|
|
// APIError provides error information returned by the OpenAI API.
|
|
|
|
type APIError struct {
|
|
|
|
Code any `json:"code,omitempty"`
|
|
|
|
Message string `json:"message"`
|
|
|
|
Param *string `json:"param,omitempty"`
|
|
|
|
Type string `json:"type"`
|
|
|
|
}
|
|
|
|
|
|
|
|
type ErrorResponse struct {
|
|
|
|
Error *APIError `json:"error,omitempty"`
|
|
|
|
}
|
|
|
|
|
2023-05-03 15:29:18 +00:00
|
|
|
type OpenAIUsage struct {
|
|
|
|
PromptTokens int `json:"prompt_tokens"`
|
|
|
|
CompletionTokens int `json:"completion_tokens"`
|
|
|
|
TotalTokens int `json:"total_tokens"`
|
|
|
|
}
|
|
|
|
|
2023-05-05 09:20:06 +00:00
|
|
|
type Item struct {
|
|
|
|
Embedding []float32 `json:"embedding"`
|
|
|
|
Index int `json:"index"`
|
|
|
|
Object string `json:"object,omitempty"`
|
|
|
|
}
|
|
|
|
|
2023-04-27 04:18:18 +00:00
|
|
|
type OpenAIResponse struct {
|
2023-05-05 09:20:06 +00:00
|
|
|
Created int `json:"created,omitempty"`
|
|
|
|
Object string `json:"object,omitempty"`
|
|
|
|
ID string `json:"id,omitempty"`
|
|
|
|
Model string `json:"model,omitempty"`
|
|
|
|
Choices []Choice `json:"choices,omitempty"`
|
|
|
|
Data []Item `json:"data,omitempty"`
|
|
|
|
|
|
|
|
Usage OpenAIUsage `json:"usage"`
|
2023-04-27 04:18:18 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
type Choice struct {
|
|
|
|
Index int `json:"index,omitempty"`
|
|
|
|
FinishReason string `json:"finish_reason,omitempty"`
|
|
|
|
Message *Message `json:"message,omitempty"`
|
|
|
|
Delta *Message `json:"delta,omitempty"`
|
|
|
|
Text string `json:"text,omitempty"`
|
|
|
|
}
|
|
|
|
|
|
|
|
type Message struct {
|
|
|
|
Role string `json:"role,omitempty" yaml:"role"`
|
|
|
|
Content string `json:"content,omitempty" yaml:"content"`
|
|
|
|
}
|
|
|
|
|
|
|
|
type OpenAIModel struct {
|
|
|
|
ID string `json:"id"`
|
|
|
|
Object string `json:"object"`
|
|
|
|
}
|
|
|
|
|
|
|
|
type OpenAIRequest struct {
|
|
|
|
Model string `json:"model" yaml:"model"`
|
|
|
|
|
|
|
|
// Prompt is read only by completion API calls
|
2023-05-03 11:13:31 +00:00
|
|
|
Prompt interface{} `json:"prompt" yaml:"prompt"`
|
2023-04-27 04:18:18 +00:00
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
// Edit endpoint
|
2023-05-05 13:54:59 +00:00
|
|
|
Instruction string `json:"instruction" yaml:"instruction"`
|
|
|
|
Input interface{} `json:"input" yaml:"input"`
|
2023-04-29 07:22:09 +00:00
|
|
|
|
2023-05-02 21:30:00 +00:00
|
|
|
Stop interface{} `json:"stop" yaml:"stop"`
|
2023-04-27 04:18:18 +00:00
|
|
|
|
|
|
|
// Messages is read only by chat/completion API calls
|
|
|
|
Messages []Message `json:"messages" yaml:"messages"`
|
|
|
|
|
|
|
|
Stream bool `json:"stream"`
|
|
|
|
Echo bool `json:"echo"`
|
|
|
|
// Common options between all the API calls
|
|
|
|
TopP float64 `json:"top_p" yaml:"top_p"`
|
|
|
|
TopK int `json:"top_k" yaml:"top_k"`
|
|
|
|
Temperature float64 `json:"temperature" yaml:"temperature"`
|
|
|
|
Maxtokens int `json:"max_tokens" yaml:"max_tokens"`
|
|
|
|
|
|
|
|
N int `json:"n"`
|
|
|
|
|
|
|
|
// Custom parameters - not present in the OpenAI API
|
|
|
|
Batch int `json:"batch" yaml:"batch"`
|
|
|
|
F16 bool `json:"f16" yaml:"f16"`
|
|
|
|
IgnoreEOS bool `json:"ignore_eos" yaml:"ignore_eos"`
|
|
|
|
RepeatPenalty float64 `json:"repeat_penalty" yaml:"repeat_penalty"`
|
|
|
|
Keep int `json:"n_keep" yaml:"n_keep"`
|
|
|
|
|
2023-05-05 11:45:37 +00:00
|
|
|
MirostatETA float64 `json:"mirostat_eta" yaml:"mirostat_eta"`
|
|
|
|
MirostatTAU float64 `json:"mirostat_tau" yaml:"mirostat_tau"`
|
|
|
|
Mirostat int `json:"mirostat" yaml:"mirostat"`
|
|
|
|
|
2023-04-27 04:18:18 +00:00
|
|
|
Seed int `json:"seed" yaml:"seed"`
|
|
|
|
}
|
|
|
|
|
|
|
|
func defaultRequest(modelFile string) OpenAIRequest {
|
|
|
|
return OpenAIRequest{
|
|
|
|
TopP: 0.7,
|
|
|
|
TopK: 80,
|
|
|
|
Maxtokens: 512,
|
|
|
|
Temperature: 0.9,
|
|
|
|
Model: modelFile,
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
// https://platform.openai.com/docs/api-reference/completions
|
|
|
|
func completionEndpoint(cm ConfigMerger, debug bool, loader *model.ModelLoader, threads, ctx int, f16 bool) func(c *fiber.Ctx) error {
|
|
|
|
return func(c *fiber.Ctx) error {
|
|
|
|
config, input, err := readConfig(cm, c, loader, debug, threads, ctx, f16)
|
|
|
|
if err != nil {
|
|
|
|
return fmt.Errorf("failed reading parameters from request:%w", err)
|
2023-04-27 04:18:18 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
log.Debug().Msgf("Parameter Config: %+v", config)
|
|
|
|
|
|
|
|
templateFile := config.Model
|
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
if config.TemplateConfig.Completion != "" {
|
2023-04-27 04:18:18 +00:00
|
|
|
templateFile = config.TemplateConfig.Completion
|
|
|
|
}
|
|
|
|
|
2023-05-03 11:13:31 +00:00
|
|
|
var result []Choice
|
2023-05-05 13:54:59 +00:00
|
|
|
for _, i := range config.PromptStrings {
|
2023-05-03 11:13:31 +00:00
|
|
|
// A model can have a "file.bin.tmpl" file associated with a prompt template prefix
|
|
|
|
templatedInput, err := loader.TemplatePrefix(templateFile, struct {
|
|
|
|
Input string
|
|
|
|
}{Input: i})
|
|
|
|
if err == nil {
|
|
|
|
i = templatedInput
|
|
|
|
log.Debug().Msgf("Template found, input modified to: %s", i)
|
|
|
|
}
|
|
|
|
|
|
|
|
r, err := ComputeChoices(i, input, config, loader, func(s string, c *[]Choice) {
|
|
|
|
*c = append(*c, Choice{Text: s})
|
|
|
|
}, nil)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
|
|
|
|
result = append(result, r...)
|
2023-04-29 07:22:09 +00:00
|
|
|
}
|
2023-04-27 04:18:18 +00:00
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
resp := &OpenAIResponse{
|
|
|
|
Model: input.Model, // we have to return what the user sent here, due to OpenAI spec.
|
|
|
|
Choices: result,
|
|
|
|
Object: "text_completion",
|
2023-04-27 04:18:18 +00:00
|
|
|
}
|
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
jsonResult, _ := json.Marshal(resp)
|
|
|
|
log.Debug().Msgf("Response: %s", jsonResult)
|
|
|
|
|
|
|
|
// Return the prediction in the response body
|
|
|
|
return c.JSON(resp)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2023-05-05 22:42:08 +00:00
|
|
|
// https://platform.openai.com/docs/api-reference/embeddings
|
2023-05-05 09:20:06 +00:00
|
|
|
func embeddingsEndpoint(cm ConfigMerger, debug bool, loader *model.ModelLoader, threads, ctx int, f16 bool) func(c *fiber.Ctx) error {
|
|
|
|
return func(c *fiber.Ctx) error {
|
|
|
|
config, input, err := readConfig(cm, c, loader, debug, threads, ctx, f16)
|
|
|
|
if err != nil {
|
|
|
|
return fmt.Errorf("failed reading parameters from request:%w", err)
|
|
|
|
}
|
|
|
|
|
|
|
|
log.Debug().Msgf("Parameter Config: %+v", config)
|
2023-05-05 13:54:59 +00:00
|
|
|
items := []Item{}
|
2023-05-05 09:20:06 +00:00
|
|
|
|
2023-05-05 13:54:59 +00:00
|
|
|
for i, s := range config.InputStrings {
|
2023-05-05 09:20:06 +00:00
|
|
|
|
2023-05-05 13:54:59 +00:00
|
|
|
// get the model function to call for the result
|
|
|
|
embedFn, err := ModelEmbedding(s, loader, *config)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
|
|
|
|
embeddings, err := embedFn()
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
items = append(items, Item{Embedding: embeddings, Index: i, Object: "embedding"})
|
2023-05-05 09:20:06 +00:00
|
|
|
}
|
2023-05-05 13:54:59 +00:00
|
|
|
|
2023-05-05 09:20:06 +00:00
|
|
|
resp := &OpenAIResponse{
|
|
|
|
Model: input.Model, // we have to return what the user sent here, due to OpenAI spec.
|
2023-05-05 13:54:59 +00:00
|
|
|
Data: items,
|
2023-05-05 09:20:06 +00:00
|
|
|
Object: "list",
|
|
|
|
}
|
|
|
|
|
|
|
|
jsonResult, _ := json.Marshal(resp)
|
|
|
|
log.Debug().Msgf("Response: %s", jsonResult)
|
|
|
|
|
|
|
|
// Return the prediction in the response body
|
|
|
|
return c.JSON(resp)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
func chatEndpoint(cm ConfigMerger, debug bool, loader *model.ModelLoader, threads, ctx int, f16 bool) func(c *fiber.Ctx) error {
|
2023-05-04 17:49:43 +00:00
|
|
|
|
|
|
|
process := func(s string, req *OpenAIRequest, config *Config, loader *model.ModelLoader, responses chan OpenAIResponse) {
|
|
|
|
ComputeChoices(s, req, config, loader, func(s string, c *[]Choice) {}, func(s string) bool {
|
|
|
|
resp := OpenAIResponse{
|
|
|
|
Model: req.Model, // we have to return what the user sent here, due to OpenAI spec.
|
|
|
|
Choices: []Choice{{Delta: &Message{Role: "assistant", Content: s}}},
|
|
|
|
Object: "chat.completion.chunk",
|
|
|
|
}
|
|
|
|
log.Debug().Msgf("Sending goroutine: %s", s)
|
|
|
|
|
|
|
|
responses <- resp
|
|
|
|
return true
|
|
|
|
})
|
|
|
|
close(responses)
|
|
|
|
}
|
2023-04-29 07:22:09 +00:00
|
|
|
return func(c *fiber.Ctx) error {
|
|
|
|
config, input, err := readConfig(cm, c, loader, debug, threads, ctx, f16)
|
2023-04-27 04:18:18 +00:00
|
|
|
if err != nil {
|
2023-04-29 07:22:09 +00:00
|
|
|
return fmt.Errorf("failed reading parameters from request:%w", err)
|
2023-04-27 04:18:18 +00:00
|
|
|
}
|
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
log.Debug().Msgf("Parameter Config: %+v", config)
|
2023-04-27 04:18:18 +00:00
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
var predInput string
|
2023-04-27 04:18:18 +00:00
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
mess := []string{}
|
|
|
|
for _, i := range input.Messages {
|
|
|
|
r := config.Roles[i.Role]
|
|
|
|
if r == "" {
|
|
|
|
r = i.Role
|
2023-04-27 04:18:18 +00:00
|
|
|
}
|
2023-04-29 07:22:09 +00:00
|
|
|
|
|
|
|
content := fmt.Sprint(r, " ", i.Content)
|
|
|
|
mess = append(mess, content)
|
2023-04-27 04:18:18 +00:00
|
|
|
}
|
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
predInput = strings.Join(mess, "\n")
|
2023-04-27 04:18:18 +00:00
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
if input.Stream {
|
|
|
|
log.Debug().Msgf("Stream request received")
|
2023-05-02 18:03:35 +00:00
|
|
|
c.Context().SetContentType("text/event-stream")
|
2023-04-29 07:22:09 +00:00
|
|
|
//c.Response().Header.SetContentType(fiber.MIMETextHTMLCharsetUTF8)
|
2023-05-02 18:03:35 +00:00
|
|
|
// c.Set("Content-Type", "text/event-stream")
|
2023-04-29 07:22:09 +00:00
|
|
|
c.Set("Cache-Control", "no-cache")
|
|
|
|
c.Set("Connection", "keep-alive")
|
|
|
|
c.Set("Transfer-Encoding", "chunked")
|
|
|
|
}
|
2023-04-27 04:18:18 +00:00
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
templateFile := config.Model
|
|
|
|
|
|
|
|
if config.TemplateConfig.Chat != "" {
|
|
|
|
templateFile = config.TemplateConfig.Chat
|
|
|
|
}
|
|
|
|
|
|
|
|
// A model can have a "file.bin.tmpl" file associated with a prompt template prefix
|
|
|
|
templatedInput, err := loader.TemplatePrefix(templateFile, struct {
|
|
|
|
Input string
|
|
|
|
}{Input: predInput})
|
|
|
|
if err == nil {
|
|
|
|
predInput = templatedInput
|
|
|
|
log.Debug().Msgf("Template found, input modified to: %s", predInput)
|
|
|
|
}
|
|
|
|
|
2023-04-27 04:18:18 +00:00
|
|
|
if input.Stream {
|
2023-05-02 18:03:35 +00:00
|
|
|
responses := make(chan OpenAIResponse)
|
|
|
|
|
2023-05-04 17:49:43 +00:00
|
|
|
go process(predInput, input, config, loader, responses)
|
2023-05-02 18:03:35 +00:00
|
|
|
|
2023-04-27 04:18:18 +00:00
|
|
|
c.Context().SetBodyStreamWriter(fasthttp.StreamWriter(func(w *bufio.Writer) {
|
|
|
|
|
2023-05-02 18:03:35 +00:00
|
|
|
for ev := range responses {
|
|
|
|
var buf bytes.Buffer
|
|
|
|
enc := json.NewEncoder(&buf)
|
|
|
|
enc.Encode(ev)
|
2023-04-27 04:18:18 +00:00
|
|
|
|
2023-05-02 18:03:35 +00:00
|
|
|
fmt.Fprintf(w, "event: data\n\n")
|
|
|
|
fmt.Fprintf(w, "data: %v\n\n", buf.String())
|
|
|
|
log.Debug().Msgf("Sending chunk: %s", buf.String())
|
|
|
|
w.Flush()
|
|
|
|
}
|
2023-04-27 04:18:18 +00:00
|
|
|
|
2023-05-02 18:03:35 +00:00
|
|
|
w.WriteString("event: data\n\n")
|
2023-04-27 04:18:18 +00:00
|
|
|
resp := &OpenAIResponse{
|
|
|
|
Model: input.Model, // we have to return what the user sent here, due to OpenAI spec.
|
2023-04-29 07:22:09 +00:00
|
|
|
Choices: []Choice{{FinishReason: "stop"}},
|
2023-04-27 04:18:18 +00:00
|
|
|
}
|
|
|
|
respData, _ := json.Marshal(resp)
|
|
|
|
|
2023-05-02 18:03:35 +00:00
|
|
|
w.WriteString(fmt.Sprintf("data: %s\n\n", respData))
|
2023-04-27 04:18:18 +00:00
|
|
|
w.Flush()
|
|
|
|
}))
|
|
|
|
return nil
|
|
|
|
}
|
2023-04-29 07:22:09 +00:00
|
|
|
|
2023-05-02 18:03:35 +00:00
|
|
|
result, err := ComputeChoices(predInput, input, config, loader, func(s string, c *[]Choice) {
|
|
|
|
*c = append(*c, Choice{Message: &Message{Role: "assistant", Content: s}})
|
|
|
|
}, nil)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
|
|
|
|
resp := &OpenAIResponse{
|
|
|
|
Model: input.Model, // we have to return what the user sent here, due to OpenAI spec.
|
|
|
|
Choices: result,
|
|
|
|
Object: "chat.completion",
|
|
|
|
}
|
2023-05-04 15:32:23 +00:00
|
|
|
respData, _ := json.Marshal(resp)
|
|
|
|
log.Debug().Msgf("Response: %s", respData)
|
2023-05-02 18:03:35 +00:00
|
|
|
|
2023-04-29 07:22:09 +00:00
|
|
|
// Return the prediction in the response body
|
|
|
|
return c.JSON(resp)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
func editEndpoint(cm ConfigMerger, debug bool, loader *model.ModelLoader, threads, ctx int, f16 bool) func(c *fiber.Ctx) error {
|
|
|
|
return func(c *fiber.Ctx) error {
|
|
|
|
config, input, err := readConfig(cm, c, loader, debug, threads, ctx, f16)
|
|
|
|
if err != nil {
|
|
|
|
return fmt.Errorf("failed reading parameters from request:%w", err)
|
|
|
|
}
|
|
|
|
|
|
|
|
log.Debug().Msgf("Parameter Config: %+v", config)
|
|
|
|
|
|
|
|
templateFile := config.Model
|
|
|
|
|
|
|
|
if config.TemplateConfig.Edit != "" {
|
|
|
|
templateFile = config.TemplateConfig.Edit
|
|
|
|
}
|
|
|
|
|
2023-05-05 13:54:59 +00:00
|
|
|
var result []Choice
|
|
|
|
for _, i := range config.InputStrings {
|
|
|
|
// A model can have a "file.bin.tmpl" file associated with a prompt template prefix
|
|
|
|
templatedInput, err := loader.TemplatePrefix(templateFile, struct {
|
|
|
|
Input string
|
|
|
|
Instruction string
|
|
|
|
}{Input: i})
|
|
|
|
if err == nil {
|
|
|
|
i = templatedInput
|
|
|
|
log.Debug().Msgf("Template found, input modified to: %s", i)
|
|
|
|
}
|
2023-04-29 07:22:09 +00:00
|
|
|
|
2023-05-05 13:54:59 +00:00
|
|
|
r, err := ComputeChoices(i, input, config, loader, func(s string, c *[]Choice) {
|
|
|
|
*c = append(*c, Choice{Text: s})
|
|
|
|
}, nil)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
|
|
|
|
result = append(result, r...)
|
2023-04-29 07:22:09 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
resp := &OpenAIResponse{
|
|
|
|
Model: input.Model, // we have to return what the user sent here, due to OpenAI spec.
|
|
|
|
Choices: result,
|
|
|
|
Object: "edit",
|
|
|
|
}
|
|
|
|
|
|
|
|
jsonResult, _ := json.Marshal(resp)
|
|
|
|
log.Debug().Msgf("Response: %s", jsonResult)
|
|
|
|
|
|
|
|
// Return the prediction in the response body
|
|
|
|
return c.JSON(resp)
|
2023-04-27 04:18:18 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
func listModels(loader *model.ModelLoader, cm ConfigMerger) func(ctx *fiber.Ctx) error {
|
|
|
|
return func(c *fiber.Ctx) error {
|
|
|
|
models, err := loader.ListModels()
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
var mm map[string]interface{} = map[string]interface{}{}
|
|
|
|
|
|
|
|
dataModels := []OpenAIModel{}
|
|
|
|
for _, m := range models {
|
|
|
|
mm[m] = nil
|
|
|
|
dataModels = append(dataModels, OpenAIModel{ID: m, Object: "model"})
|
|
|
|
}
|
|
|
|
|
|
|
|
for k := range cm {
|
|
|
|
if _, exists := mm[k]; !exists {
|
|
|
|
dataModels = append(dataModels, OpenAIModel{ID: k, Object: "model"})
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return c.JSON(struct {
|
|
|
|
Object string `json:"object"`
|
|
|
|
Data []OpenAIModel `json:"data"`
|
|
|
|
}{
|
|
|
|
Object: "list",
|
|
|
|
Data: dataModels,
|
|
|
|
})
|
|
|
|
}
|
|
|
|
}
|