Merge branch 'feature/fireworks' into 'main'

add:添加fireworks

See merge request goku/apinto!257

app/apinto/driver.go

@@ -1,6 +1,7 @@
 package main
 
 import (
+	"github.com/eolinker/apinto/drivers/ai-provider/fireworks"
 	"github.com/eolinker/apinto/drivers/ai-provider/google"
 	"github.com/eolinker/apinto/drivers/ai-provider/moonshot"
 	"github.com/eolinker/apinto/drivers/ai-provider/openAI"
@@ -95,4 +96,5 @@ func driverRegister(extenderRegister eosc.IExtenderDriverRegister) {
 	moonshot.Register(extenderRegister)
 	tongyi.Register(extenderRegister)
 	zhipuai.Register(extenderRegister)
+	fireworks.Register(extenderRegister)
 }

app/apinto/profession.go

@@ -324,6 +324,12 @@ func ApintoProfession() []*eosc.ProfessionConfig {
 					Label: "zhipuai",
 					Desc:  "zhipuai",
 				},
+				{
+					Id:    "eolinker.com:apinto:fireworks", // 插件ID
+					Name:  "fireworks",                     // 驱动名称，应和定义文件的provider字段一致
+					Label: "fireworks",
+					Desc:  "fireworks",
+				},
 			},
 			Mod: eosc.ProfessionConfig_Worker,
 		},

drivers/ai-provider/fireworks/config.go

@@ -0,0 +1,22 @@
+package fireworks
+
+import (
+	"fmt"
+	"github.com/eolinker/eosc"
+)
+
+type Config struct {
+	APIKey       string `json:"fireworks_api_key"`
+	Organization string `json:"fireworks_organization"`
+}
+
+func checkConfig(v interface{}) (*Config, error) {
+	conf, ok := v.(*Config)
+	if !ok {
+		return nil, eosc.ErrorConfigType
+	}
+	if conf.APIKey == "" {
+		return nil, fmt.Errorf("api_key is required")
+	}
+	return conf, nil
+}

drivers/ai-provider/fireworks/executor.go

@@ -0,0 +1,215 @@
+package fireworks
+
+import (
+	"embed"
+	"encoding/json"
+	"fmt"
+	"github.com/eolinker/eosc/log"
+	"reflect"
+	"strconv"
+
+	"github.com/eolinker/apinto/drivers"
+
+	http_context "github.com/eolinker/eosc/eocontext/http-context"
+
+	ai_provider "github.com/eolinker/apinto/drivers/ai-provider"
+
+	"github.com/eolinker/apinto/convert"
+	"github.com/eolinker/eosc"
+	"github.com/eolinker/eosc/eocontext"
+)
+
+var (
+	//go:embed fireworks.yaml
+	providerContent []byte
+	//go:embed *
+	providerDir  embed.FS
+	modelConvert = make(map[string]convert.IConverter)
+
+	_ convert.IConverterDriver = (*executor)(nil)
+)
+
+func init() {
+	models, err := ai_provider.LoadModels(providerContent, providerDir)
+	if err != nil {
+		panic(err)
+	}
+	for key, value := range models {
+		if value.ModelProperties != nil {
+			if v, ok := modelModes[value.ModelProperties.Mode]; ok {
+				modelConvert[key] = v
+			}
+		}
+	}
+}
+
+type Converter struct {
+	apikey         string
+	balanceHandler eocontext.BalanceHandler
+	converter      convert.IConverter
+}
+
+func (c *Converter) RequestConvert(ctx eocontext.EoContext, extender map[string]interface{}) error {
+	if c.balanceHandler != nil {
+		ctx.SetBalance(c.balanceHandler)
+	}
+	httpContext, err := http_context.Assert(ctx)
+	if err != nil {
+		return err
+	}
+	httpContext.Proxy().Header().SetHeader("Authorization", "Bearer "+c.apikey)
+
+	return c.converter.RequestConvert(httpContext, extender)
+}
+
+func (c *Converter) ResponseConvert(ctx eocontext.EoContext) error {
+	return c.converter.ResponseConvert(ctx)
+}
+
+type executor struct {
+	drivers.WorkerBase
+	apikey string
+	eocontext.BalanceHandler
+}
+
+func (e *executor) GetConverter(model string) (convert.IConverter, bool) {
+	converter, ok := modelConvert[model]
+	if !ok {
+		return nil, false
+	}
+
+	return &Converter{balanceHandler: e.BalanceHandler, converter: converter, apikey: e.apikey}, true
+}
+
+func (e *executor) GetModel(model string) (convert.FGenerateConfig, bool) {
+	if _, ok := modelConvert[model]; !ok {
+		return nil, false
+	}
+	return func(cfg string) (map[string]interface{}, error) {
+
+		result := map[string]interface{}{
+			"model": model,
+		}
+		if cfg != "" {
+			tmp := make(map[string]interface{})
+			if err := json.Unmarshal([]byte(cfg), &tmp); err != nil {
+				log.Errorf("unmarshal config error: %v, cfg: %s", err, cfg)
+				return result, nil
+			}
+			modelCfg := mapToStruct[ModelConfig](tmp)
+			result["frequency_penalty"] = modelCfg.FrequencyPenalty
+			if modelCfg.MaxTokens >= 1 {
+				result["max_tokens"] = modelCfg.MaxTokens
+			}
+
+			result["presence_penalty"] = modelCfg.PresencePenalty
+			result["temperature"] = modelCfg.Temperature
+			result["top_p"] = modelCfg.TopP
+			if modelCfg.ResponseFormat == "" {
+				modelCfg.ResponseFormat = "text"
+			}
+			result["response_format"] = map[string]interface{}{
+				"type": modelCfg.ResponseFormat,
+			}
+		}
+		return result, nil
+	}, true
+}
+
+func (e *executor) Start() error {
+	return nil
+}
+
+func (e *executor) Reset(conf interface{}, workers map[eosc.RequireId]eosc.IWorker) error {
+	cfg, ok := conf.(*Config)
+	if !ok {
+		return fmt.Errorf("invalid config")
+	}
+
+	return e.reset(cfg, workers)
+}
+
+func (e *executor) reset(conf *Config, workers map[eosc.RequireId]eosc.IWorker) error {
+	e.BalanceHandler = nil
+	e.apikey = conf.APIKey
+	convert.Set(e.Id(), e)
+
+	return nil
+}
+
+func (e *executor) Stop() error {
+	e.BalanceHandler = nil
+	convert.Del(e.Id())
+	return nil
+}
+
+func (e *executor) CheckSkill(skill string) bool {
+	return convert.CheckSkill(skill)
+}
+
+type ModelConfig struct {
+	FrequencyPenalty float64 `json:"frequency_penalty"`
+	MaxTokens        int     `json:"max_tokens"`
+	PresencePenalty  float64 `json:"presence_penalty"`
+	ResponseFormat   string  `json:"response_format"`
+	Temperature      float64 `json:"temperature"`
+	TopP             float64 `json:"top_p"`
+}
+
+func mapToStruct[T any](tmp map[string]interface{}) *T {
+	// 创建目标结构体的实例
+	var result T
+	val := reflect.ValueOf(&result).Elem()
+
+	// 获取结构体的类型
+	t := val.Type()
+
+	// 遍历 map 中的键值对
+	for k, v := range tmp {
+		// 查找结构体中与键名匹配的字段
+		for i := 0; i < t.NumField(); i++ {
+			field := t.Field(i)
+			jsonTag := field.Tag.Get("json")
+			if jsonTag == k {
+				// 获取字段的值
+				fieldVal := val.Field(i)
+
+				// 如果字段不可设置，跳过
+				if !fieldVal.CanSet() {
+					continue
+				}
+
+				// 根据字段的类型，进行类型转换
+				switch fieldVal.Kind() {
+				case reflect.Float64:
+					if strVal, ok := v.(string); ok && strVal != "" {
+						// 如果是 string 类型且非空，转换为 float64
+						if floatVal, err := strconv.ParseFloat(strVal, 64); err == nil {
+							fieldVal.SetFloat(floatVal)
+						}
+					} else if floatVal, ok := v.(float64); ok {
+						fieldVal.SetFloat(floatVal)
+					}
+
+				case reflect.Int:
+					if intVal, ok := v.(int); ok {
+						fieldVal.SetInt(int64(intVal))
+					} else if strVal, ok := v.(string); ok && strVal != "" {
+						if intVal, err := strconv.Atoi(strVal); err == nil {
+							fieldVal.SetInt(int64(intVal))
+						}
+					}
+				case reflect.String:
+					if strVal, ok := v.(string); ok {
+						fieldVal.SetString(strVal)
+					}
+
+				default:
+					// 其他类型不进行转换
+				}
+			}
+		}
+	}
+
+	return &result
+}

drivers/ai-provider/fireworks/factory.go

@@ -0,0 +1,31 @@
+package fireworks
+
+import (
+	"github.com/eolinker/apinto/drivers"
+	"github.com/eolinker/eosc"
+)
+
+var name = "fireworks"
+
+// Register 注册驱动
+func Register(register eosc.IExtenderDriverRegister) {
+	register.RegisterExtenderDriver(name, NewFactory())
+}
+
+// NewFactory 创建service_http驱动工厂
+func NewFactory() eosc.IExtenderDriverFactory {
+	return drivers.NewFactory[Config](Create)
+}
+
+// Create 创建驱动实例
+func Create(id, name string, v *Config, workers map[eosc.RequireId]eosc.IWorker) (eosc.IWorker, error) {
+	_, err := checkConfig(v)
+	if err != nil {
+		return nil, err
+	}
+	w := &executor{
+		WorkerBase: drivers.Worker(id, name),
+	}
+	w.reset(v, workers)
+	return w, nil
+}

drivers/ai-provider/fireworks/fireworks.yaml

@@ -0,0 +1,18 @@
+provider: fireworks
+label:
+  zh_Hans: Fireworks AI
+  en_US: Fireworks AI
+icon_small:
+  en_US: icon_s_en.svg
+icon_large:
+  en_US: icon_l_en.svg
+background: "#FCFDFF"
+help:
+  title:
+    en_US: Get your API Key from Fireworks AI
+    zh_Hans: 从 Fireworks AI 获取 API Key
+  url:
+    en_US: https://fireworks.ai/account/api-keys
+supported_model_types:
+  - llm
+  - text-embedding

drivers/ai-provider/fireworks/llm/firefunction-v1.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/firefunction-v1
+label:
+  zh_Hans: Firefunction V1
+  en_US: Firefunction V1
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 32768
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.5'
+  output: '0.5'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/firefunction-v2.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/firefunction-v2
+label:
+  zh_Hans: Firefunction V2
+  en_US: Firefunction V2
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.9'
+  output: '0.9'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/gemma2-9b-it.yaml

@@ -0,0 +1,45 @@
+model: accounts/fireworks/models/gemma2-9b-it
+label:
+  zh_Hans: Gemma2 9B Instruct
+  en_US: Gemma2 9B Instruct
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.2'
+  output: '0.2'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/llama-v3-70b-instruct-hf.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/llama-v3-70b-instruct-hf
+label:
+  zh_Hans: Llama3 70B Instruct(HF version)
+  en_US: Llama3 70B Instruct(HF version)
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.9'
+  output: '0.9'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/llama-v3-70b-instruct.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/llama-v3-70b-instruct
+label:
+  zh_Hans: Llama3 70B Instruct
+  en_US: Llama3 70B Instruct
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.9'
+  output: '0.9'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/llama-v3-8b-instruct-hf.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/llama-v3-8b-instruct-hf
+label:
+  zh_Hans: Llama3 8B Instruct(HF version)
+  en_US: Llama3 8B Instruct(HF version)
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.2'
+  output: '0.2'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/llama-v3-8b-instruct.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/llama-v3-8b-instruct
+label:
+  zh_Hans: Llama3 8B Instruct
+  en_US: Llama3 8B Instruct
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.2'
+  output: '0.2'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/llama-v3p1-405b-instruct.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/llama-v3p1-405b-instruct
+label:
+  zh_Hans: Llama3.1 405B Instruct
+  en_US: Llama3.1 405B Instruct
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 131072
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '3'
+  output: '3'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/llama-v3p1-70b-instruct.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/llama-v3p1-70b-instruct
+label:
+  zh_Hans: Llama3.1 70B Instruct
+  en_US: Llama3.1 70B Instruct
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 131072
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.2'
+  output: '0.2'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/llama-v3p1-8b-instruct.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/llama-v3p1-8b-instruct
+label:
+  zh_Hans: Llama3.1 8B Instruct
+  en_US: Llama3.1 8B Instruct
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 131072
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.2'
+  output: '0.2'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/mixtral-8x22b-instruct.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/mixtral-8x22b-instruct
+label:
+  zh_Hans: Mixtral MoE 8x22B Instruct
+  en_US: Mixtral MoE 8x22B Instruct
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 65536
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '1.2'
+  output: '1.2'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/mixtral-8x7b-instruct-hf.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/mixtral-8x7b-instruct-hf
+label:
+  zh_Hans: Mixtral MoE 8x7B Instruct(HF version)
+  en_US: Mixtral MoE 8x7B Instruct(HF version)
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 32768
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.5'
+  output: '0.5'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/mixtral-8x7b-instruct.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/mixtral-8x7b-instruct
+label:
+  zh_Hans: Mixtral MoE 8x7B Instruct
+  en_US: Mixtral MoE 8x7B Instruct
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 32768
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.5'
+  output: '0.5'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/mythomax-l2-13b.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/mythomax-l2-13b
+label:
+  zh_Hans: MythoMax L2 13b
+  en_US: MythoMax L2 13b
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+model_properties:
+  mode: chat
+  context_size: 4096
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.2'
+  output: '0.2'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/phi-3-vision-128k-instruct.yaml

@@ -0,0 +1,46 @@
+model: accounts/fireworks/models/phi-3-vision-128k-instruct
+label:
+  zh_Hans: Phi3.5 Vision Instruct
+  en_US: Phi3.5 Vision Instruct
+model_type: llm
+features:
+  - agent-thought
+  - vision
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '0.2'
+  output: '0.2'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/llm/yi-large.yaml

@@ -0,0 +1,45 @@
+model: accounts/yi-01-ai/models/yi-large
+label:
+  zh_Hans: Yi-Large
+  en_US: Yi-Large
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 32768
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: top_p
+    use_template: top_p
+  - name: top_k
+    label:
+      zh_Hans: 取样数量
+      en_US: Top k
+    type: int
+    help:
+      zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
+      en_US: Only sample from the top K options for each subsequent token.
+  - name: max_tokens
+    use_template: max_tokens
+  - name: context_length_exceeded_behavior
+    default: None
+    label:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    help:
+      zh_Hans: 上下文长度超出行为
+      en_US: Context Length Exceeded Behavior
+    type: string
+    options:
+      - None
+      - truncate
+      - error
+  - name: response_format
+    use_template: response_format
+pricing:
+  input: '3'
+  output: '3'
+  unit: '0.000001'
+  currency: USD

drivers/ai-provider/fireworks/message.go

@@ -0,0 +1,38 @@
+package fireworks
+
+type ClientRequest struct {
+	Messages []*Message `json:"messages"`
+}
+
+type Message struct {
+	Role    string `json:"role"`
+	Content string `json:"content"`
+}
+
+type Response struct {
+	Id                string           `json:"id"`
+	Object            string           `json:"object"`
+	Created           int              `json:"created"`
+	Model             string           `json:"model"`
+	SystemFingerprint string           `json:"system_fingerprint"`
+	Choices           []ResponseChoice `json:"choices"`
+	Usage             Usage            `json:"usage"`
+}
+
+type ResponseChoice struct {
+	Index        int         `json:"index"`
+	Message      Message     `json:"message"`
+	Logprobs     interface{} `json:"logprobs"`
+	FinishReason string      `json:"finish_reason"`
+}
+
+type Usage struct {
+	PromptTokens            int                     `json:"prompt_tokens"`
+	CompletionTokens        int                     `json:"completion_tokens"`
+	TotalTokens             int                     `json:"total_tokens"`
+	CompletionTokensDetails CompletionTokensDetails `json:"completion_tokens_details"`
+}
+
+type CompletionTokensDetails struct {
+	ReasoningTokens int `json:"reasoning_tokens"`
+}

drivers/ai-provider/fireworks/mode.go

@@ -0,0 +1,104 @@
+package fireworks
+
+import (
+	"encoding/json"
+	"github.com/eolinker/apinto/convert"
+	ai_provider "github.com/eolinker/apinto/drivers/ai-provider"
+	"github.com/eolinker/eosc"
+	"github.com/eolinker/eosc/eocontext"
+	http_context "github.com/eolinker/eosc/eocontext/http-context"
+)
+
+var (
+	modelModes = map[string]IModelMode{
+		ai_provider.ModeChat.String(): NewChat(),
+	}
+)
+
+type IModelMode interface {
+	Endpoint() string
+	convert.IConverter
+}
+
+type Chat struct {
+	endPoint string
+}
+
+func NewChat() *Chat {
+	return &Chat{
+		endPoint: "/inference/v1/chat/completions",
+	}
+}
+
+func (c *Chat) Endpoint() string {
+	return c.endPoint
+}
+
+func (c *Chat) RequestConvert(ctx eocontext.EoContext, extender map[string]interface{}) error {
+	httpContext, err := http_context.Assert(ctx)
+	if err != nil {
+		return err
+	}
+	body, err := httpContext.Proxy().Body().RawBody()
+	if err != nil {
+		return err
+	}
+	// 设置转发地址
+	httpContext.Proxy().URI().SetPath(c.endPoint)
+	baseCfg := eosc.NewBase[ai_provider.ClientRequest]()
+	err = json.Unmarshal(body, baseCfg)
+	if err != nil {
+		return err
+	}
+	messages := make([]Message, 0, len(baseCfg.Config.Messages)+1)
+	for _, m := range baseCfg.Config.Messages {
+		messages = append(messages, Message{
+			Role:    m.Role,
+			Content: m.Content,
+		})
+	}
+	baseCfg.SetAppend("messages", messages)
+	for k, v := range extender {
+		baseCfg.SetAppend(k, v)
+	}
+	body, err = json.Marshal(baseCfg)
+	if err != nil {
+		return err
+	}
+	httpContext.Proxy().Body().SetRaw("application/json", body)
+	return nil
+}
+
+func (c *Chat) ResponseConvert(ctx eocontext.EoContext) error {
+	httpContext, err := http_context.Assert(ctx)
+	if err != nil {
+		return err
+	}
+	if httpContext.Response().StatusCode() != 200 {
+		return nil
+	}
+	body := httpContext.Response().GetBody()
+	data := eosc.NewBase[Response]()
+	err = json.Unmarshal(body, data)
+	if err != nil {
+		return err
+	}
+	responseBody := &ai_provider.ClientResponse{}
+	if len(data.Config.Choices) > 0 {
+		msg := data.Config.Choices[0]
+		responseBody.Message = ai_provider.Message{
+			Role:    msg.Message.Role,
+			Content: msg.Message.Content,
+		}
+		responseBody.FinishReason = msg.FinishReason
+	} else {
+		responseBody.Code = -1
+		responseBody.Error = "no response"
+	}
+	body, err = json.Marshal(responseBody)
+	if err != nil {
+		return err
+	}
+	httpContext.Response().SetBody(body)
+	return nil
+}