{status === "success" && chatHistories.length === 0 && (
@@ -244,4 +245,4 @@ export const Sidebar = ({ onClose }: Props) => {
)}
)
-}
+}
\ No newline at end of file
diff --git a/src/components/Sidepanel/Chat/body.tsx b/src/components/Sidepanel/Chat/body.tsx
index 61d90715..f9f14079 100644
--- a/src/components/Sidepanel/Chat/body.tsx
+++ b/src/components/Sidepanel/Chat/body.tsx
@@ -47,6 +47,7 @@ export const SidePanelBody = () => {
setIsSourceOpen(true)
}}
isTTSEnabled={ttsEnabled}
+ generationInfo={message?.generationInfo}
/>
))}
diff --git a/src/components/Sidepanel/Chat/form.tsx b/src/components/Sidepanel/Chat/form.tsx
index e1cc3f01..618431d4 100644
--- a/src/components/Sidepanel/Chat/form.tsx
+++ b/src/components/Sidepanel/Chat/form.tsx
@@ -75,10 +75,10 @@ export const SidepanelForm = ({ dropedFile }: Props) => {
) {
e.preventDefault()
form.onSubmit(async (value) => {
- await stopListening()
- if (value.message.trim().length === 0) {
+ if (value.message.trim().length === 0 && value.image.length === 0) {
return
}
+ await stopListening()
if (!selectedModel || selectedModel.length === 0) {
form.setFieldError("message", t("formError.noModel"))
return
@@ -237,6 +237,9 @@ export const SidepanelForm = ({ dropedFile }: Props) => {
}
}
await stopListening()
+ if (value.message.trim().length === 0 && value.image.length === 0) {
+ return
+ }
form.reset()
textAreaFocus()
await sendMessage({
@@ -260,7 +263,6 @@ export const SidepanelForm = ({ dropedFile }: Props) => {
onKeyDown={(e) => handleKeyDown(e)}
ref={textareaRef}
className="px-2 py-2 w-full resize-none bg-transparent focus-within:outline-none focus:ring-0 focus-visible:ring-0 ring-0 dark:ring-0 border-0 dark:text-gray-100"
- required
onPaste={handlePaste}
rows={1}
style={{ minHeight: "60px" }}
diff --git a/src/db/index.ts b/src/db/index.ts
index 4a775532..da837b29 100644
--- a/src/db/index.ts
+++ b/src/db/index.ts
@@ -33,6 +33,7 @@ type Message = {
search?: WebSearch
createdAt: number
messageType?: string
+ generationInfo?: any
}
type Webshare = {
@@ -254,7 +255,8 @@ export const saveMessage = async (
images: string[],
source?: any[],
time?: number,
- message_type?: string
+ message_type?: string,
+ generationInfo?: any
) => {
const id = generateID()
let createdAt = Date.now()
@@ -270,7 +272,8 @@ export const saveMessage = async (
images,
createdAt,
sources: source,
- messageType: message_type
+ messageType: message_type,
+ generationInfo: generationInfo
}
const db = new PageAssitDatabase()
await db.addMessage(message)
@@ -298,7 +301,8 @@ export const formatToMessage = (messages: MessageHistory): MessageType[] => {
message: message.content,
name: message.name,
sources: message?.sources || [],
- images: message.images || []
+ images: message.images || [],
+ generationInfo: message?.generationInfo,
}
})
}
@@ -351,10 +355,14 @@ export const updateMessageByIndex = async (
index: number,
message: string
) => {
+ try {
const db = new PageAssitDatabase()
const chatHistory = (await db.getChatHistory(history_id)).reverse()
chatHistory[index].content = message
await db.db.set({ [history_id]: chatHistory.reverse() })
+ } catch(e) {
+ // temp chat will break
+ }
}
export const deleteChatForEdit = async (history_id: string, index: number) => {
diff --git a/src/db/models.ts b/src/db/models.ts
index 4575c677..928beb99 100644
--- a/src/db/models.ts
+++ b/src/db/models.ts
@@ -24,6 +24,8 @@ export const removeModelSuffix = (id: string) => {
return id
.replace(/_model-[a-f0-9]{4}-[a-f0-9]{4}-[a-f0-9]{3,4}-[a-f0-9]{4}/, "")
.replace(/_lmstudio_openai-[a-f0-9]{4}-[a-f0-9]{3}-[a-f0-9]{4}/, "")
+ .replace(/_llamafile_openai-[a-f0-9]{4}-[a-f0-9]{3}-[a-f0-9]{4}/, "")
+ .replace(/_ollama2_openai-[a-f0-9]{4}-[a-f0-9]{3}-[a-f0-9]{4}/, "")
}
export const isLMStudioModel = (model: string) => {
const lmstudioModelRegex =
@@ -31,6 +33,16 @@ export const isLMStudioModel = (model: string) => {
return lmstudioModelRegex.test(model)
}
+export const isLlamafileModel = (model: string) => {
+ const llamafileModelRegex =
+ /_llamafile_openai-[a-f0-9]{4}-[a-f0-9]{3}-[a-f0-9]{4}/
+ return llamafileModelRegex.test(model)
+}
+export const isOllamaModel = (model: string) => {
+ const ollamaModelRegex =
+ /_ollama2_openai-[a-f0-9]{4}-[a-f0-9]{3}-[a-f0-9]{4}/
+ return ollamaModelRegex.test(model)
+}
export const getLMStudioModelId = (
model: string
): { model_id: string; provider_id: string } => {
@@ -44,10 +56,45 @@ export const getLMStudioModelId = (
}
return null
}
+export const getOllamaModelId = (
+ model: string
+): { model_id: string; provider_id: string } => {
+ const ollamaModelRegex =
+ /_ollama2_openai-[a-f0-9]{4}-[a-f0-9]{3}-[a-f0-9]{4}/
+ const match = model.match(ollamaModelRegex)
+ if (match) {
+ const modelId = match[0]
+ const providerId = match[0].replace("_ollama2_openai-", "")
+ return { model_id: modelId, provider_id: providerId }
+ }
+ return null
+}
+export const getLlamafileModelId = (
+ model: string
+): { model_id: string; provider_id: string } => {
+ const llamafileModelRegex =
+ /_llamafile_openai-[a-f0-9]{4}-[a-f0-9]{3}-[a-f0-9]{4}/
+ const match = model.match(llamafileModelRegex)
+ if (match) {
+ const modelId = match[0]
+ const providerId = match[0].replace("_llamafile_openai-", "")
+ return { model_id: modelId, provider_id: providerId }
+ }
+ return null
+}
export const isCustomModel = (model: string) => {
if (isLMStudioModel(model)) {
return true
}
+
+ if (isLlamafileModel(model)) {
+ return true
+ }
+
+ if (isOllamaModel(model)) {
+ return true
+ }
+
const customModelRegex =
/_model-[a-f0-9]{4}-[a-f0-9]{4}-[a-f0-9]{3,4}-[a-f0-9]{4}/
return customModelRegex.test(model)
@@ -201,6 +248,44 @@ export const getModelInfo = async (id: string) => {
}
}
+
+ if (isLlamafileModel(id)) {
+ const llamafileId = getLlamafileModelId(id)
+ if (!llamafileId) {
+ throw new Error("Invalid LMStudio model ID")
+ }
+ return {
+ model_id: id.replace(
+ /_llamafile_openai-[a-f0-9]{4}-[a-f0-9]{3}-[a-f0-9]{4}/,
+ ""
+ ),
+ provider_id: `openai-${llamafileId.provider_id}`,
+ name: id.replace(
+ /_llamafile_openai-[a-f0-9]{4}-[a-f0-9]{3}-[a-f0-9]{4}/,
+ ""
+ )
+ }
+ }
+
+
+ if (isOllamaModel(id)) {
+ const ollamaId = getOllamaModelId(id)
+ if (!ollamaId) {
+ throw new Error("Invalid LMStudio model ID")
+ }
+ return {
+ model_id: id.replace(
+ /_ollama2_openai-[a-f0-9]{4}-[a-f0-9]{3}-[a-f0-9]{4}/,
+ ""
+ ),
+ provider_id: `openai-${ollamaId.provider_id}`,
+ name: id.replace(
+ /_ollama2_openai-[a-f0-9]{4}-[a-f0-9]{3}-[a-f0-9]{4}/,
+ ""
+ )
+ }
+ }
+
const model = await db.getById(id)
return model
}
@@ -264,6 +349,48 @@ export const dynamicFetchLMStudio = async ({
return lmstudioModels
}
+export const dynamicFetchOllama2 = async ({
+ baseUrl,
+ providerId
+}: {
+ baseUrl: string
+ providerId: string
+}) => {
+ const models = await getAllOpenAIModels(baseUrl)
+ const ollama2Models = models.map((e) => {
+ return {
+ name: e?.name || e?.id,
+ id: `${e?.id}_ollama2_${providerId}`,
+ provider: providerId,
+ lookup: `${e?.id}_${providerId}`,
+ provider_id: providerId
+ }
+ })
+
+ return ollama2Models
+}
+
+export const dynamicFetchLlamafile = async ({
+ baseUrl,
+ providerId
+}: {
+ baseUrl: string
+ providerId: string
+}) => {
+ const models = await getAllOpenAIModels(baseUrl)
+ const llamafileModels = models.map((e) => {
+ return {
+ name: e?.name || e?.id,
+ id: `${e?.id}_llamafile_${providerId}`,
+ provider: providerId,
+ lookup: `${e?.id}_${providerId}`,
+ provider_id: providerId
+ }
+ })
+
+ return llamafileModels
+}
+
export const ollamaFormatAllCustomModels = async (
modelType: "all" | "chat" | "embedding" = "all"
) => {
@@ -276,6 +403,14 @@ export const ollamaFormatAllCustomModels = async (
(provider) => provider.provider === "lmstudio"
)
+ const llamafileProviders = allProviders.filter(
+ (provider) => provider.provider === "llamafile"
+ )
+
+ const ollamaProviders = allProviders.filter(
+ (provider) => provider.provider === "ollama2"
+ )
+
const lmModelsPromises = lmstudioProviders.map((provider) =>
dynamicFetchLMStudio({
baseUrl: provider.baseUrl,
@@ -283,16 +418,39 @@ export const ollamaFormatAllCustomModels = async (
})
)
+ const llamafileModelsPromises = llamafileProviders.map((provider) =>
+ dynamicFetchLlamafile({
+ baseUrl: provider.baseUrl,
+ providerId: provider.id
+ })
+ )
+
+ const ollamaModelsPromises = ollamaProviders.map((provider) =>
+ dynamicFetchOllama2({
+ baseUrl: provider.baseUrl,
+ providerId: provider.id
+ }))
+
const lmModelsFetch = await Promise.all(lmModelsPromises)
+ const llamafileModelsFetch = await Promise.all(llamafileModelsPromises)
+
+ const ollamaModelsFetch = await Promise.all(ollamaModelsPromises)
+
const lmModels = lmModelsFetch.flat()
+ const llamafileModels = llamafileModelsFetch.flat()
+
+ const ollama2Models = ollamaModelsFetch.flat()
+
// merge allModels and lmModels
const allModlesWithLMStudio = [
...(modelType !== "all"
? allModles.filter((model) => model.model_type === modelType)
: allModles),
- ...lmModels
+ ...lmModels,
+ ...llamafileModels,
+ ...ollama2Models
]
const ollamaModels = allModlesWithLMStudio.map((model) => {
diff --git a/src/hooks/chat-helper/index.ts b/src/hooks/chat-helper/index.ts
index 18f89bea..3f725c02 100644
--- a/src/hooks/chat-helper/index.ts
+++ b/src/hooks/chat-helper/index.ts
@@ -118,7 +118,7 @@ export const saveMessageOnSuccess = async ({
fullText,
source,
message_source = "web-ui",
- message_type
+ message_type, generationInfo
}: {
historyId: string | null
setHistoryId: (historyId: string) => void
@@ -130,6 +130,7 @@ export const saveMessageOnSuccess = async ({
source: any[]
message_source?: "copilot" | "web-ui",
message_type?: string
+ generationInfo?: any
}) => {
if (historyId) {
if (!isRegenerate) {
@@ -141,7 +142,8 @@ export const saveMessageOnSuccess = async ({
[image],
[],
1,
- message_type
+ message_type,
+ generationInfo
)
}
await saveMessage(
@@ -152,7 +154,8 @@ export const saveMessageOnSuccess = async ({
[],
source,
2,
- message_type
+ message_type,
+ generationInfo
)
await setLastUsedChatModel(historyId, selectedModel!)
} else {
@@ -166,7 +169,8 @@ export const saveMessageOnSuccess = async ({
[image],
[],
1,
- message_type
+ message_type,
+ generationInfo
)
await saveMessage(
newHistoryId.id,
@@ -176,7 +180,8 @@ export const saveMessageOnSuccess = async ({
[],
source,
2,
- message_type
+ message_type,
+ generationInfo
)
setHistoryId(newHistoryId.id)
await setLastUsedChatModel(newHistoryId.id, selectedModel!)
diff --git a/src/hooks/useMessage.tsx b/src/hooks/useMessage.tsx
index c85324a9..d06102db 100644
--- a/src/hooks/useMessage.tsx
+++ b/src/hooks/useMessage.tsx
@@ -133,7 +133,9 @@ export const useMessage = () => {
currentChatModelSettings?.numCtx ?? userDefaultModelSettings?.numCtx,
seed: currentChatModelSettings?.seed,
numGpu:
- currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu
+ currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu,
+ numPredict: currentChatModelSettings?.numPredict ?? userDefaultModelSettings?.numPredict,
+
})
let newMessage: Message[] = []
@@ -261,7 +263,9 @@ export const useMessage = () => {
userDefaultModelSettings?.numCtx,
seed: currentChatModelSettings?.seed,
numGpu:
- currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu
+ currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu,
+ numPredict: currentChatModelSettings?.numPredict ?? userDefaultModelSettings?.numPredict,
+
})
const response = await questionOllama.invoke(promptForQuestion)
query = response.content.toString()
@@ -328,16 +332,31 @@ export const useMessage = () => {
const applicationChatHistory = generateHistory(history, selectedModel)
+ let generationInfo: any | undefined = undefined
+
const chunks = await ollama.stream(
[...applicationChatHistory, humanMessage],
{
- signal: signal
+ signal: signal,
+ callbacks: [
+ {
+ handleLLMEnd(
+ output: any,
+ ): any {
+ try {
+ generationInfo = output?.generations?.[0][0]?.generationInfo
+ } catch (e) {
+ console.log("handleLLMEnd error", e)
+ }
+ }
+ }
+ ]
}
)
let count = 0
for await (const chunk of chunks) {
- contentToSave += chunk.content
- fullText += chunk.content
+ contentToSave += chunk?.content
+ fullText += chunk?.content
if (count === 0) {
setIsProcessing(true)
}
@@ -361,7 +380,8 @@ export const useMessage = () => {
return {
...message,
message: fullText,
- sources: source
+ sources: source,
+ generationInfo
}
}
return message
@@ -390,7 +410,8 @@ export const useMessage = () => {
image,
fullText,
source,
- message_source: "copilot"
+ message_source: "copilot",
+ generationInfo
})
setIsProcessing(false)
@@ -458,7 +479,9 @@ export const useMessage = () => {
currentChatModelSettings?.numCtx ?? userDefaultModelSettings?.numCtx,
seed: currentChatModelSettings?.seed,
numGpu:
- currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu
+ currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu,
+ numPredict: currentChatModelSettings?.numPredict ?? userDefaultModelSettings?.numPredict,
+
})
let newMessage: Message[] = []
@@ -544,16 +567,31 @@ export const useMessage = () => {
)
}
+ let generationInfo: any | undefined = undefined
+
const chunks = await ollama.stream(
[...applicationChatHistory, humanMessage],
{
- signal: signal
+ signal: signal,
+ callbacks: [
+ {
+ handleLLMEnd(
+ output: any,
+ ): any {
+ try {
+ generationInfo = output?.generations?.[0][0]?.generationInfo
+ } catch (e) {
+ console.log("handleLLMEnd error", e)
+ }
+ }
+ }
+ ]
}
)
let count = 0
for await (const chunk of chunks) {
- contentToSave += chunk.content
- fullText += chunk.content
+ contentToSave += chunk?.content
+ fullText += chunk?.content
if (count === 0) {
setIsProcessing(true)
}
@@ -576,7 +614,8 @@ export const useMessage = () => {
if (message.id === generateMessageId) {
return {
...message,
- message: fullText
+ message: fullText,
+ generationInfo
}
}
return message
@@ -605,7 +644,8 @@ export const useMessage = () => {
image,
fullText,
source: [],
- message_source: "copilot"
+ message_source: "copilot",
+ generationInfo
})
setIsProcessing(false)
@@ -668,7 +708,9 @@ export const useMessage = () => {
currentChatModelSettings?.numCtx ?? userDefaultModelSettings?.numCtx,
seed: currentChatModelSettings?.seed,
numGpu:
- currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu
+ currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu,
+ numPredict: currentChatModelSettings?.numPredict ?? userDefaultModelSettings?.numPredict,
+
})
let newMessage: Message[] = []
@@ -743,7 +785,9 @@ export const useMessage = () => {
userDefaultModelSettings?.numCtx,
seed: currentChatModelSettings?.seed,
numGpu:
- currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu
+ currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu,
+ numPredict: currentChatModelSettings?.numPredict ?? userDefaultModelSettings?.numPredict,
+
})
const response = await questionOllama.invoke(promptForQuestion)
query = response.content.toString()
@@ -789,16 +833,30 @@ export const useMessage = () => {
)
}
+ let generationInfo: any | undefined = undefined
const chunks = await ollama.stream(
[...applicationChatHistory, humanMessage],
{
- signal: signal
+ signal: signal,
+ callbacks: [
+ {
+ handleLLMEnd(
+ output: any,
+ ): any {
+ try {
+ generationInfo = output?.generations?.[0][0]?.generationInfo
+ } catch (e) {
+ console.log("handleLLMEnd error", e)
+ }
+ }
+ }
+ ]
}
)
let count = 0
for await (const chunk of chunks) {
- contentToSave += chunk.content
- fullText += chunk.content
+ contentToSave += chunk?.content
+ fullText += chunk?.content
if (count === 0) {
setIsProcessing(true)
}
@@ -822,7 +880,8 @@ export const useMessage = () => {
return {
...message,
message: fullText,
- sources: source
+ sources: source,
+ generationInfo
}
}
return message
@@ -850,7 +909,8 @@ export const useMessage = () => {
message,
image,
fullText,
- source
+ source,
+ generationInfo
})
setIsProcessing(false)
@@ -914,7 +974,9 @@ export const useMessage = () => {
currentChatModelSettings?.numCtx ?? userDefaultModelSettings?.numCtx,
seed: currentChatModelSettings?.seed,
numGpu:
- currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu
+ currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu,
+ numPredict: currentChatModelSettings?.numPredict ?? userDefaultModelSettings?.numPredict,
+
})
let newMessage: Message[] = []
@@ -982,13 +1044,28 @@ export const useMessage = () => {
})
}
+ let generationInfo: any | undefined = undefined
+
const chunks = await ollama.stream([humanMessage], {
- signal: signal
+ signal: signal,
+ callbacks: [
+ {
+ handleLLMEnd(
+ output: any,
+ ): any {
+ try {
+ generationInfo = output?.generations?.[0][0]?.generationInfo
+ } catch (e) {
+ console.log("handleLLMEnd error", e)
+ }
+ }
+ }
+ ]
})
let count = 0
for await (const chunk of chunks) {
- contentToSave += chunk.content
- fullText += chunk.content
+ contentToSave += chunk?.content
+ fullText += chunk?.content
if (count === 0) {
setIsProcessing(true)
}
@@ -1011,7 +1088,8 @@ export const useMessage = () => {
if (message.id === generateMessageId) {
return {
...message,
- message: fullText
+ message: fullText,
+ generationInfo
}
}
return message
@@ -1042,7 +1120,8 @@ export const useMessage = () => {
fullText,
source: [],
message_source: "copilot",
- message_type: messageType
+ message_type: messageType,
+ generationInfo
})
setIsProcessing(false)
diff --git a/src/hooks/useMessageOption.tsx b/src/hooks/useMessageOption.tsx
index 58d56bfc..322287c4 100644
--- a/src/hooks/useMessageOption.tsx
+++ b/src/hooks/useMessageOption.tsx
@@ -22,7 +22,10 @@ import { notification } from "antd"
import { getSystemPromptForWeb } from "~/web/web"
import { generateHistory } from "@/utils/generate-history"
import { useTranslation } from "react-i18next"
-import { saveMessageOnError, saveMessageOnSuccess } from "./chat-helper"
+import {
+ saveMessageOnError as saveError,
+ saveMessageOnSuccess as saveSuccess
+} from "./chat-helper"
import { usePageAssist } from "@/context"
import { PageAssistVectorStore } from "@/libs/PageAssistVectorStore"
import { formatDocs } from "@/chain/chat-with-x"
@@ -65,7 +68,9 @@ export const useMessageOption = () => {
selectedSystemPrompt,
setSelectedSystemPrompt,
selectedKnowledge,
- setSelectedKnowledge
+ setSelectedKnowledge,
+ temporaryChat,
+ setTemporaryChat
} = useStoreMessageOption()
const currentChatModelSettings = useStoreChatModelSettings()
const [selectedModel, setSelectedModel] = useStorage("selectedModel")
@@ -122,7 +127,10 @@ export const useMessageOption = () => {
currentChatModelSettings?.numCtx ?? userDefaultModelSettings?.numCtx,
seed: currentChatModelSettings?.seed,
numGpu:
- currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu
+ currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu,
+ numPredict:
+ currentChatModelSettings?.numPredict ??
+ userDefaultModelSettings?.numPredict
})
let newMessage: Message[] = []
@@ -197,7 +205,11 @@ export const useMessageOption = () => {
userDefaultModelSettings?.numCtx,
seed: currentChatModelSettings?.seed,
numGpu:
- currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu
+ currentChatModelSettings?.numGpu ??
+ userDefaultModelSettings?.numGpu,
+ numPredict:
+ currentChatModelSettings?.numPredict ??
+ userDefaultModelSettings?.numPredict
})
const response = await questionOllama.invoke(promptForQuestion)
query = response.content.toString()
@@ -243,16 +255,29 @@ export const useMessageOption = () => {
)
}
+ let generationInfo: any | undefined = undefined
+
const chunks = await ollama.stream(
[...applicationChatHistory, humanMessage],
{
- signal: signal
+ signal: signal,
+ callbacks: [
+ {
+ handleLLMEnd(output: any): any {
+ try {
+ generationInfo = output?.generations?.[0][0]?.generationInfo
+ } catch (e) {
+ console.log("handleLLMEnd error", e)
+ }
+ }
+ }
+ ]
}
)
let count = 0
for await (const chunk of chunks) {
- contentToSave += chunk.content
- fullText += chunk.content
+ contentToSave += chunk?.content
+ fullText += chunk?.content
if (count === 0) {
setIsProcessing(true)
}
@@ -276,7 +301,8 @@ export const useMessageOption = () => {
return {
...message,
message: fullText,
- sources: source
+ sources: source,
+ generationInfo
}
}
return message
@@ -304,7 +330,8 @@ export const useMessageOption = () => {
message,
image,
fullText,
- source
+ source,
+ generationInfo
})
setIsProcessing(false)
@@ -336,6 +363,39 @@ export const useMessageOption = () => {
}
}
+ const saveMessageOnSuccess = async (e: any) => {
+ if (!temporaryChat) {
+ return await saveSuccess(e)
+ } else {
+ setHistoryId("temp")
+ }
+
+ return true
+ }
+
+ const saveMessageOnError = async (e: any) => {
+ if (!temporaryChat) {
+ return await saveError(e)
+ } else {
+ setHistory([
+ ...history,
+ {
+ role: "user",
+ content: e.userMessage,
+ image: e.image
+ },
+ {
+ role: "assistant",
+ content: e.botMessage
+ }
+ ])
+
+ setHistoryId("temp")
+ }
+
+ return true
+ }
+
const normalChatMode = async (
message: string,
image: string,
@@ -366,7 +426,10 @@ export const useMessageOption = () => {
currentChatModelSettings?.numCtx ?? userDefaultModelSettings?.numCtx,
seed: currentChatModelSettings?.seed,
numGpu:
- currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu
+ currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu,
+ numPredict:
+ currentChatModelSettings?.numPredict ??
+ userDefaultModelSettings?.numPredict
})
let newMessage: Message[] = []
@@ -465,17 +528,30 @@ export const useMessageOption = () => {
)
}
+ let generationInfo: any | undefined = undefined
+
const chunks = await ollama.stream(
[...applicationChatHistory, humanMessage],
{
- signal: signal
+ signal: signal,
+ callbacks: [
+ {
+ handleLLMEnd(output: any): any {
+ try {
+ generationInfo = output?.generations?.[0][0]?.generationInfo
+ } catch (e) {
+ console.log("handleLLMEnd error", e)
+ }
+ }
+ }
+ ]
}
)
let count = 0
for await (const chunk of chunks) {
- contentToSave += chunk.content
- fullText += chunk.content
+ contentToSave += chunk?.content
+ fullText += chunk?.content
if (count === 0) {
setIsProcessing(true)
}
@@ -498,7 +574,8 @@ export const useMessageOption = () => {
if (message.id === generateMessageId) {
return {
...message,
- message: fullText
+ message: fullText,
+ generationInfo
}
}
return message
@@ -526,7 +603,8 @@ export const useMessageOption = () => {
message,
image,
fullText,
- source: []
+ source: [],
+ generationInfo
})
setIsProcessing(false)
@@ -586,7 +664,10 @@ export const useMessageOption = () => {
currentChatModelSettings?.numCtx ?? userDefaultModelSettings?.numCtx,
seed: currentChatModelSettings?.seed,
numGpu:
- currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu
+ currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu,
+ numPredict:
+ currentChatModelSettings?.numPredict ??
+ userDefaultModelSettings?.numPredict
})
let newMessage: Message[] = []
@@ -677,7 +758,11 @@ export const useMessageOption = () => {
userDefaultModelSettings?.numCtx,
seed: currentChatModelSettings?.seed,
numGpu:
- currentChatModelSettings?.numGpu ?? userDefaultModelSettings?.numGpu
+ currentChatModelSettings?.numGpu ??
+ userDefaultModelSettings?.numGpu,
+ numPredict:
+ currentChatModelSettings?.numPredict ??
+ userDefaultModelSettings?.numPredict
})
const response = await questionOllama.invoke(promptForQuestion)
query = response.content.toString()
@@ -711,16 +796,29 @@ export const useMessageOption = () => {
const applicationChatHistory = generateHistory(history, selectedModel)
+ let generationInfo: any | undefined = undefined
+
const chunks = await ollama.stream(
[...applicationChatHistory, humanMessage],
{
- signal: signal
+ signal: signal,
+ callbacks: [
+ {
+ handleLLMEnd(output: any): any {
+ try {
+ generationInfo = output?.generations?.[0][0]?.generationInfo
+ } catch (e) {
+ console.log("handleLLMEnd error", e)
+ }
+ }
+ }
+ ]
}
)
let count = 0
for await (const chunk of chunks) {
- contentToSave += chunk.content
- fullText += chunk.content
+ contentToSave += chunk?.content
+ fullText += chunk?.content
if (count === 0) {
setIsProcessing(true)
}
@@ -744,7 +842,8 @@ export const useMessageOption = () => {
return {
...message,
message: fullText,
- sources: source
+ sources: source,
+ generationInfo
}
}
return message
@@ -772,7 +871,8 @@ export const useMessageOption = () => {
message,
image,
fullText,
- source
+ source,
+ generationInfo
})
setIsProcessing(false)
@@ -984,6 +1084,8 @@ export const useMessageOption = () => {
textareaRef,
selectedKnowledge,
setSelectedKnowledge,
- ttsEnabled
+ ttsEnabled,
+ temporaryChat,
+ setTemporaryChat
}
}
diff --git a/src/i18n/index.ts b/src/i18n/index.ts
index 97bbf6f8..effddb4a 100644
--- a/src/i18n/index.ts
+++ b/src/i18n/index.ts
@@ -14,6 +14,7 @@ import { de } from "./lang/de";
import { da } from "./lang/da";
import { no } from "./lang/no";
import { sv } from "./lang/sv";
+import { ko } from "./lang/ko";
i18n
@@ -37,7 +38,8 @@ i18n
da: da,
no: no,
de: de,
- sv: sv
+ sv: sv,
+ ko: ko,
},
fallbackLng: "en",
lng: localStorage.getItem("i18nextLng") || "en",
diff --git a/src/i18n/lang/ko.ts b/src/i18n/lang/ko.ts
new file mode 100644
index 00000000..037ce8f2
--- /dev/null
+++ b/src/i18n/lang/ko.ts
@@ -0,0 +1,19 @@
+import option from "@/assets/locale/ko/option.json";
+import playground from "@/assets/locale/ko/playground.json";
+import common from "@/assets/locale/ko/common.json";
+import sidepanel from "@/assets/locale/ko/sidepanel.json";
+import settings from "@/assets/locale/ko/settings.json";
+import knowledge from "@/assets/locale/ko/knowledge.json";
+import chrome from "@/assets/locale/ko/chrome.json";
+import openai from "@/assets/locale/ko/openai.json";
+
+export const ko = {
+ option,
+ playground,
+ common,
+ sidepanel,
+ settings,
+ knowledge,
+ chrome,
+ openai
+}
\ No newline at end of file
diff --git a/src/i18n/support-language.ts b/src/i18n/support-language.ts
index 2c62ad40..273964d9 100644
--- a/src/i18n/support-language.ts
+++ b/src/i18n/support-language.ts
@@ -55,5 +55,9 @@ export const supportLanguage = [
{
value: "sv",
label: "Svenska"
+ },
+ {
+ value: "ko",
+ label: "한국어"
}
]
diff --git a/src/models/embedding.ts b/src/models/embedding.ts
index 03eb663a..0c1026b6 100644
--- a/src/models/embedding.ts
+++ b/src/models/embedding.ts
@@ -1,4 +1,4 @@
-import { getModelInfo, isCustomModel } from "@/db/models"
+import { getModelInfo, isCustomModel, isOllamaModel } from "@/db/models"
import { OllamaEmbeddingsPageAssist } from "./OllamaEmbedding"
import { OAIEmbedding } from "./OAIEmbedding"
import { getOpenAIConfigById } from "@/db/openai"
diff --git a/src/models/index.ts b/src/models/index.ts
index d459e664..135025fa 100644
--- a/src/models/index.ts
+++ b/src/models/index.ts
@@ -1,8 +1,9 @@
-import { getModelInfo, isCustomModel } from "@/db/models"
+import { getModelInfo, isCustomModel, isOllamaModel } from "@/db/models"
import { ChatChromeAI } from "./ChatChromeAi"
import { ChatOllama } from "./ChatOllama"
import { getOpenAIConfigById } from "@/db/openai"
import { ChatOpenAI } from "@langchain/openai"
+import { urlRewriteRuntime } from "@/libs/runtime"
export const pageAssistModel = async ({
model,
@@ -13,7 +14,8 @@ export const pageAssistModel = async ({
topP,
numCtx,
seed,
- numGpu
+ numGpu,
+ numPredict,
}: {
model: string
baseUrl: string
@@ -24,12 +26,13 @@ export const pageAssistModel = async ({
numCtx?: number
seed?: number
numGpu?: number
+ numPredict?: number
}) => {
if (model === "chrome::gemini-nano::page-assist") {
return new ChatChromeAI({
temperature,
- topK
+ topK,
})
}
@@ -41,15 +44,20 @@ export const pageAssistModel = async ({
const modelInfo = await getModelInfo(model)
const providerInfo = await getOpenAIConfigById(modelInfo.provider_id)
+ if (isOllamaModel(model)) {
+ await urlRewriteRuntime(providerInfo.baseUrl || "")
+ }
+
return new ChatOpenAI({
modelName: modelInfo.model_id,
openAIApiKey: providerInfo.apiKey || "temp",
temperature,
topP,
+ maxTokens: numPredict,
configuration: {
apiKey: providerInfo.apiKey || "temp",
baseURL: providerInfo.baseUrl || "",
- }
+ },
}) as any
}
@@ -64,7 +72,8 @@ export const pageAssistModel = async ({
numCtx,
seed,
model,
- numGpu
+ numGpu,
+ numPredict
})
diff --git a/src/models/utils/chrome.ts b/src/models/utils/chrome.ts
index 16e0b8b8..3d426863 100644
--- a/src/models/utils/chrome.ts
+++ b/src/models/utils/chrome.ts
@@ -3,13 +3,19 @@ export const checkChromeAIAvailability = async (): Promise<"readily" | "no" | "a
try {
const ai = (window as any).ai;
- // upcoming version change
+ // latest i guess
+ if (ai?.languageModel?.capabilities) {
+ const capabilities = await ai.languageModel.capabilities();
+ return capabilities?.available ?? "no";
+ }
+
+ // old version change
if (ai?.assistant?.capabilities) {
const capabilities = await ai.assistant.capabilities();
return capabilities?.available ?? "no";
}
- // old version
+ // too old version
if (ai?.canCreateTextSession) {
const available = await ai.canCreateTextSession();
return available ?? "no";
@@ -33,7 +39,15 @@ export interface AITextSession {
export const createAITextSession = async (data: any): Promise
=> {
const ai = (window as any).ai;
- // upcoming version change
+ // new version i guess
+ if (ai?.languageModel?.create) {
+ const session = await ai.languageModel.create({
+ ...data
+ })
+ return session
+ }
+
+ // old version change
if (ai?.assistant?.create) {
const session = await ai.assistant.create({
...data
@@ -41,7 +55,7 @@ export const createAITextSession = async (data: any): Promise =>
return session
}
- // old version
+ // too old version
if (ai.createTextSession) {
const session = await ai.createTextSession({
...data
diff --git a/src/store/option.tsx b/src/store/option.tsx
index 8a55501e..49911cd3 100644
--- a/src/store/option.tsx
+++ b/src/store/option.tsx
@@ -65,6 +65,9 @@ type State = {
setSpeechToTextLanguage: (language: string) => void
speechToTextLanguage: string
+
+ temporaryChat: boolean
+ setTemporaryChat: (temporaryChat: boolean) => void
}
export const useStoreMessageOption = create((set) => ({
@@ -102,5 +105,8 @@ export const useStoreMessageOption = create((set) => ({
setSelectedQuickPrompt: (selectedQuickPrompt) => set({ selectedQuickPrompt }),
selectedKnowledge: null,
- setSelectedKnowledge: (selectedKnowledge) => set({ selectedKnowledge })
+ setSelectedKnowledge: (selectedKnowledge) => set({ selectedKnowledge }),
+
+ temporaryChat: false,
+ setTemporaryChat: (temporaryChat) => set({ temporaryChat }),
}))
diff --git a/src/types/message.ts b/src/types/message.ts
index 3be7cdc0..1ec1d5ae 100644
--- a/src/types/message.ts
+++ b/src/types/message.ts
@@ -1,19 +1,20 @@
type WebSearch = {
- search_engine: string
- search_url: string
- search_query: string
- search_results: {
- title: string
- link: string
- }[]
- }
- export type Message = {
- isBot: boolean
- name: string
- message: string
- sources: any[]
- images?: string[]
- search?: WebSearch
- messageType?: string
- id?: string
- }
\ No newline at end of file
+ search_engine: string
+ search_url: string
+ search_query: string
+ search_results: {
+ title: string
+ link: string
+ }[]
+}
+export type Message = {
+ isBot: boolean
+ name: string
+ message: string
+ sources: any[]
+ images?: string[]
+ search?: WebSearch
+ messageType?: string
+ id?: string
+ generationInfo?: any
+}
\ No newline at end of file
diff --git a/src/utils/human-message.tsx b/src/utils/human-message.tsx
index 67123394..9c1200fd 100644
--- a/src/utils/human-message.tsx
+++ b/src/utils/human-message.tsx
@@ -1,43 +1,46 @@
import { isCustomModel } from "@/db/models"
import { HumanMessage, type MessageContent } from "@langchain/core/messages"
-
type HumanMessageType = {
- content: MessageContent,
- model: string
+ content: MessageContent
+ model: string
}
export const humanMessageFormatter = ({ content, model }: HumanMessageType) => {
-
- const isCustom = isCustomModel(model)
-
- if(isCustom) {
- if(typeof content !== 'string') {
- if(content.length > 1) {
- // this means that we need to reformat the image_url
- const newContent: MessageContent = [
- {
- type: "text",
- //@ts-ignore
- text: content[0].text
- },
- {
- type: "image_url",
- image_url: {
- //@ts-ignore
- url: content[1].image_url
- }
- }
- ]
+ const isCustom = isCustomModel(model)
- return new HumanMessage({
- content: newContent
- })
+ if (isCustom) {
+ if (typeof content !== "string") {
+ if (content.length > 1) {
+ // this means that we need to reformat the image_url
+ const newContent: MessageContent = [
+ {
+ type: "text",
+ //@ts-ignore
+ text: content[0].text
+ },
+ {
+ type: "image_url",
+ image_url: {
+ //@ts-ignore
+ url: content[1].image_url
}
- }
+ }
+ ]
+
+ return new HumanMessage({
+ content: newContent
+ })
+ } else {
+ return new HumanMessage({
+ //@ts-ignore
+ content: content[0].text
+ })
+ }
}
-
- return new HumanMessage({
- content,
- })
-}
\ No newline at end of file
+ }
+
+ return new HumanMessage({
+ content
+ })
+}
diff --git a/src/utils/oai-api-providers.ts b/src/utils/oai-api-providers.ts
index 5d651053..40c1e21f 100644
--- a/src/utils/oai-api-providers.ts
+++ b/src/utils/oai-api-providers.ts
@@ -1,9 +1,24 @@
export const OAI_API_PROVIDERS = [
+ {
+ label: "Custom",
+ value: "custom",
+ baseUrl: ""
+ },
{
label: "LM Studio",
value: "lmstudio",
baseUrl: "http://localhost:1234/v1"
},
+ {
+ label: "Llamafile",
+ value: "llamafile",
+ baseUrl: "http://127.0.0.1:8080/v1"
+ },
+ {
+ label: "Ollama",
+ value: "ollama2",
+ baseUrl: "http://localhost:11434/v1"
+ },
{
label: "OpenAI",
value: "openai",
@@ -29,9 +44,5 @@ export const OAI_API_PROVIDERS = [
value: "openrouter",
baseUrl: "https://openrouter.ai/api/v1"
},
- {
- label: "Custom",
- value: "custom",
- baseUrl: ""
- }
+
]
\ No newline at end of file
diff --git a/wxt.config.ts b/wxt.config.ts
index c870e8e4..288a5553 100644
--- a/wxt.config.ts
+++ b/wxt.config.ts
@@ -50,7 +50,7 @@ export default defineConfig({
outDir: "build",
manifest: {
- version: "1.3.3",
+ version: "1.3.4",
name:
process.env.TARGET === "firefox"
? "Page Assist - A Web UI for Local AI Models"