feat: added conversation history

ROADMAP.md

@@ -1,4 +1,5 @@
 # Roadmap
-- [ ] add support for conversation history
-- [ ] add support for setting a model for a conversation
-- [ ] add support for multiple conversations
+- [x] add support for conversation history
+- [x] add support for setting a model for a conversation
+- [ ] add support for multiple conversations
+- [ ] implement stopping a stream

adapters/outbound/ollama_llm_service.go

@@ -4,15 +4,17 @@ import (
 	cont "context"
 	"errors"
 
+	"github.com/ChristianSch/Theta/domain/models"
 	"github.com/ChristianSch/Theta/domain/ports/outbound"
 	"github.com/jmorganca/ollama/api"
 	"github.com/tmc/langchaingo/llms"
 	"github.com/tmc/langchaingo/llms/ollama"
+	"github.com/tmc/langchaingo/schema"
 )
 
 type OllamaLlmService struct {
 	client *api.Client
-	llm    *ollama.LLM
+	llm    *ollama.Chat
 	model  *string
 	log    outbound.Log
 }
@@ -52,7 +54,7 @@ func (s *OllamaLlmService) ListModels() ([]string, error) {
 func (s *OllamaLlmService) SetModel(model string) error {
 	s.model = &model
 
-	llm, err := ollama.New(ollama.WithModel(model))
+	llm, err := ollama.NewChat(ollama.WithLLMOptions(ollama.WithModel(model)))
 	if err != nil {
 		return err
 	}
@@ -63,15 +65,29 @@ func (s *OllamaLlmService) SetModel(model string) error {
 	return nil
 }
 
-func (s *OllamaLlmService) SendMessage(prompt string, context []string, resHandler outbound.ResponseHandler) error {
+func (s *OllamaLlmService) SendMessage(prompt string, context []models.Message, resHandler outbound.ResponseHandler) error {
 	if s.llm == nil {
 		return errors.New(ModelNotSetError)
 	}
 
-	ctx := cont.Background()
+	var messages []schema.ChatMessage
+
+	for _, msg := range context {
+		if msg.Type == models.UserMessage {
+			messages = append(messages, schema.HumanChatMessage{
+				Content: msg.Text,
+			})
+		} else {
+			messages = append(messages, schema.AIChatMessage{
+				Content: msg.Text,
+			})
+		}
+	}
+
+	messages = append(messages, schema.HumanChatMessage{
+		Content: prompt,
+	})
 
-	_, err := s.llm.Call(ctx, prompt,
-		llms.WithStreamingFunc(resHandler),
-	)
+	_, err := s.llm.Call(cont.Background(), messages, llms.WithStreamingFunc(resHandler))
 	return err
 }

adapters/outbound/repo/in_memory_conversation_repo.go

@@ -0,0 +1,68 @@
+package repo
+
+import (
+	"errors"
+	"sync"
+
+	"github.com/ChristianSch/Theta/domain/models"
+	"github.com/google/uuid"
+)
+
+const (
+	ErrConversationNotFound = "conversation not found"
+)
+
+type InMemoryConversationRepo struct {
+	conversations map[string]models.Conversation
+	mu            sync.Mutex
+}
+
+func NewInMemoryConversationRepo() *InMemoryConversationRepo {
+	return &InMemoryConversationRepo{
+		conversations: make(map[string]models.Conversation),
+	}
+}
+
+func nextId() string {
+	return uuid.New().String()
+}
+
+func (r *InMemoryConversationRepo) CreateConversation(model string) (models.Conversation, error) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+
+	conv := models.Conversation{
+		Id:    nextId(),
+		Model: model,
+	}
+	r.conversations[conv.Id] = conv
+
+	return conv, nil
+}
+
+func (r *InMemoryConversationRepo) GetConversation(id string) (models.Conversation, error) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+
+	conv, ok := r.conversations[id]
+	if !ok {
+		return models.Conversation{}, errors.New(ErrConversationNotFound)
+	}
+
+	return conv, nil
+}
+
+func (r *InMemoryConversationRepo) AddMessage(id string, message models.Message) (models.Conversation, error) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+
+	conv, ok := r.conversations[id]
+	if !ok {
+		return models.Conversation{}, errors.New(ErrConversationNotFound)
+	}
+
+	conv.Messages = append(conv.Messages, message)
+	r.conversations[id] = conv
+
+	return conv, nil
+}

domain/models/conversation.go

@@ -0,0 +1,12 @@
+package models
+
+import "time"
+
+type Conversation struct {
+	Id                string
+	ConversationStart time.Time
+	Model             string
+	Messages          []Message
+	// Active means that this conversation can be continued (needs the model to be available in the main context)
+	Active bool
+}

domain/ports/outbound/llm_service.go

@@ -1,12 +1,16 @@
 package outbound
 
-import "context"
+import (
+	"context"
+
+	"github.com/ChristianSch/Theta/domain/models"
+)
 
 // ResponseHandler is a function that is called for every data chunk that is received. EOF is indicated by an empty chunk.
 type ResponseHandler func(ctx context.Context, chunk []byte) error
 
 type LlmService interface {
 	ListModels() ([]string, error)
 	SetModel(model string) error
-	SendMessage(prompt string, context []string, resHandler ResponseHandler) error
+	SendMessage(prompt string, context []models.Message, resHandler ResponseHandler) error
 }

domain/ports/outbound/repo/conversation_repo.go

@@ -0,0 +1,11 @@
+package repo
+
+import "github.com/ChristianSch/Theta/domain/models"
+
+type ConversationRepo interface {
+	// CreateConversation creates a new conversation with the given model
+	CreateConversation(model string) (models.Conversation, error)
+	// GetConversation returns the conversation with the given id
+	GetConversation(id string) (models.Conversation, error)
+	AddMessage(id string, message models.Message) (models.Conversation, error)
+}

domain/usecases/chat/handle_incoming_message.go

@@ -8,16 +8,18 @@ import (
 
 	"github.com/ChristianSch/Theta/domain/models"
 	"github.com/ChristianSch/Theta/domain/ports/outbound"
+	"github.com/ChristianSch/Theta/domain/ports/outbound/repo"
 	"github.com/gofiber/fiber/v2/log"
 	"github.com/google/uuid"
 )
 
 type IncomingMessageHandlerConfig struct {
 	// dependencies
-	Sender         outbound.SendMessageService
-	Formatter      outbound.MessageFormatter
-	Llm            outbound.LlmService
-	PostProcessors []outbound.PostProcessor
+	Sender           outbound.SendMessageService
+	Formatter        outbound.MessageFormatter
+	Llm              outbound.LlmService
+	PostProcessors   []outbound.PostProcessor
+	ConversationRepo repo.ConversationRepo
 }
 
 type IncomingMessageHandler struct {
@@ -43,7 +45,7 @@ func NewIncomingMessageHandler(cfg IncomingMessageHandlerConfig) *IncomingMessag
 	}
 }
 
-func (h *IncomingMessageHandler) Handle(message models.Message, connection interface{}) error {
+func (h *IncomingMessageHandler) Handle(message models.Message, conversation models.Conversation, connection interface{}) error {
 	msgId := fmt.Sprintf("msg-%s", strings.Split(uuid.New().String(), "-")[0])
 	log.Debug("starting processing of message", outbound.LogField{Key: "messageId", Value: msgId})
 
@@ -119,7 +121,7 @@ func (h *IncomingMessageHandler) Handle(message models.Message, connection inter
 
 	// send message to llm via a goroutine so we can wait for the answer
 	go func() {
-		err = h.cfg.Llm.SendMessage(message.Text, []string{}, fn)
+		err = h.cfg.Llm.SendMessage(message.Text, conversation.Messages, fn)
 		if err != nil {
 			done <- true
 		}
@@ -128,6 +130,13 @@ func (h *IncomingMessageHandler) Handle(message models.Message, connection inter
 	// wait for answer to be finished
 	<-done
 
+	// add message and answer to conversation
+	h.cfg.ConversationRepo.AddMessage(conversation.Id, message)
+	h.cfg.ConversationRepo.AddMessage(conversation.Id, models.Message{
+		Text: string(chunks),
+		Type: models.GptMessage,
+	})
+
 	if err != nil {
 		log.Error("error while receiving answer",
 			outbound.LogField{Key: "component", Value: "handle_incoming_message"},

infrastructure/views/chat.gohtml

@@ -1,25 +1,30 @@
 <div
     class="container mx-auto p-4 max-w-4xl text-lg flex flex-col h-screen"
     hx-ext="ws"
-    ws-connect="/ws/chat"
+    ws-connect="/ws/chat/{{.ConversationId}}"
     id="chat-socket"
+    x-data="{}"
 >
     <div id="content" class="flex-grow">
         <!-- Chat messages will be displayed here -->
+        {{ range.Messages }}
+        {{ . }}
+        {{ end }}
     </div>
 
-    <!-- prevent adding a newline when pressing enter by x-data and x-on -->
+    <!-- prevent adding a newline when pressing enter by x-data and x-on. note that we also trigger by load but catch empty
+    prompts in the backend part -->
     <textarea
         ws-send
-        x-data="{}"
         x-on:keydown.enter.prevent
-        hx-trigger="keydown[keyCode==13] from:body"
+        hx-trigger="load, keydown[keyCode==13] from:body"
         name="message"
         id="user-message"
         class="w-full display-block leading-8 p-4 mt-4 rounded-xl white bg-gray-900 border border-gray-700"
         placeholder="Message GPT..."
         style="min-height: 4rem"
-    ></textarea>
+        >{{.UserMessage}}</textarea
+    >
 
     <div class="text-sm pt-4 text-gray-500">
         Don't blindly trust the LLM. Use at your own risk. We don't secure you
@@ -31,8 +36,6 @@
 </div>
 
 <script>
-    const messages = [];
-
     htmx.on("htmx:wsAfterSend", function (evt) {
         const textarea = document.getElementById("user-message");
         textarea.value = "";

infrastructure/views/components/message.gohtml

@@ -1,8 +1,20 @@
 <div hx-swap-oob="beforeend:#content">
     <div class="text-left">
         <div class="font-semibold select-none pt-4">{{.Author}}</div>
-        <div class="inline-block text-white p-2 {{ if .IsGpt }}gpt-message{{end}}" id={{.MessageId}}>
-            {{.Text}}
+        <div class="text-white p-2 {{ if .IsGpt }}gpt-message{{end}}" id={{.MessageId}}>
+            {{ if .Text }}
+                {{.Text}}
+            {{ else }}
+                {{ if .IsGpt }}
+                    <div role="status">
+                        <svg aria-hidden="true" class="w-8 h-8 text-gray-200 animate-spin fill-gray-600 mx-auto" viewBox="0 0 100 101" fill="none" xmlns="http://www.w3.org/2000/svg">
+                            <path d="M100 50.5908C100 78.2051 77.6142 100.591 50 100.591C22.3858 100.591 0 78.2051 0 50.5908C0 22.9766 22.3858 0.59082 50 0.59082C77.6142 0.59082 100 22.9766 100 50.5908ZM9.08144 50.5908C9.08144 73.1895 27.4013 91.5094 50 91.5094C72.5987 91.5094 90.9186 73.1895 90.9186 50.5908C90.9186 27.9921 72.5987 9.67226 50 9.67226C27.4013 9.67226 9.08144 27.9921 9.08144 50.5908Z" fill="currentColor"/>
+                            <path d="M93.9676 39.0409C96.393 38.4038 97.8624 35.9116 97.0079 33.5539C95.2932 28.8227 92.871 24.3692 89.8167 20.348C85.8452 15.1192 80.8826 10.7238 75.2124 7.41289C69.5422 4.10194 63.2754 1.94025 56.7698 1.05124C51.7666 0.367541 46.6976 0.446843 41.7345 1.27873C39.2613 1.69328 37.813 4.19778 38.4501 6.62326C39.0873 9.04874 41.5694 10.4717 44.0505 10.1071C47.8511 9.54855 51.7191 9.52689 55.5402 10.0491C60.8642 10.7766 65.9928 12.5457 70.6331 15.2552C75.2735 17.9648 79.3347 21.5619 82.5849 25.841C84.9175 28.9121 86.7997 32.2913 88.1811 35.8758C89.083 38.2158 91.5421 39.6781 93.9676 39.0409Z" fill="currentFill"/>
+                        </svg>
+                        <span class="sr-only">Loading...</span>
+                    </div>
+                {{ end }}
+            {{ end }}
         </div>
     </div>
 </div>

infrastructure/views/layouts/empty.gohtml

@@ -0,0 +1 @@
+{{embed}}

infrastructure/views/layouts/main.gohtml

@@ -4,11 +4,11 @@
     <meta charset="UTF-8">
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
     <link rel="stylesheet" href="/static/base.css">
-    <link href="https://cdn.jsdelivr.net/npm/tailwindcss@2.2.15/dist/tailwind.min.css" rel="stylesheet">
+    <script src="https://cdn.tailwindcss.com"></script>
     <script src="https://unpkg.com/htmx.org/dist/htmx.min.js"></script>
     <script src="https://unpkg.com/htmx.org/dist/ext/ws.js"></script>
     <script src="https://unpkg.com/htmx.org/dist/ext/morphdom-swap.js"></script>
-    <script src="//unpkg.com/alpinejs" defer></script>
+    <script src="https://unpkg.com/alpinejs" defer></script>
     <script src="https://unpkg.com/[email protected]"></script>
     <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/11.9.0/styles/default.min.css">
     <script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/11.9.0/highlight.min.js"></script>

infrastructure/views/new_chat.gohtml

@@ -0,0 +1,48 @@
+<div
+    id="container"
+    class="container mx-auto p-4 max-w-4xl text-lg flex flex-col h-screen"
+    x-data="{}"
+>
+    <form hx-post="/chat" hx-target="#container" x-ref="newChatForm">
+        <div class="flex flex-col">
+            <label for="model" class="text-sm text-gray-500">Model</label>
+            <select
+                name="model"
+                id="model"
+                class="w-full display-block leading-8 p-4 mt-4 rounded-xl white bg-gray-900 border border-gray-700"
+            >
+            {{ range .Models }}
+                <option value="{{.}}">{{.}}</option>
+            {{ end }}
+            </select>
+        <div id="content" class="flex-grow">
+            <!-- Chat messages will be displayed here -->
+        </div>
+
+        <!-- prevent adding a newline when pressing enter by x-data and x-on -->
+        <textarea
+            name="message"
+            id="user-message"
+            class="w-full display-block leading-8 p-4 mt-4 rounded-xl white bg-gray-900 border border-gray-700"
+            placeholder="Message GPT..."
+            style="min-height: 4rem"
+        ></textarea>
+
+        <input type="submit" style="display: none;">
+
+        <div class="text-sm pt-4 text-gray-500">
+            Don't blindly trust the LLM. Use at your own risk. We don't secure you
+            against XSS or other attacks on purpose, as that would mean censored
+            output.
+        </div>
+    </form>
+</div>
+
+<script>
+    document.getElementById('user-message').addEventListener('keydown', function(event) {
+        if (event.key === 'Enter' && !event.shiftKey) {
+            this.form.requestSubmit();
+        }
+    });
+</script>
+