messages.ts

import { composeContext } from "@ai16z/eliza/src/context.ts";
import {
    generateMessageResponse,
    generateShouldRespond,
} from "@ai16z/eliza/src/generation.ts";
import { embeddingZeroVector } from "@ai16z/eliza/src/memory.ts";
import {
    messageCompletionFooter,
    shouldRespondFooter,
} from "@ai16z/eliza/src/parsing.ts";
import {
    Content,
    HandlerCallback,
    IAgentRuntime,
    IBrowserService,
    ISpeechService,
    IVideoService,
    Media,
    Memory,
    ModelClass,
    ServiceType,
    State,
    UUID,
} from "@ai16z/eliza/src/types.ts";
import { stringToUuid } from "@ai16z/eliza/src/uuid.ts";
import { generateText, trimTokens } from "@ai16z/eliza/src/generation.ts";
import { parseJSONObjectFromText } from "@ai16z/eliza/src/parsing.ts";
import {
    ChannelType,
    Client,
    Message as DiscordMessage,
    PermissionsBitField,
    TextChannel,
    ThreadChannel,
} from "discord.js";
import { elizaLogger } from "@ai16z/eliza/src/logger.ts";
import { AttachmentManager } from "./attachments.ts";
import { VoiceManager } from "./voice.ts";
import { Service } from "@ai16z/eliza";

const MAX_MESSAGE_LENGTH = 1900;
async function generateSummary(
    runtime: IAgentRuntime,
    text: string
): Promise<{ title: string; description: string }> {
    // make sure text is under 128k characters
    text = trimTokens(text, 100000, "gpt-4o-mini"); // TODO: clean this up

    const prompt = `Please generate a concise summary for the following text:
  
  Text: """
  ${text}
  """
  
  Respond with a JSON object in the following format:
  \`\`\`json
  {
    "title": "Generated Title",
    "summary": "Generated summary and/or description of the text"
  }
  \`\`\``;

    const response = await generateText({
        runtime,
        context: prompt,
        modelClass: ModelClass.SMALL,
    });

    const parsedResponse = parseJSONObjectFromText(response);

    if (parsedResponse) {
        return {
            title: parsedResponse.title,
            description: parsedResponse.summary,
        };
    }

    return {
        title: "",
        description: "",
    };
}

export type InterestChannels = {
    [key: string]: {
        lastMessageSent: number;
        messages: { userId: UUID; userName: string; content: Content }[];
    };
};

const discordShouldRespondTemplate =
    `# Task: Decide if {{agentName}} should respond.
About {{agentName}}:
{{bio}}

# INSTRUCTIONS: Determine if {{agentName}} should respond to the message and participate in the conversation. Do not comment. Just respond with "RESPOND" or "IGNORE" or "STOP".

# RESPONSE EXAMPLES
<user 1>: I just saw a really great movie
<user 2>: Oh? Which movie?
Result: [IGNORE]

{{agentName}}: Oh, this is my favorite scene
<user 1>: sick
<user 2>: wait, why is it your favorite scene
Result: [RESPOND]

<user>: stfu bot
Result: [STOP]

<user>: Hey {{agent}}, can you help me with something
Result: [RESPOND]

<user>: {{agentName}} stfu plz
Result: [STOP]

<user>: i need help
{{agentName}}: how can I help you?
<user>: no. i need help from someone else
Result: [IGNORE]

<user>: Hey {{agent}}, can I ask you a question
{{agentName}}: Sure, what is it
<user>: can you ask claude to create a basic react module that demonstrates a counter
Result: [RESPOND]

<user>: {{agentName}} can you tell me a story
<user>: {about a girl named elara
{{agentName}}: Sure.
{{agentName}}: Once upon a time, in a quaint little village, there was a curious girl named Elara.
{{agentName}}: Elara was known for her adventurous spirit and her knack for finding beauty in the mundane.
<user>: I'm loving it, keep going
Result: [RESPOND]

<user>: {{agentName}} stop responding plz
Result: [STOP]

<user>: okay, i want to test something. can you say marco?
{{agentName}}: marco
<user>: great. okay, now do it again
Result: [RESPOND]

Response options are [RESPOND], [IGNORE] and [STOP].

{{agentName}} is in a room with other users and is very worried about being annoying and saying too much.
Respond with [RESPOND] to messages that are directed at {{agentName}}, or participate in conversations that are interesting or relevant to their background.
If a message is not interesting or relevant, respond with [IGNORE]
Unless directly responding to a user, respond with [IGNORE] to messages that are very short or do not contain much information.
If a user asks {{agentName}} to be quiet, respond with [STOP]
If {{agentName}} concludes a conversation and isn't part of the conversation anymore, respond with [STOP]

IMPORTANT: {{agentName}} is particularly sensitive about being annoying, so if there is any doubt, it is better to respond with [IGNORE].
If {{agentName}} is conversing with a user and they have not asked to stop, it is better to respond with [RESPOND].

{{recentMessages}}

# INSTRUCTIONS: Choose the option that best describes {{agentName}}'s response to the last message. Ignore messages if they are addressed to someone else.
` + shouldRespondFooter;

export const discordMessageHandlerTemplate =
    // {{goals}}
    `# Action Examples
{{actionExamples}}
(Action examples are for reference only. Do not use the information from them in your response.)

# Knowledge
{{knowledge}}

# Task: Generate dialog and actions for the character {{agentName}}.
About {{agentName}}:
{{bio}}
{{lore}}

Examples of {{agentName}}'s dialog and actions:
{{characterMessageExamples}}

{{providers}}

{{attachments}}

{{actions}}

# Capabilities
Note that {{agentName}} is capable of reading/seeing/hearing various forms of media, including images, videos, audio, plaintext and PDFs. Recent attachments have been included above under the "Attachments" section.

{{messageDirections}}

{{recentMessages}}

# Instructions: Write the next message for {{agentName}}. Include an action, if appropriate. {{actionNames}}
` + messageCompletionFooter;

export async function sendMessageInChunks(
    channel: TextChannel,
    content: string,
    inReplyTo: string,
    files: any[]
): Promise<DiscordMessage[]> {
    const sentMessages: DiscordMessage[] = [];
    const messages = splitMessage(content);
    try {
        for (let i = 0; i < messages.length; i++) {
            const message = messages[i];
            if (
                message.trim().length > 0 ||
                (i === messages.length - 1 && files && files.length > 0)
            ) {
                const options: any = {
                    content: message.trim(),
                };

                // if (i === 0 && inReplyTo) {
                //   // Reply to the specified message for the first chunk
                //   options.reply = {
                //     messageReference: inReplyTo,
                //   };
                // }

                if (i === messages.length - 1 && files && files.length > 0) {
                    // Attach files to the last message chunk
                    options.files = files;
                }

                const m = await channel.send(options);
                sentMessages.push(m);
            }
        }
    } catch (error) {
        elizaLogger.error("Error sending message:", error);
    }

    return sentMessages;
}

function splitMessage(content: string): string[] {
    const messages: string[] = [];
    let currentMessage = "";

    const rawLines = content?.split("\n") || [];
    // split all lines into MAX_MESSAGE_LENGTH chunks so any long lines are split
    const lines = rawLines
        .map((line) => {
            const chunks = [];
            while (line.length > MAX_MESSAGE_LENGTH) {
                chunks.push(line.slice(0, MAX_MESSAGE_LENGTH));
                line = line.slice(MAX_MESSAGE_LENGTH);
            }
            chunks.push(line);
            return chunks;
        })
        .flat();

    for (const line of lines) {
        if (currentMessage.length + line.length + 1 > MAX_MESSAGE_LENGTH) {
            messages.push(currentMessage.trim());
            currentMessage = "";
        }
        currentMessage += line + "\n";
    }

    if (currentMessage.trim().length > 0) {
        messages.push(currentMessage.trim());
    }

    return messages;
}

function canSendMessage(channel) {
    // if it is a DM channel, we can always send messages
    if (channel.type === ChannelType.DM) {
        return {
            canSend: true,
            reason: null,
        };
    }
    const botMember = channel.guild?.members.cache.get(channel.client.user.id);

    if (!botMember) {
        return {
            canSend: false,
            reason: "Not a guild channel or bot member not found",
        };
    }

    // Required permissions for sending messages
    const requiredPermissions = [
        PermissionsBitField.Flags.ViewChannel,
        PermissionsBitField.Flags.SendMessages,
        PermissionsBitField.Flags.ReadMessageHistory,
    ];

    // Add thread-specific permission if it's a thread
    if (channel instanceof ThreadChannel) {
        requiredPermissions.push(
            PermissionsBitField.Flags.SendMessagesInThreads
        );
    }

    // Check permissions
    const permissions = channel.permissionsFor(botMember);

    if (!permissions) {
        return {
            canSend: false,
            reason: "Could not retrieve permissions",
        };
    }

    // Check each required permission
    const missingPermissions = requiredPermissions.filter(
        (perm) => !permissions.has(perm)
    );

    return {
        canSend: missingPermissions.length === 0,
        missingPermissions: missingPermissions,
        reason:
            missingPermissions.length > 0
                ? `Missing permissions: ${missingPermissions.map((p) => String(p)).join(", ")}`
                : null,
    };
}

export class MessageManager {
    private client: Client;
    private runtime: IAgentRuntime;
    private attachmentManager: AttachmentManager;
    private interestChannels: InterestChannels = {};
    private discordClient: any;
    private voiceManager: VoiceManager;

    constructor(discordClient: any, voiceManager: VoiceManager) {
        this.client = discordClient.client;
        this.voiceManager = voiceManager;
        this.discordClient = discordClient;
        this.runtime = discordClient.runtime;
        this.attachmentManager = new AttachmentManager(this.runtime);
    }

    async handleMessage(message: DiscordMessage) {
        if (
            message.interaction ||
            message.author.id ===
                this.client.user?.id /* || message.author?.bot*/
        )
            return;
        const userId = message.author.id as UUID;
        const userName = message.author.username;
        const name = message.author.displayName;
        const channelId = message.channel.id;

        try {
            const { processedContent, attachments } =
                await this.processMessageMedia(message);

            const audioAttachments = message.attachments.filter((attachment) =>
                attachment.contentType?.startsWith("audio/")
            );
            if (audioAttachments.size > 0) {
                const processedAudioAttachments =
                    await this.attachmentManager.processAttachments(
                        audioAttachments
                    );
                attachments.push(...processedAudioAttachments);
            }

            const roomId = stringToUuid(channelId + "-" + this.runtime.agentId);
            const userIdUUID = stringToUuid(userId);

            await this.runtime.ensureConnection(
                userIdUUID,
                roomId,
                userName,
                name,
                "discord"
            );

            const messageId = stringToUuid(
                message.id + "-" + this.runtime.agentId
            );

            let shouldIgnore = false;
            let shouldRespond = true;

            const content: Content = {
                text: processedContent,
                attachments: attachments,
                source: "discord",
                url: message.url,
                inReplyTo: message.reference?.messageId
                    ? stringToUuid(
                          message.reference.messageId +
                              "-" +
                              this.runtime.agentId
                      )
                    : undefined,
            };

            const userMessage = {
                content,
                userId: userIdUUID,
                agentId: this.runtime.agentId,
                roomId,
            };

            const memory: Memory = {
                id: stringToUuid(message.id + "-" + this.runtime.agentId),
                ...userMessage,
                userId: userIdUUID,
                agentId: this.runtime.agentId,
                roomId,
                content,
                createdAt: message.createdTimestamp,
                embedding: embeddingZeroVector,
            };

            if (content.text) {
                await this.runtime.messageManager.createMemory(memory);
            }

            let state = (await this.runtime.composeState(userMessage, {
                discordClient: this.client,
                discordMessage: message,
                agentName:
                    this.runtime.character.name ||
                    this.client.user?.displayName,
            })) as State;

            if (!canSendMessage(message.channel).canSend) {
                return elizaLogger.warn(
                    `Cannot send message to channel ${message.channel}`,
                    canSendMessage(message.channel)
                );
            }

            if (!shouldIgnore) {
                shouldIgnore = await this._shouldIgnore(message);
            }

            if (shouldIgnore) {
                return;
            }
            const hasInterest = this._checkInterest(channelId);

            const agentUserState =
                await this.runtime.databaseAdapter.getParticipantUserState(
                    roomId,
                    this.runtime.agentId
                );

            if (
                agentUserState === "MUTED" &&
                !message.mentions.has(this.client.user.id) &&
                !hasInterest
            ) {
                console.log("Ignoring muted room");
                // Ignore muted rooms unless explicitly mentioned
                return;
            }

            if (agentUserState === "FOLLOWED") {
                shouldRespond = true; // Always respond in followed rooms
            } else if (
                (!shouldRespond && hasInterest) ||
                (shouldRespond && !hasInterest)
            ) {
                shouldRespond = await this._shouldRespond(message, state);
            }

            if (shouldRespond) {
                const context = composeContext({
                    state,
                    template:
                        this.runtime.character.templates
                            ?.discordMessageHandlerTemplate ||
                        discordMessageHandlerTemplate,
                });

                const responseContent = await this._generateResponse(
                    memory,
                    state,
                    context
                );

                responseContent.text = responseContent.text?.trim();
                responseContent.inReplyTo = stringToUuid(
                    message.id + "-" + this.runtime.agentId
                );

                if (!responseContent.text) {
                    return;
                }

                const callback: HandlerCallback = async (
                    content: Content,
                    files: any[]
                ) => {
                    try {
                        if (message.id && !content.inReplyTo) {
                            content.inReplyTo = stringToUuid(
                                message.id + "-" + this.runtime.agentId
                            );
                        }
                        if (false) {
                            // For voice channels, use text-to-speech
                            const audioStream = await this.runtime
                                .getService(ServiceType.SPEECH_GENERATION)
                                .getInstance<ISpeechService>()
                                .generate(this.runtime, content.text);
                            await this.voiceManager.playAudioStream(
                                userId,
                                audioStream
                            );
                            const memory: Memory = {
                                id: stringToUuid(
                                    message.id + "-" + this.runtime.agentId
                                ),
                                userId: this.runtime.agentId,
                                agentId: this.runtime.agentId,
                                content,
                                roomId,
                                embedding: embeddingZeroVector,
                            };
                            return [memory];
                        } else {
                            // For text channels, send the message
                            const messages = await sendMessageInChunks(
                                message.channel as TextChannel,
                                content.text,
                                message.id,
                                files
                            );

                            const memories: Memory[] = [];
                            for (const m of messages) {
                                let action = content.action;
                                // If there's only one message or it's the last message, keep the original action
                                // For multiple messages, set all but the last to 'CONTINUE'
                                if (
                                    messages.length > 1 &&
                                    m !== messages[messages.length - 1]
                                ) {
                                    action = "CONTINUE";
                                }

                                const memory: Memory = {
                                    id: stringToUuid(
                                        m.id + "-" + this.runtime.agentId
                                    ),
                                    userId: this.runtime.agentId,
                                    agentId: this.runtime.agentId,
                                    content: {
                                        ...content,
                                        action,
                                        inReplyTo: messageId,
                                        url: m.url,
                                    },
                                    roomId,
                                    embedding: embeddingZeroVector,
                                    createdAt: m.createdTimestamp,
                                };
                                memories.push(memory);
                            }
                            for (const m of memories) {
                                await this.runtime.messageManager.createMemory(
                                    m
                                );
                            }
                            return memories;
                        }
                    } catch (error) {
                        console.error("Error sending message:", error);
                        return [];
                    }
                };

                const responseMessages = await callback(responseContent);

                state = await this.runtime.updateRecentMessageState(state);

                await this.runtime.processActions(
                    memory,
                    responseMessages,
                    state,
                    callback
                );
            }
            await this.runtime.evaluate(memory, state, shouldRespond);
        } catch (error) {
            console.error("Error handling message:", error);
            if (message.channel.type === ChannelType.GuildVoice) {
                // For voice channels, use text-to-speech for the error message
                const errorMessage = "Sorry, I had a glitch. What was that?";
                const audioStream = await this.runtime
                    .getService(ServiceType.SPEECH_GENERATION)
                    .getInstance<ISpeechService>()
                    .generate(this.runtime, errorMessage);
                await this.voiceManager.playAudioStream(userId, audioStream);
            } else {
                // For text channels, send the error message
                console.error("Error sending message:", error);
            }
        }
    }

    async cacheMessages(channel: TextChannel, count: number = 20) {
        const messages = await channel.messages.fetch({ limit: count });

        // TODO: This is throwing an error but seems to work?
        for (const [_, message] of messages) {
            await this.handleMessage(message);
        }
    }

    async processMessageMedia(
        message: DiscordMessage
    ): Promise<{ processedContent: string; attachments: Media[] }> {
        let processedContent = message.content;
        let attachments: Media[] = [];

        // Process code blocks in the message content
        const codeBlockRegex = /```([\s\S]*?)```/g;
        let match;
        while ((match = codeBlockRegex.exec(processedContent))) {
            const codeBlock = match[1];
            const lines = codeBlock.split("\n");
            const title = lines[0];
            const description = lines.slice(0, 3).join("\n");
            const attachmentId =
                `code-${Date.now()}-${Math.floor(Math.random() * 1000)}`.slice(
                    -5
                );
            attachments.push({
                id: attachmentId,
                url: "",
                title: title || "Code Block",
                source: "Code",
                description: description,
                text: codeBlock,
            });
            processedContent = processedContent.replace(
                match[0],
                `Code Block (${attachmentId})`
            );
        }

        // Process message attachments
        if (message.attachments.size > 0) {
            attachments = await this.attachmentManager.processAttachments(
                message.attachments
            );
        }

        // TODO: Move to attachments manager
        const urlRegex = /(https?:\/\/[^\s]+)/g;
        const urls = processedContent.match(urlRegex) || [];

        for (const url of urls) {
            if (
                this.runtime
                    .getService(ServiceType.VIDEO)
                    .getInstance<IVideoService>()
                    .isVideoUrl(url)
            ) {
                const videoInfo = await this.runtime
                    .getService(ServiceType.VIDEO)
                    .getInstance<IVideoService>()
                    .processVideo(url);
                attachments.push({
                    id: `youtube-${Date.now()}`,
                    url: url,
                    title: videoInfo.title,
                    source: "YouTube",
                    description: videoInfo.description,
                    text: videoInfo.text,
                });
            } else {
                const { title, bodyContent } = await this.runtime
                    .getService(ServiceType.BROWSER)
                    .getInstance<IBrowserService>()
                    .getPageContent(url, this.runtime);
                const { title: newTitle, description } = await generateSummary(
                    this.runtime,
                    title + "\n" + bodyContent
                );
                attachments.push({
                    id: `webpage-${Date.now()}`,
                    url: url,
                    title: newTitle || "Web Page",
                    source: "Web",
                    description,
                    text: bodyContent,
                });
            }
        }

        return { processedContent, attachments };
    }

    private _checkInterest(channelId: string): boolean {
        return !!this.interestChannels[channelId];
    }

    private async _shouldIgnore(message: DiscordMessage): Promise<boolean> {
        // if the message is from us, ignore
        if (message.author.id === this.client.user?.id) return true;
        let messageContent = message.content.toLowerCase();

        // Replace the bot's @ping with the character name
        const botMention = `<@!?${this.client.user?.id}>`;
        messageContent = messageContent.replace(
            new RegExp(botMention, "gi"),
            this.runtime.character.name.toLowerCase()
        );

        // Replace the bot's username with the character name
        const botUsername = this.client.user?.username.toLowerCase();
        messageContent = messageContent.replace(
            new RegExp(`\\b${botUsername}\\b`, "g"),
            this.runtime.character.name.toLowerCase()
        );

        // strip all special characters
        messageContent = messageContent.replace(/[^a-zA-Z0-9\s]/g, "");

        // short responses where ruby should stop talking and disengage unless mentioned again
        const loseInterestWords = [
            "shut up",
            "stop",
            "please shut up",
            "shut up please",
            "dont talk",
            "silence",
            "stop talking",
            "be quiet",
            "hush",
            "wtf",
            "chill",
            "stfu",
            "stupid bot",
            "dumb bot",
            "stop responding",
            "god damn it",
            "god damn",
            "goddamnit",
            "can you not",
            "can you stop",
            "be quiet",
            "hate you",
            "hate this",
            "fuck up",
        ];
        if (
            messageContent.length < 100 &&
            loseInterestWords.some((word) => messageContent.includes(word))
        ) {
            delete this.interestChannels[message.channelId];
            return true;
        }

        // If we're not interested in the channel and it's a short message, ignore it
        if (
            messageContent.length < 10 &&
            !this.interestChannels[message.channelId]
        ) {
            return true;
        }

        const targetedPhrases = [
            this.runtime.character.name + " stop responding",
            this.runtime.character.name + " stop talking",
            this.runtime.character.name + " shut up",
            this.runtime.character.name + " stfu",
            "stop talking" + this.runtime.character.name,
            this.runtime.character.name + " stop talking",
            "shut up " + this.runtime.character.name,
            this.runtime.character.name + " shut up",
            "stfu " + this.runtime.character.name,
            this.runtime.character.name + " stfu",
            "chill" + this.runtime.character.name,
            this.runtime.character.name + " chill",
        ];

        // lose interest if pinged and told to stop responding
        if (targetedPhrases.some((phrase) => messageContent.includes(phrase))) {
            delete this.interestChannels[message.channelId];
            return true;
        }

        // if the message is short, ignore but maintain interest
        if (
            !this.interestChannels[message.channelId] &&
            messageContent.length < 2
        ) {
            return true;
        }

        const ignoreResponseWords = [
            "lol",
            "nm",
            "uh",
            "wtf",
            "stfu",
            "dumb",
            "jfc",
            "omg",
        ];
        if (
            message.content.length < 4 &&
            ignoreResponseWords.some((word) =>
                message.content.toLowerCase().includes(word)
            )
        ) {
            return true;
        }
        return false;
    }

    private async _shouldRespond(
        message: DiscordMessage,
        state: State
    ): Promise<boolean> {
        if (message.author.id === this.client.user?.id) return false;
        // if (message.author.bot) return false;
        if (message.mentions.has(this.client.user?.id as string)) return true;

        const guild = message.guild;
        const member = guild?.members.cache.get(this.client.user?.id as string);
        const nickname = member?.nickname;

        if (
            message.content
                .toLowerCase()
                .includes(this.client.user?.username.toLowerCase() as string) ||
            message.content
                .toLowerCase()
                .includes(this.client.user?.tag.toLowerCase() as string) ||
            (nickname &&
                message.content.toLowerCase().includes(nickname.toLowerCase()))
        ) {
            return true;
        }

        if (!message.guild) {
            return true;
        }

        // If none of the above conditions are met, use the generateText to decide
        const shouldRespondContext = composeContext({
            state,
            template:
                this.runtime.character.templates
                    ?.discordShouldRespondTemplate ||
                this.runtime.character.templates?.shouldRespondTemplate ||
                discordShouldRespondTemplate,
        });

        const response = await generateShouldRespond({
            runtime: this.runtime,
            context: shouldRespondContext,
            modelClass: ModelClass.SMALL,
        });

        if (response === "RESPOND") {
            return true;
        } else if (response === "IGNORE") {
            return false;
        } else if (response === "STOP") {
            delete this.interestChannels[message.channelId];
            return false;
        } else {
            console.error(
                "Invalid response from response generateText:",
                response
            );
            return false;
        }
    }

    private async _generateResponse(
        message: Memory,
        state: State,
        context: string
    ): Promise<Content> {
        const { userId, roomId } = message;

        const response = await generateMessageResponse({
            runtime: this.runtime,
            context,
            modelClass: ModelClass.SMALL,
        });

        if (!response) {
            console.error("No response from generateMessageResponse");
            return;
        }

        await this.runtime.databaseAdapter.log({
            body: { message, context, response },
            userId: userId,
            roomId,
            type: "response",
        });

        return response;
    }

    async fetchBotName(botToken: string) {
        const url = "https://discord.com/api/v10/users/@me";

        const response = await fetch(url, {
            method: "GET",
            headers: {
                Authorization: `Bot ${botToken}`,
            },
        });

        if (!response.ok) {
            throw new Error(
                `Error fetching bot details: ${response.statusText}`
            );
        }

        const data = await response.json();
        return data.username;
    }
}