use crate::{Context, Error};
use crate::messaging::trace_message;
use tracing::{debug, error, info, trace, warn};
use ollama_rs::Ollama;
use std::env;
use std::time::Duration;
use ollama_rs::models::create::CreateModelRequest;
use ollama_rs::generation::completion::request::GenerationRequest;
use chrono::{DateTime, Utc};
use tokio_stream::StreamExt;
use poise::CreateReply;

/// Prompt the super advanced EwiAI
#[poise::command(slash_command, prefix_command)]
pub async fn prompt(
    ctx: Context<'_>,
    #[description = "The prompt to send to EwiAI"]
    prompt: String,
    #[description = "Include recent messages (max 10)"]
    include_messages: Option<u8>,
) -> Result<(), Error> {
    let channel_str = ctx.channel_id().to_string();
    let guild_str = ctx.guild_id().map_or_else(|| "DM".to_string(), |g| g.to_string());

    trace!("prompt command called by user {} (ID: {}) in channel {} (guild: {})", ctx.author().name, ctx.author().id, channel_str, guild_str);
    debug!("{} has requested to prompt EwiAI with '{}' (include_messages: {:?})", ctx.author().name, prompt, include_messages);

    trace!("Checking environment variable TG_BOT_OLLAMA_HOST...");
    let host = match env::var("TG_BOT_OLLAMA_HOST") {
        Ok(h) => {
            trace!("Successfully retrieved TG_BOT_OLLAMA_HOST = {}", h);
            h
        }
        Err(e) => {
            warn!("TG_BOT_OLLAMA_HOST environment variable missing: {}", e);
            let err_msg = "Error: Expected an ollama url in the environment (`TG_BOT_OLLAMA_HOST`).";
            trace_message(err_msg, channel_str.clone(), guild_str.clone()).await;
            ctx.say(err_msg).await?;
            return Ok(());
        }
    };

    let formatted_host = if !host.starts_with("http://") && !host.starts_with("https://") {
        let f = format!("http://{}", host);
        trace!("Formatted Ollama host address to include http:// prefix: {}", f);
        f
    } else {
        trace!("Ollama host address already has URL scheme: {}", host);
        host
    };

    trace!("Building Ollama client instance for host {} on port 11434...", formatted_host);
    let ollama = Ollama::builder()
        .host(&formatted_host)
        .port(11434)
        .build();

    trace!("Requesting local model list from Ollama server...");
    let model_list = match ollama.list_local_models().await {
        Ok(m) => {
            debug!("Successfully retrieved {} models from Ollama server", m.len());
            trace!("Local models available: {:?}", m.iter().map(|model| &model.name).collect::<Vec<_>>());
            m
        }
        Err(e) => {
            error!("Failed to connect to Ollama server at {}: {}", formatted_host, e);
            let err_msg = format!("Error: Failed to connect to Ollama: {}", e);
            trace_message(&err_msg, channel_str.clone(), guild_str.clone()).await;
            ctx.say(err_msg).await?;
            return Ok(());
        }
    };

    let mut needs_create = true;
    let yesterday = Utc::now() - chrono::Duration::days(1);
    trace!("Checking model list against yesterday's cutoff timestamp ({})", yesterday);

    for model in &model_list {
        trace!("Inspecting model entry: '{}', modified_at: '{}'", model.name, model.modified_at);
        if model.name.starts_with("EwiAI") {
            debug!("Found matching model candidate: '{}'", model.name);
            match DateTime::parse_from_rfc3339(&model.modified_at) {
                Ok(modified) => {
                    let modified_utc = modified.with_timezone(&Utc);
                    trace!("Parsed model modified_at: {} (UTC: {})", modified, modified_utc);
                    if modified_utc > yesterday {
                        info!("Model '{}' is up to date (modified {} > cutoff {})", model.name, modified_utc, yesterday);
                        needs_create = false;
                    } else {
                        trace!("Model '{}' was modified at {}, which is older than cutoff {}", model.name, modified_utc, yesterday);
                    }
                }
                Err(e) => {
                    warn!("Failed to parse modified_at date '{}' for model '{}': {}", model.modified_at, model.name, e);
                }
            }
        }
    }

    if needs_create {
        info!("EwiAI model is missing or out of date. Initiating model creation/update...");
        let system_prompt = system_prompt();
        trace!("Generated system prompt for model creation (length: {} chars)", system_prompt.len());
        trace!("System prompt content: {}", system_prompt);
        let notice_msg = "EwiAI model is missing or out of date. Creating/updating model (this may take a bit)...";
        trace_message(notice_msg, channel_str.clone(), guild_str.clone()).await;
        let _ = ctx.say(notice_msg).await?;

        trace!("Sending create_model request to Ollama for 'EwiAI' based on 'gemma4:e2b-it-qat'...");
        if let Err(e) = ollama.create_model(CreateModelRequest::new("EwiAI".into())
            .system(system_prompt.into())
            .from_model("gemma4:e2b-it-qat".into())).await {
            error!("Failed to create model EwiAI: {}", e);
            let err_msg = format!("Error creating model: {}", e);
            trace_message(&err_msg, channel_str.clone(), guild_str.clone()).await;
            ctx.say(err_msg).await?;
            return Ok(());
        }
        info!("Successfully created/updated EwiAI model");
    } else {
        trace!("Skipping model creation; existing model is up to date.");
    }

    let mut final_prompt = String::new();
    if let Some(mut limit) = include_messages {
        trace!("include_messages specified: {}", limit);
        if limit > 10 {
            trace!("Clamping include_messages limit from {} to 10", limit);
            limit = 10;
        }
        trace!("Fetching last {} messages from channel {}", limit, channel_str);
        match ctx.channel_id().messages(ctx.http(), poise::serenity_prelude::GetMessages::new().limit(limit)).await {
            Ok(msgs) => {
                debug!("Retrieved {} messages from channel {}", msgs.len(), channel_str);
                final_prompt.push_str("Recent channel messages:\n");
                for msg in msgs.iter().rev() {
                    let mut content = msg.content.clone();
                    if content.len() > 100 {
                        content.truncate(100);
                        content.push_str("...");
                    }
                    trace!("Appending message from author {} (ID: {}): snippet='{}'", msg.author.name, msg.author.id, content);
                    final_prompt.push_str(&format!("{} (ID: {}): {}\n", msg.author.name, msg.author.id, content));
                }
                final_prompt.push_str("\n");
            }
            Err(e) => {
                error!("Failed to get messages from channel {}: {:?}", channel_str, e);
            }
        }
    }
    
    final_prompt.push_str(&format!("The user talking to you is {} (ID: {}).\nThat user is telling you the following: {}", ctx.author().name, ctx.author().id, prompt));
    trace!("Final prompt assembled (length: {} chars)", final_prompt.len());
    trace!("Final prompt: {}", final_prompt);

    let thinking_msg = "Thinking...\n-# The hardware this thing runs is really slow, expect a long wait";
    trace_message(thinking_msg, channel_str.clone(), guild_str.clone()).await;
    trace!("Sending initial thinking message...");
    let reply = ctx.say(thinking_msg).await?;

    let sys_prompt = system_prompt();
    trace!("Generating stream from Ollama model 'EwiAI' with system prompt length {}...", sys_prompt.len());
    let mut stream = match ollama.generate_stream(GenerationRequest::new("EwiAI".into(), final_prompt).system(sys_prompt)).await {
        Ok(s) => {
            debug!("Successfully initiated generation stream");
            s
        }
        Err(e) => {
            error!("Error starting generation stream with Ollama: {}", e);
            let err_msg = format!("Error during generation: {}", e);
            trace_message(&err_msg, channel_str.clone(), guild_str.clone()).await;
            reply.edit(ctx, CreateReply::default().content(err_msg)).await?;
            return Ok(());
        }
    };

    let mut response_text = String::new();
    let mut last_update = std::time::Instant::now();
    let mut final_response = None;
    let mut chunk_count = 0usize;

    trace!("Reading chunks from generation stream...");
    while let Some(res) = stream.next().await {
        match res {
            Ok(chunks) => {
                chunk_count += chunks.len();
                trace!("Received stream batch containing {} chunk(s)", chunks.len());
                for chunk in chunks {
                    trace!("Chunk response segment: '{}', done: {}", chunk.response, chunk.done);
                    response_text.push_str(&chunk.response);
                    if chunk.done {
                        debug!("Stream chunk marked done");
                        final_response = Some(chunk);
                    }
                }

                if last_update.elapsed() >= Duration::from_secs(1) && !response_text.is_empty() {
                    trace!("Throttled stream update: updating reply message (response_text length: {} chars)...", response_text.len());
                    let _ = reply.edit(ctx, CreateReply::default().content(&response_text)).await;
                    last_update = std::time::Instant::now();
                }
            }
            Err(e) => {
                error!("Error encountered while reading stream chunk: {}", e);
                let err_msg = format!("{} [Stream Error: {}]", response_text, e);
                trace_message(&err_msg, channel_str.clone(), guild_str.clone()).await;
                let _ = reply.edit(ctx, CreateReply::default().content(err_msg)).await;
                return Ok(());
            }
        }
    }
    info!("Stream completed. Total chunks received: {}, output length: {} chars", chunk_count, response_text.len());

    if let Some(stats) = final_response {
        let total_duration = stats.total_duration.unwrap_or(0) as f64 / 1_000_000_000.0;
        let eval_count = stats.eval_count.unwrap_or(0);
        let eval_duration = stats.eval_duration.unwrap_or(0) as f64 / 1_000_000_000.0;
        let tokens_per_sec = if eval_duration > 0.0 { eval_count as f64 / eval_duration } else { 0.0 };

        debug!("Generation stats: total_duration={:.2}s, eval_count={}, eval_duration={:.2}s, tok/s={:.2}", total_duration, eval_count, eval_duration, tokens_per_sec);

        let stats_text = format!(
            "\n\n*Generated in {:.2}s ({:.2} tok/s)*",
            total_duration, tokens_per_sec
        );
        response_text.push_str(&stats_text);
    } else {
        trace!("No final response stats available from stream");
    }
    
    trace_message(&response_text, channel_str.clone(), guild_str.clone()).await;
    trace!("Sending final edit to reply message...");
    let _ = reply.edit(ctx, CreateReply::default().content(&response_text)).await;
    debug!("prompt command finished successfully for user {}", ctx.author().name);

    Ok(())
}

/// Ask EwiAI to answer to recent messages
#[poise::command(slash_command, prefix_command)]
pub async fn answer(
    ctx: Context<'_>,
    #[description = "How many messages to include (max 10)"]
    messages_count: Option<u8>,
) -> Result<(), Error> {
    let channel_str = ctx.channel_id().to_string();
    let guild_str = ctx.guild_id().map_or_else(|| "DM".to_string(), |g| g.to_string());

    trace!("answer command called by user {} (ID: {}) in channel {} (guild: {})", ctx.author().name, ctx.author().id, channel_str, guild_str);
    debug!("{} has requested EwiAI to answer (messages_count: {:?})", ctx.author().name, messages_count);

    trace!("Checking environment variable TG_BOT_OLLAMA_HOST...");
    let host = match env::var("TG_BOT_OLLAMA_HOST") {
        Ok(h) => {
            trace!("Successfully retrieved TG_BOT_OLLAMA_HOST = {}", h);
            h
        }
        Err(e) => {
            warn!("TG_BOT_OLLAMA_HOST environment variable missing: {}", e);
            let err_msg = "Error: Expected an ollama url in the environment (`TG_BOT_OLLAMA_HOST`).";
            trace_message(err_msg, channel_str.clone(), guild_str.clone()).await;
            ctx.say(err_msg).await?;
            return Ok(());
        }
    };

    let formatted_host = if !host.starts_with("http://") && !host.starts_with("https://") {
        let f = format!("http://{}", host);
        trace!("Formatted Ollama host address to include http:// prefix: {}", f);
        f
    } else {
        trace!("Ollama host address already has URL scheme: {}", host);
        host
    };

    trace!("Building Ollama client instance for host {} on port 11434...", formatted_host);
    let ollama = Ollama::builder()
        .host(&formatted_host)
        .port(11434)
        .build();

    trace!("Requesting local model list from Ollama server...");
    let model_list = match ollama.list_local_models().await {
        Ok(m) => {
            debug!("Successfully retrieved {} models from Ollama server", m.len());
            trace!("Local models available: {:?}", m.iter().map(|model| &model.name).collect::<Vec<_>>());
            m
        }
        Err(e) => {
            error!("Failed to connect to Ollama server at {}: {}", formatted_host, e);
            let err_msg = format!("Error: Failed to connect to Ollama: {}", e);
            trace_message(&err_msg, channel_str.clone(), guild_str.clone()).await;
            ctx.say(err_msg).await?;
            return Ok(());
        }
    };

    let mut needs_create = true;
    let yesterday = Utc::now() - chrono::Duration::days(1);
    trace!("Checking model list against yesterday's cutoff timestamp ({})", yesterday);

    for model in &model_list {
        trace!("Inspecting model entry: '{}', modified_at: '{}'", model.name, model.modified_at);
        if model.name.starts_with("EwiAI") {
            debug!("Found matching model candidate: '{}'", model.name);
            match DateTime::parse_from_rfc3339(&model.modified_at) {
                Ok(modified) => {
                    let modified_utc = modified.with_timezone(&Utc);
                    trace!("Parsed model modified_at: {} (UTC: {})", modified, modified_utc);
                    if modified_utc > yesterday {
                        info!("Model '{}' is up to date (modified {} > cutoff {})", model.name, modified_utc, yesterday);
                        needs_create = false;
                    } else {
                        trace!("Model '{}' was modified at {}, which is older than cutoff {}", model.name, modified_utc, yesterday);
                    }
                }
                Err(e) => {
                    warn!("Failed to parse modified_at date '{}' for model '{}': {}", model.modified_at, model.name, e);
                }
            }
        }
    }

    if needs_create {
        info!("EwiAI model is missing or out of date. Initiating model creation/update...");
        let system_prompt = system_prompt();
        trace!("Generated system prompt for model creation (length: {} chars)", system_prompt.len());
        trace!("System prompt content: {}", system_prompt);
        let notice_msg = "EwiAI model is missing or out of date. Creating/updating model (this may take a bit)...";
        trace_message(notice_msg, channel_str.clone(), guild_str.clone()).await;
        let _ = ctx.say(notice_msg).await?;

        trace!("Sending create_model request to Ollama for 'EwiAI' based on 'gemma4:e2b-it-qat'...");
        if let Err(e) = ollama.create_model(CreateModelRequest::new("EwiAI".into())
            .system(system_prompt.into())
            .from_model("gemma4:e2b-it-qat".into())).await {
            error!("Failed to create model EwiAI: {}", e);
            let err_msg = format!("Error creating model: {}", e);
            trace_message(&err_msg, channel_str.clone(), guild_str.clone()).await;
            ctx.say(err_msg).await?;
            return Ok(());
        }
        info!("Successfully created/updated EwiAI model");
    } else {
        trace!("Skipping model creation; existing model is up to date.");
    }

    let mut final_prompt = String::new();
    let mut limit = messages_count.unwrap_or(5);
    trace!("messages_count limit: {}", limit);
    if limit > 10 {
        trace!("Clamping messages_count limit from {} to 10", limit);
        limit = 10;
    }
    trace!("Fetching last {} messages from channel {}", limit, channel_str);
    match ctx.channel_id().messages(ctx.http(), poise::serenity_prelude::GetMessages::new().limit(limit)).await {
        Ok(msgs) => {
            debug!("Retrieved {} messages from channel {}", msgs.len(), channel_str);
            final_prompt.push_str("Recent channel messages:\n");
            for msg in msgs.iter().rev() {
                let mut content = msg.content.clone();
                if content.len() > 100 {
                    content.truncate(100);
                    content.push_str("...");
                }
                trace!("Appending message from author {} (ID: {}): snippet='{}'", msg.author.name, msg.author.id, content);
                final_prompt.push_str(&format!("{} (ID: {}): {}\n", msg.author.name, msg.author.id, content));
            }
            final_prompt.push_str("\n");
        }
        Err(e) => {
            error!("Failed to get messages from channel {}: {:?}", channel_str, e);
        }
    }
    
    final_prompt.push_str("Please respond to the messages above.");
    trace!("Final prompt assembled (length: {} chars)", final_prompt.len());
    trace!("Final prompt: {}", final_prompt);

    let thinking_msg = "Thinking...\n-# The hardware this thing runs is really slow, expect a long wait";
    trace_message(thinking_msg, channel_str.clone(), guild_str.clone()).await;
    trace!("Sending initial thinking message...");
    let reply = ctx.say(thinking_msg).await?;

    let sys_prompt = system_prompt();
    trace!("Generating stream from Ollama model 'EwiAI' with system prompt length {}...", sys_prompt.len());
    let mut stream = match ollama.generate_stream(GenerationRequest::new("EwiAI".into(), final_prompt).system(sys_prompt)).await {
        Ok(s) => {
            debug!("Successfully initiated generation stream");
            s
        }
        Err(e) => {
            error!("Error starting generation stream with Ollama: {}", e);
            let err_msg = format!("Error during generation: {}", e);
            trace_message(&err_msg, channel_str.clone(), guild_str.clone()).await;
            reply.edit(ctx, CreateReply::default().content(err_msg)).await?;
            return Ok(());
        }
    };

    let mut response_text = String::new();
    let mut last_update = std::time::Instant::now();
    let mut final_response = None;
    let mut chunk_count = 0usize;

    trace!("Reading chunks from generation stream...");
    while let Some(res) = stream.next().await {
        match res {
            Ok(chunks) => {
                chunk_count += chunks.len();
                trace!("Received stream batch containing {} chunk(s)", chunks.len());
                for chunk in chunks {
                    trace!("Chunk response segment: '{}', done: {}", chunk.response, chunk.done);
                    response_text.push_str(&chunk.response);
                    if chunk.done {
                        debug!("Stream chunk marked done");
                        final_response = Some(chunk);
                    }
                }

                if last_update.elapsed() >= Duration::from_secs(1) && !response_text.is_empty() {
                    trace!("Throttled stream update: updating reply message (response_text length: {} chars)...", response_text.len());
                    let _ = reply.edit(ctx, CreateReply::default().content(&response_text)).await;
                    last_update = std::time::Instant::now();
                }
            }
            Err(e) => {
                error!("Error encountered while reading stream chunk: {}", e);
                let err_msg = format!("{} [Stream Error: {}]", response_text, e);
                trace_message(&err_msg, channel_str.clone(), guild_str.clone()).await;
                let _ = reply.edit(ctx, CreateReply::default().content(err_msg)).await;
                return Ok(());
            }
        }
    }
    info!("Stream completed. Total chunks received: {}, output length: {} chars", chunk_count, response_text.len());

    if let Some(stats) = final_response {
        let total_duration = stats.total_duration.unwrap_or(0) as f64 / 1_000_000_000.0;
        let eval_count = stats.eval_count.unwrap_or(0);
        let eval_duration = stats.eval_duration.unwrap_or(0) as f64 / 1_000_000_000.0;
        let tokens_per_sec = if eval_duration > 0.0 { eval_count as f64 / eval_duration } else { 0.0 };

        debug!("Generation stats: total_duration={:.2}s, eval_count={}, eval_duration={:.2}s, tok/s={:.2}", total_duration, eval_count, eval_duration, tokens_per_sec);

        let stats_text = format!(
            "\n\n*Generated in {:.2}s ({:.2} tok/s)*",
            total_duration, tokens_per_sec
        );
        response_text.push_str(&stats_text);
    } else {
        trace!("No final response stats available from stream");
    }
    
    trace_message(&response_text, channel_str.clone(), guild_str.clone()).await;
    trace!("Sending final edit to reply message...");
    let _ = reply.edit(ctx, CreateReply::default().content(&response_text)).await;
    debug!("answer command finished successfully for user {}", ctx.author().name);

    Ok(())
}

fn system_prompt() -> String {
    let now_str = Utc::now().to_string();
    trace!("Generating system_prompt with current UTC timestamp: {}", now_str);
    let prompt = format!("You are EwiAI, an AI in the TeenGovernment Development Server. Your developer is Ewi/ewenlau (discord user id: 713354021124964422, discord username: ewenlau). Be slightly unhelpful, in a sarcastic way, but still answer the question. The current date is {}", now_str);
    trace!("Generated system prompt: {}", prompt);
    prompt
}