//! Request body modification for intercepted LLM API calls.
//!
//! Aggressively strips everything except identity and actual conversation
//! from the Gemini API request. No integrity checks exist on the request
//! body — Google validates OAuth, project, model, and JSON structure only.

use regex::Regex;
use serde_json::Value;
use tracing::info;

/// Strip ALL tool definitions.
/// Set to false to allow tools through (for tool call testing).
const STRIP_ALL_TOOLS: bool = false;

/// Modify a streamGenerateContent request body in-place.
/// Returns the modified JSON bytes, or None if modification wasn't possible.
pub fn modify_request(body: &[u8]) -> Option<Vec<u8>> {
    let mut json: Value = serde_json::from_slice(body).ok()?;

    let original_size = body.len();
    let mut changes: Vec<String> = Vec::new();

    // ── 1. System instruction: keep ONLY <identity>, nuke everything else ──
    if let Some(sys) = json
        .pointer_mut("/request/systemInstruction/parts/0/text")
        .and_then(|v| v.as_str())
        .map(|s| s.to_string())
    {
        let original_len = sys.len();

        // Extract <identity>...</identity> block
        let identity = extract_xml_section(&sys, "identity");

        if let Some(identity_text) = identity {
            let new_sys = format!("<identity>\n{}\n</identity>", identity_text.trim());
            let stripped = original_len - new_sys.len();
            if stripped > 0 {
                changes.push(format!(
                    "system instruction: keep <identity> only ({original_len} → {} chars, -{stripped})",
                    new_sys.len()
                ));
                json["request"]["systemInstruction"]["parts"][0]["text"] =
                    Value::String(new_sys);
            }
        } else {
            // No identity tag found — clear the whole thing
            changes.push(format!("system instruction: cleared ({original_len} chars)"));
            json["request"]["systemInstruction"]["parts"][0]["text"] =
                Value::String(String::new());
        }
    }

    // ── 2. Content messages: keep only actual conversation turns ───────────
    if let Some(contents) = json
        .pointer_mut("/request/contents")
        .and_then(|v| v.as_array_mut())
    {
        let before = contents.len();

        // Remove messages that are pure Antigravity context injection
        contents.retain(|msg| {
            if let Some(text) = msg["parts"][0]["text"].as_str() {
                // Strip user_information (OS, workspace paths)
                if text.starts_with("<user_information>") {
                    return false;
                }
                // Strip user_rules / MEMORY blocks
                if text.starts_with("<user_rules>") {
                    return false;
                }
                // Strip workflows
                if text.starts_with("<workflows>") {
                    return false;
                }
                // Strip MCP servers block
                if text.starts_with("<mcp_servers>") {
                    return false;
                }
            }
            true
        });

        // For remaining messages, strip embedded metadata
        for msg in contents.iter_mut() {
            if let Some(text) = msg["parts"][0]["text"].as_str().map(|s| s.to_string()) {
                let mut modified = text.clone();

                // Strip conversation summaries block
                if let Some(cleaned) = strip_between(&modified, "# Conversation History\n", "</conversation_summaries>") {
                    modified = cleaned;
                }

                // Strip <ADDITIONAL_METADATA> blocks (cursor pos, open files, etc.)
                if let Some(cleaned) = strip_xml_section(&modified, "ADDITIONAL_METADATA") {
                    modified = cleaned;
                }

                // Strip <EPHEMERAL_MESSAGE> blocks
                if let Some(cleaned) = strip_xml_section(&modified, "EPHEMERAL_MESSAGE") {
                    modified = cleaned;
                }

                // Strip "Step Id: N\n" prefixes
                if modified.starts_with("Step Id:") {
                    if let Some(newline_pos) = modified.find('\n') {
                        modified = modified[newline_pos + 1..].to_string();
                    }
                }

                // Strip knowledge item blocks
                if let Some(cleaned) = strip_between(&modified, "Here are the ", "</knowledge_item>") {
                    // Only strip if it's about knowledge items
                    if cleaned.len() < modified.len() && modified.contains("knowledge item") {
                        modified = cleaned;
                    }
                }

                // Clean up excessive whitespace from stripping
                let modified = collapse_newlines(&modified);

                if modified.len() < text.len() {
                    msg["parts"][0]["text"] = Value::String(modified);
                }
            }
        }

        // Remove now-empty messages
        contents.retain(|msg| {
            if let Some(text) = msg["parts"][0]["text"].as_str() {
                !text.trim().is_empty()
            } else {
                true
            }
        });

        let removed = before - contents.len();
        if removed > 0 {
            changes.push(format!("remove {removed}/{before} content messages"));
        }
    }

    // ── 3. Strip all tool definitions ────────────────────────────────────
    if STRIP_ALL_TOOLS {
        if let Some(tools) = json
            .pointer_mut("/request/tools")
            .and_then(|v| v.as_array_mut())
        {
            let count = tools.len();
            if count > 0 {
                tools.clear();
                changes.push(format!("strip all {count} tools"));
            }
        }
    }

    // ── 4. Inject includeThoughts to capture thinking text ───────────────
    // Without this flag, Google only reports thinking token counts
    // but doesn't send the thinking text in SSE parts.
    {
        // Ensure request.generationConfig.thinkingConfig.includeThoughts = true
        let request = json.get_mut("request").and_then(|v| v.as_object_mut());
        if let Some(req) = request {
            let gen_config = req
                .entry("generationConfig")
                .or_insert_with(|| serde_json::json!({}));
            if let Some(gc) = gen_config.as_object_mut() {
                let thinking_config = gc
                    .entry("thinkingConfig")
                    .or_insert_with(|| serde_json::json!({}));
                if let Some(tc) = thinking_config.as_object_mut() {
                    if !tc.contains_key("includeThoughts") {
                        tc.insert("includeThoughts".to_string(), Value::Bool(true));
                        changes.push("inject includeThoughts".to_string());
                    }
                }
            }
        } else {
            // Not wrapped in request — try top-level (public API format)
            let gen_config = json.as_object_mut().and_then(|o| {
                Some(o.entry("generationConfig")
                    .or_insert_with(|| serde_json::json!({})))
            });
            if let Some(gc) = gen_config.and_then(|v| v.as_object_mut()) {
                let thinking_config = gc
                    .entry("thinkingConfig")
                    .or_insert_with(|| serde_json::json!({}));
                if let Some(tc) = thinking_config.as_object_mut() {
                    if !tc.contains_key("includeThoughts") {
                        tc.insert("includeThoughts".to_string(), Value::Bool(true));
                        changes.push("inject includeThoughts (top-level)".to_string());
                    }
                }
            }
        }
    }

    if changes.is_empty() {
        return None; // Nothing modified
    }

    let modified_bytes = serde_json::to_vec(&json).ok()?;
    let saved = original_size as i64 - modified_bytes.len() as i64;
    let pct = if original_size > 0 {
        (saved as f64 / original_size as f64 * 100.0) as i32
    } else {
        0
    };

    info!(
        original = original_size,
        modified = modified_bytes.len(),
        saved_bytes = saved,
        saved_pct = pct,
        "MITM: request modified [{}]",
        changes.join(", ")
    );

    Some(modified_bytes)
}

/// Extract the inner text of an XML-style section.
fn extract_xml_section(text: &str, tag: &str) -> Option<String> {
    let open = format!("<{tag}>");
    let close = format!("</{tag}>");
    let start = text.find(&open)?;
    let end = text.find(&close)?;
    let inner_start = start + open.len();
    if inner_start >= end {
        return None;
    }
    Some(text[inner_start..end].to_string())
}

/// Strip an XML-style section and return the modified text.
fn strip_xml_section(text: &str, tag: &str) -> Option<String> {
    let open = format!("<{tag}>");
    let close = format!("</{tag}>");
    let start = text.find(&open)?;
    let end = text.find(&close)?;
    let end_pos = end + close.len();
    Some(format!("{}{}", &text[..start], &text[end_pos..]))
}

/// Strip everything between two markers (inclusive of markers).
fn strip_between(text: &str, start_marker: &str, end_marker: &str) -> Option<String> {
    let start = text.find(start_marker)?;
    let end = text.find(end_marker)?;
    let end_pos = end + end_marker.len();
    // Skip any trailing whitespace after end marker
    let rest = text[end_pos..].trim_start();
    Some(format!("{}{}", &text[..start], rest))
}

/// Collapse 3+ consecutive newlines into 2.
fn collapse_newlines(text: &str) -> String {
    let re = Regex::new(r"\n{3,}").unwrap();
    re.replace_all(text, "\n\n").to_string()
}

/// Dechunk an HTTP chunked-encoded body into raw bytes.
pub fn dechunk(data: &[u8]) -> Vec<u8> {
    let mut result = Vec::with_capacity(data.len());
    let mut pos = 0;

    while pos < data.len() {
        let line_end = match data[pos..].windows(2).position(|w| w == b"\r\n") {
            Some(p) => pos + p,
            None => break,
        };

        let size_str = std::str::from_utf8(&data[pos..line_end])
            .unwrap_or("")
            .split(';')
            .next()
            .unwrap_or("")
            .trim();

        let chunk_size = match usize::from_str_radix(size_str, 16) {
            Ok(0) => break,
            Ok(n) => n,
            Err(_) => break,
        };

        let data_start = line_end + 2;
        let data_end = (data_start + chunk_size).min(data.len());
        result.extend_from_slice(&data[data_start..data_end]);

        pos = data_end + 2;
    }

    result
}

/// Re-encode data as a single HTTP chunk + terminal chunk.
pub fn rechunk(data: &[u8]) -> Vec<u8> {
    let hex_size = format!("{:x}", data.len());
    let mut result = Vec::with_capacity(hex_size.len() + 2 + data.len() + 2 + 5);
    result.extend_from_slice(hex_size.as_bytes());
    result.extend_from_slice(b"\r\n");
    result.extend_from_slice(data);
    result.extend_from_slice(b"\r\n0\r\n\r\n");
    result
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_dechunk_basic() {
        let chunked = b"5\r\nhello\r\n6\r\n world\r\n0\r\n\r\n";
        let result = dechunk(chunked);
        assert_eq!(result, b"hello world");
    }

    #[test]
    fn test_dechunk_single() {
        let chunked = b"b\r\nhello world\r\n0\r\n\r\n";
        let result = dechunk(chunked);
        assert_eq!(result, b"hello world");
    }

    #[test]
    fn test_rechunk() {
        let data = b"hello world";
        let chunked = rechunk(data);
        let expected = b"b\r\nhello world\r\n0\r\n\r\n";
        assert_eq!(chunked, expected);
    }

    #[test]
    fn test_dechunk_rechunk_roundtrip() {
        let original = b"5\r\nhello\r\n6\r\n world\r\n0\r\n\r\n";
        let data = dechunk(original);
        let rechunked = rechunk(&data);
        let data2 = dechunk(&rechunked);
        assert_eq!(data, data2);
    }

    #[test]
    fn test_modify_strips_all_tools() {
        let body = serde_json::json!({
            "project": "test",
            "requestId": "test/1",
            "request": {
                "contents": [{"role": "user", "parts": [{"text": "hello"}]}],
                "tools": [
                    {"functionDeclarations": [{"name": "view_file", "description": "view", "parameters": {}}]},
                    {"functionDeclarations": [{"name": "browser_subagent", "description": "browse", "parameters": {}}]},
                ],
                "generationConfig": {}
            },
            "model": "test"
        });

        let bytes = serde_json::to_vec(&body).unwrap();
        let modified = modify_request(&bytes).unwrap();
        let result: Value = serde_json::from_slice(&modified).unwrap();

        let tools = result["request"]["tools"].as_array().unwrap();
        assert!(tools.is_empty(), "all tools should be stripped");
    }

    #[test]
    fn test_modify_keeps_only_identity() {
        let sys_text = "<identity>\nYou are a helpful AI.\n</identity>\n\n<tool_calling>\nUse absolute paths.\n</tool_calling>\n<web_application_development>\nlots of web dev stuff\n</web_application_development>\n<communication_style>\nbe helpful\n</communication_style>";
        let body = serde_json::json!({
            "project": "test",
            "requestId": "test/1",
            "request": {
                "contents": [{"role": "user", "parts": [{"text": "hello"}]}],
                "systemInstruction": {"parts": [{"text": sys_text}]},
                "tools": [],
                "generationConfig": {}
            },
            "model": "test"
        });

        let bytes = serde_json::to_vec(&body).unwrap();
        let modified = modify_request(&bytes).unwrap();
        let result: Value = serde_json::from_slice(&modified).unwrap();

        let new_sys = result["request"]["systemInstruction"]["parts"][0]["text"]
            .as_str()
            .unwrap();

        assert!(new_sys.contains("<identity>"));
        assert!(new_sys.contains("You are a helpful AI."));
        assert!(!new_sys.contains("tool_calling"));
        assert!(!new_sys.contains("web_application_development"));
        assert!(!new_sys.contains("communication_style"));
    }

    #[test]
    fn test_modify_strips_context_messages() {
        let body = serde_json::json!({
            "project": "test",
            "requestId": "test/1",
            "request": {
                "contents": [
                    {"role": "user", "parts": [{"text": "<user_information>\nLinux\n</user_information>"}]},
                    {"role": "user", "parts": [{"text": "<user_rules>\nno rules\n</user_rules>"}]},
                    {"role": "user", "parts": [{"text": "<workflows>\nsome workflows\n</workflows>"}]},
                    {"role": "user", "parts": [{"text": "Step Id: 0\n\n<USER_REQUEST>\nSay hello\n</USER_REQUEST>\n<ADDITIONAL_METADATA>\ncursor stuff\n</ADDITIONAL_METADATA>"}]},
                    {"role": "model", "parts": [{"text": "Hello!"}]},
                ],
                "tools": [],
                "generationConfig": {}
            },
            "model": "test"
        });

        let bytes = serde_json::to_vec(&body).unwrap();
        let modified = modify_request(&bytes).unwrap();
        let result: Value = serde_json::from_slice(&modified).unwrap();

        let contents = result["request"]["contents"].as_array().unwrap();
        // Should have removed user_information, user_rules, workflows (3 messages)
        // Kept: USER_REQUEST message (with ADDITIONAL_METADATA stripped) + model response
        assert_eq!(contents.len(), 2, "should keep only user request + model response");

        // Check USER_REQUEST message had metadata stripped
        let user_msg = contents[0]["parts"][0]["text"].as_str().unwrap();
        assert!(user_msg.contains("Say hello"), "should keep user request");
        assert!(!user_msg.contains("ADDITIONAL_METADATA"), "should strip metadata");
        assert!(!user_msg.contains("cursor stuff"), "should strip cursor info");
        assert!(!user_msg.starts_with("Step Id:"), "should strip step id");

        // Model response kept intact
        assert_eq!(contents[1]["parts"][0]["text"].as_str().unwrap(), "Hello!");
    }

    #[test]
    fn test_extract_xml_section() {
        let text = "before <identity>\nI am AI\n</identity> after";
        let result = extract_xml_section(text, "identity").unwrap();
        assert_eq!(result, "\nI am AI\n");
    }

    #[test]
    fn test_strip_xml_section() {
        let text = "before <META>\nstuff\n</META> after";
        let result = strip_xml_section(text, "META").unwrap();
        assert_eq!(result, "before  after");
    }

    #[test]
    fn test_strip_between() {
        let text = "keep this # Conversation History\nlots of stuff\n</conversation_summaries>\nand this";
        let result = strip_between(text, "# Conversation History\n", "</conversation_summaries>").unwrap();
        assert_eq!(result, "keep this and this");
    }
}