diff --git a/src/bin/codegen.rs b/src/bin/codegen.rs
new file mode 100644
index 0000000..9ce10f4
--- /dev/null
+++ b/src/bin/codegen.rs
@@ -0,0 +1,343 @@
+use std::fs;
+
+use serde::{Deserialize, Serialize};
+
+#[derive(Debug, Deserialize)]
+enum ValueType {
+    #[serde(rename = "string")]
+    String,
+    #[serde(rename = "number")]
+    Number,
+    #[serde(skip_deserializing)]
+    Array,
+    #[serde(skip_deserializing)]
+    ArrayElement,
+}
+
+#[derive(Debug, Serialize, Deserialize)]
+#[serde(untagged)]
+enum Key {
+    String(String),
+    Num(i64),
+}
+
+#[derive(Debug, Deserialize)]
+struct Data {
+    keys: Vec<Key>,
+    value_type: ValueType,
+}
+
+const PRAGMA: &str = "pragma circom 2.1.9;\n\n";
+
+fn extract_string(data: Data, cfb: &mut String) {
+    *cfb += "template ExtractValue2(DATA_BYTES, MAX_STACK_HEIGHT, ";
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => *cfb += &format!("keyLen{}, depth{}, ", i + 1, i + 1),
+            Key::Num(_) => *cfb += &format!("index{}, depth{}, ", i + 1, i + 1),
+        }
+    }
+    *cfb += "maxValueLen) {\n";
+
+    *cfb += "    signal input data[DATA_BYTES];\n\n";
+
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => *cfb += &format!("    signal input key{}[keyLen{}];\n", i + 1, i + 1),
+            _ => (),
+        }
+    }
+
+    *cfb += r#"
+    signal output value[maxValueLen];
+
+    signal value_starting_index[DATA_BYTES];
+
+    value_starting_index <== ExtractMultiDepthNestedObject(DATA_BYTES, MAX_STACK_HEIGHT, keyLen1, depth1, keyLen2, depth2, index3, depth3, index4, depth4, maxValueLen)(data, key1, key2);
+
+    log("value_starting_index", value_starting_index[DATA_BYTES-2]);
+    // TODO: why +1 not required here,when required on all other string implss?
+    value <== SelectSubArray(DATA_BYTES, maxValueLen)(data, value_starting_index[DATA_BYTES-2], maxValueLen);
+
+    for (var i=0 ; i<maxValueLen; i++) {
+        log("value[",i,"]=", value[i]);
+    }
+}
+"#;
+}
+
+fn extract_number(data: Data, cfb: &mut String) {
+    *cfb += "template ExtractValue2(DATA_BYTES, MAX_STACK_HEIGHT, ";
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => *cfb += &format!("keyLen{}, depth{}, ", i + 1, i + 1),
+            Key::Num(_) => *cfb += &format!("index{}, depth{}, ", i + 1, i + 1),
+        }
+    }
+    *cfb += "maxValueLen) {\n";
+
+    *cfb += "    signal input data[DATA_BYTES];\n\n";
+
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => *cfb += &format!("    signal input key{}[keyLen{}];\n", i + 1, i + 1),
+            _ => (),
+        }
+    }
+
+    *cfb += r#"
+    signal value_string[maxValueLen];
+    signal output value;
+
+    signal value_starting_index[DATA_BYTES];
+
+    value_starting_index <== ExtractMultiDepthNestedObject(DATA_BYTES, MAX_STACK_HEIGHT, keyLen1, depth1, keyLen2, depth2, index3, depth3, index4, depth4, maxValueLen)(data, key1, key2);
+
+    log("value_starting_index", value_starting_index[DATA_BYTES-2]);
+    // TODO: why +1 not required here,when required on all other string implss?
+    value_string <== SelectSubArray(DATA_BYTES, maxValueLen)(data, value_starting_index[DATA_BYTES-2], maxValueLen);
+
+    for (var i=0 ; i<maxValueLen; i++) {
+        log("value[",i,"]=", value_string[i]);
+    }
+
+    signal number_value[maxValueLen];
+    number_value[0] <== (value_string[0]-48);
+    for (var i=1 ; i<maxValueLen ; i++) {
+        number_value[i] <== number_value[i-1] * 10 + (value_string[i]-48);
+    }
+
+    value <== number_value[maxValueLen-1];
+"#;
+}
+
+fn parse_json_request() -> std::io::Result<()> {
+    let request = r#"
+    {
+        "keys": ["a"],
+        "value_type": "string"
+    }
+    "#;
+
+    let data: Data = serde_json::from_str(request)?;
+    // let key_bytes = data
+    //     .keys
+    //     .iter()
+    //     .map(|k| match k {
+    //         Key::String(key) => key.as_bytes().to_owned(),
+    //         Key::Num(num) => num.to_string().as_bytes().to_owned(),
+    //     })
+    //     .collect::<Vec<Vec<u8>>>();
+    println!("{:?}", data);
+
+    let mut cfb = String::new();
+    cfb += PRAGMA;
+    cfb += "import ./fetcher.circom;\n\n";
+
+    cfb += "template ExtractValue2(DATA_BYTES, MAX_STACK_HEIGHT,";
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => cfb += &format!("keyLen{}, depth{}, ", i + 1, i + 1),
+            Key::Num(_) => cfb += &format!("index{}, depth{}, ", i + 1, i + 1),
+        }
+    }
+    cfb += "maxValueLen) {\n";
+
+    cfb += "    signal input data[DATA_BYTES];\n\n";
+
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => cfb += &format!("    signal input key{}[keyLen{}];\n", i + 1, i + 1),
+            _ => (),
+        }
+    }
+
+    cfb += r#"
+    signal output value_starting_index[DATA_BYTES];
+
+    signal mask[DATA_BYTES];
+    // mask[0] <== 0;
+
+    var logDataLen = log2Ceil(DATA_BYTES);
+
+    component State[DATA_BYTES];
+    State[0] = StateUpdate(MAX_STACK_HEIGHT);
+    State[0].byte           <== data[0];
+    for(var i = 0; i < MAX_STACK_HEIGHT; i++) {
+        State[0].stack[i]   <== [0,0];
+    }
+    State[0].parsing_string <== 0;
+    State[0].parsing_number <== 0;
+
+    signal parsing_key[DATA_BYTES];
+    signal parsing_value[DATA_BYTES];
+"#;
+
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => {
+                cfb += &format!("    signal parsing_object{}_value[DATA_BYTES];\n", i + 1)
+            }
+            Key::Num(_) => cfb += &format!("    signal parsing_array{}[DATA_BYTES];\n", i + 1),
+        }
+    }
+
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => cfb += &format!("    signal is_key{}_match[DATA_BYTES];\n    signal is_key{}_match_for_value[DATA_BYTES];\n    is_key{}_match_for_value[0] <== 0;\n    signal is_next_pair_at_depth{}[DATA_BYTES];\n", i+1, i+1, i+1, i+1),
+            _ => (),
+        }
+    }
+
+    cfb += r#"
+    signal is_value_match[DATA_BYTES];
+    is_value_match[0] <== 0;
+    signal value_mask[DATA_BYTES];
+    for(var data_idx = 1; data_idx < DATA_BYTES; data_idx++) {
+        // Debugging
+        for(var i = 0; i<MAX_STACK_HEIGHT; i++) {
+            log("State[", data_idx-1, "].stack[", i,"]    ", "= [",State[data_idx-1].next_stack[i][0], "][", State[data_idx-1].next_stack[i][1],"]" );
+        }
+        log("State[", data_idx-1, "].byte", "= ", data[data_idx-1]);
+        log("State[", data_idx-1, "].parsing_string", "= ", State[data_idx-1].next_parsing_string);
+        log("State[", data_idx-1, "].parsing_number", "= ", State[data_idx-1].next_parsing_number);
+
+        State[data_idx]                  = StateUpdate(MAX_STACK_HEIGHT);
+        State[data_idx].byte           <== data[data_idx];
+        State[data_idx].stack          <== State[data_idx - 1].next_stack;
+        State[data_idx].parsing_string <== State[data_idx - 1].next_parsing_string;
+        State[data_idx].parsing_number <== State[data_idx - 1].next_parsing_number;
+
+        // - parsing key
+        // - parsing value (different for string/numbers and array)
+        // - key match (key 1, key 2)
+        // - is next pair
+        // - is key match for value
+        // - value_mask
+        // - mask
+
+        parsing_key[data_idx-1] <== InsideKey(MAX_STACK_HEIGHT)(State[data_idx].stack, State[data_idx].parsing_string, State[data_idx].parsing_number);
+        // log("parsing key:", parsing_key[data_idx]);
+
+"#;
+
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => {
+                cfb += &format!("        parsing_object{}_value[data_idx-1] <== InsideObjectAtDepth(MAX_STACK_HEIGHT, depth{})(State[data_idx].stack, State[data_idx].parsing_string, State[data_idx].parsing_number);\n", i+1, i+1);
+            }
+            Key::Num(_) => {
+                cfb += &format!("        parsing_array{}[data_idx-1] <== InsideArrayIndexAtDepth(MAX_STACK_HEIGHT, index{}, depth{})(State[data_idx].stack, State[data_idx].parsing_string, State[data_idx].parsing_number);\n", i+1, i+1, i+1);
+            }
+        }
+    }
+
+    cfb += &format!(
+        "        parsing_value[data_idx-1] <== MultiAnd({})([",
+        data.keys.len()
+    );
+
+    for (i, key) in data.keys.iter().take(data.keys.len() - 1).enumerate() {
+        match key {
+            Key::String(_) => cfb += &format!("parsing_object{}_value[data_idx-1], ", i + 1),
+            Key::Num(_) => cfb += &format!("parsing_array{}[data_idx-1], ", i + 1),
+        }
+    }
+    match data.keys[data.keys.len() - 1] {
+        Key::String(_) => {
+            cfb += &format!("parsing_object{}_value[data_idx-1]]);\n", data.keys.len())
+        }
+        Key::Num(_) => cfb += &format!("parsing_array{}[data_idx-1]]);\n)", data.keys.len()),
+    }
+
+    // optional debug logs
+    cfb += "        // log(\"parsing value:\", ";
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => cfb += &format!("parsing_object{}_value[data_idx-1], ", i + 1),
+            Key::Num(_) => cfb += &format!("parsing_array{}[data_idx-1], ", i + 1),
+        }
+    }
+    cfb += "parsing_value[data_idx-1]);\n\n";
+
+    let mut num_objects = 0;
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => {
+                num_objects += 1;
+                cfb += &format!("        is_key{}_match[data_idx-1] <== KeyMatchAtDepth(DATA_BYTES, MAX_STACK_HEIGHT, keyLen{}, depth{})(data, key{}, 100, data_idx-1, parsing_key[data_idx-1], State[data_idx-1].stack);\n", i+1, i+1, i+1, i+1);
+                cfb += &format!("        is_next_pair_at_depth{}[data_idx-1] <== NextKVPairAtDepth(MAX_STACK_HEIGHT, depth{})(State[data_idx-1].stack, data[data_idx-1]);\n", i+1, i+1);
+                cfb += &format!("        is_key{}_match_for_value[data_idx] <== Mux1()([is_key{}_match_for_value[data_idx-1] * (1-is_next_pair_at_depth[data_idx-1]), is_key{}_match[data_idx-1] * (1-is_next_pair_at_depth{}[data_idx-1])], is_key{}_match[data_idx-1]);\n", i+1, i+1, i+1, i+1, i+1);
+            }
+            _ => (),
+        }
+    }
+
+    cfb += &format!(
+        "        is_value_match[data_idx] <== MultiAnd({})([",
+        num_objects
+    );
+    for (i, key) in data.keys.iter().enumerate() {
+        match key {
+            Key::String(_) => cfb += &format!("is_key{}_match_for_value[data_idx], ", i + 1),
+            Key::Num(_) => (),
+        }
+    }
+
+    // remove last 2 chars `, ` from string buffer
+    cfb.pop();
+    cfb.pop();
+    cfb += "]);\n";
+
+    cfb += r#"        // log("is_value_match", is_value_match[data_idx]);
+
+        // mask[i] = data[i] * parsing_value[i] * is_key_match_for_value[i]
+        value_mask[data_idx-1] <== data[data_idx-1] * parsing_value[data_idx-1];
+        mask[data_idx-1] <== value_mask[data_idx-1] * is_value_match[data_idx];
+        log("mask", mask[data_idx-1]);
+        log("xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx");
+    }
+
+    // Debugging
+    for(var i = 0; i < MAX_STACK_HEIGHT; i++) {
+        log("State[", DATA_BYTES-1, "].stack[", i,"]    ", "= [",State[DATA_BYTES -1].next_stack[i][0], "][", State[DATA_BYTES - 1].next_stack[i][1],"]" );
+    }
+    log("State[", DATA_BYTES-1, "].parsing_string", "= ", State[DATA_BYTES-1].next_parsing_string);
+    log("State[", DATA_BYTES-1, "].parsing_number", "= ", State[DATA_BYTES-1].next_parsing_number);
+    log("xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx");
+
+    // signal value_starting_index[DATA_BYTES];
+    signal is_zero_mask[DATA_BYTES];
+    signal is_prev_starting_index[DATA_BYTES];
+    value_starting_index[0] <== 0;
+    is_zero_mask[0] <== IsZero()(mask[0]);
+    for (var i=1 ; i<DATA_BYTES-1 ; i++) {
+        is_zero_mask[i] <== IsZero()(mask[i]);
+        is_prev_starting_index[i] <== IsZero()(value_starting_index[i-1]);
+        value_starting_index[i] <== value_starting_index[i-1] + i * (1-is_zero_mask[i]) * is_prev_starting_index[i];
+    }
+"#;
+
+    // template ends
+    cfb += "}\n";
+
+    match data.value_type {
+        ValueType::String => extract_string(data, &mut cfb),
+        ValueType::Number => extract_number(data, &mut cfb),
+        _ => unimplemented!(),
+    }
+
+    // write circuits to file
+    let mut file_path = std::env::current_dir()?;
+    file_path.push("circuits");
+    file_path.push("extractor.circom");
+
+    println!("file_path: {:?}", file_path);
+    fs::write(file_path, cfb)?;
+    Ok(())
+}
+
+pub fn main() -> std::io::Result<()> {
+    parse_json_request()?;
+    Ok(())
+}