diff --git a/src/cmd/describegpt.rs b/src/cmd/describegpt.rs
index 2ce85729b..dfac8e090 100644
--- a/src/cmd/describegpt.rs
+++ b/src/cmd/describegpt.rs
@@ -391,6 +391,7 @@ pub fn run(argv: &[&str]) -> CliResult<()> {
     )?;
     // safety: we just checked that there is at least one input file
     let arg_input = work_input[0]
+        .canonicalize()?
         .clone()
         .into_os_string()
         .into_string()
@@ -404,15 +405,18 @@ pub fn run(argv: &[&str]) -> CliResult<()> {
         return fail!("Error: --all option cannot be specified with other inference flags.");
     }
 
+    // Get qsv executable's path
+    let root = env::current_exe().unwrap();
+
     // Get stats from qsv stats on input file with --everything flag
     eprintln!("Generating stats from {arg_input} using qsv stats --everything...");
-    let Ok(stats) = Command::new("qsv")
+    let Ok(stats) = Command::new(root.clone())
         .arg("stats")
         .arg("--everything")
         .arg(arg_input.clone())
         .output()
     else {
-        return fail!("Error: Unable to parse stats as &str.");
+        return fail!("Error: Error while generating stats.");
     };
 
     // Parse the stats as &str
@@ -422,8 +426,8 @@ pub fn run(argv: &[&str]) -> CliResult<()> {
 
     // Get frequency from qsv frequency on input file
     eprintln!("Generating frequency from {arg_input} using qsv frequency...");
-    let Ok(frequency) = Command::new("qsv").arg("frequency").arg(arg_input).output() else {
-        return fail!("Error: Unable to get frequency from qsv.");
+    let Ok(frequency) = Command::new(root).arg("frequency").arg(arg_input).output() else {
+        return fail!("Error: Error while generating frequency.");
     };
 
     // Parse the frequency as &str
diff --git a/tests/test_describegpt.rs b/tests/test_describegpt.rs
index f4bcb7474..0b4e1de37 100644
--- a/tests/test_describegpt.rs
+++ b/tests/test_describegpt.rs
@@ -1,8 +1,12 @@
-use crate::{qcheck, workdir::Workdir, CsvData};
+use crate::workdir::Workdir;
 
+// Providing an invalid API key with --openai-key without
+// the environment variable set should result in an error
 #[test]
-fn count_simple() {
-    let wrk = Workdir::new("count_simple");
+// #[ignore = "Requires environment variable to NOT be set."]
+fn describegpt_invalid_api_key() {
+    let wrk = Workdir::new("describegpt");
+    // Create a CSV file with sample data
     wrk.create_indexed(
         "in.csv",
         vec![
@@ -12,172 +16,110 @@ fn count_simple() {
             svec!["gamma", "37"],
         ],
     );
-    let mut cmd = wrk.command("count");
-    cmd.arg("in.csv");
 
-    let got: String = wrk.stdout(&mut cmd);
-    let expected = "3";
-    assert_eq!(got, expected.to_string());
+    // Run the command
+    let mut cmd = wrk.command("describegpt");
+    cmd.arg("in.csv")
+        .arg("--all")
+        .arg("--json")
+        .args(["--openai-key", "INVALIDKEY"])
+        .args(["--max-tokens", "1000"]);
+
+    // Error message
+    let got_stderr = wrk.output_stderr(&mut cmd);
+    // Check that we receive the correct error message
+    assert!(got_stderr.contains("Incorrect API key provided: INVALIDKEY"));
 }
 
+// Verify --user-agent is passed to OpenAI
 #[test]
-fn count_width() {
-    let wrk = Workdir::new("count_width");
+#[ignore = "Requires environment variable to be set."]
+fn describegpt_user_agent() {
+    let wrk = Workdir::new("describegpt");
+    // Create a CSV file with sample data
     wrk.create_indexed(
         "in.csv",
         vec![
             svec!["letter", "number"],
             svec!["alpha", "13"],
             svec!["beta", "24"],
+            svec!["gamma", "37"],
         ],
     );
-    let mut cmd = wrk.command("count");
-    cmd.arg("--width").arg("in.csv");
 
-    let got: String = wrk.stdout(&mut cmd);
-    let expected = "2;9";
-    assert_eq!(got, expected.to_string());
+    // Run the command
+    let mut cmd = wrk.command("describegpt");
+    cmd.arg("in.csv")
+        .arg("--all")
+        .arg("--json")
+        .args([
+            "--user-agent",
+            "Mozilla/5.0 (platform; rv:geckoversion) Gecko/geckotrail Firefox/firefoxversion",
+        ])
+        .args(["--max-tokens", "1000"]);
+
+    // Check that the command ran successfully
+    wrk.assert_success(&mut cmd);
 }
 
+// Valid use of describegpt
 #[test]
-fn count_comments() {
-    let wrk = Workdir::new("count_comments");
+#[ignore = "Requires environment variable to be set."]
+fn describegpt_valid() {
+    let wrk = Workdir::new("describegpt");
 
-    wrk.create(
+    // Create a CSV file with sample data
+    wrk.create_indexed(
         "in.csv",
         vec![
-            svec!["# this is a comment", ""],
-            svec!["# next comment", ""],
             svec!["letter", "number"],
             svec!["alpha", "13"],
             svec!["beta", "24"],
-            svec!["# comment here too!", "24"],
+            svec!["gamma", "37"],
         ],
     );
-    let mut cmd = wrk.command("count");
-    cmd.arg("in.csv").env("QSV_COMMENT_CHAR", "#");
-
-    let got: String = wrk.stdout(&mut cmd);
-    let expected = "2";
-    assert_eq!(got, expected.to_string());
-}
-
-/// This tests whether `qsv count` gets the right answer.
-///
-/// It does some simple case analysis to handle whether we want to test counts
-/// in the presence of headers and/or indexes.
-fn prop_count_len(
-    name: &str,
-    rows: CsvData,
-    headers: bool,
-    idx: bool,
-    noheaders_env: bool,
-    human_readable: bool,
-) -> bool {
-    let mut expected_count = rows.len();
-    if headers && expected_count > 0 {
-        expected_count -= 1;
-    }
-
-    let wrk = Workdir::new(name);
-    if idx {
-        wrk.create_indexed("in.csv", rows);
-    } else {
-        wrk.create("in.csv", rows);
-    }
-
-    let mut cmd = wrk.command("count");
-    if !headers {
-        cmd.arg("--no-headers");
-    }
-    if noheaders_env {
-        cmd.env("QSV_NO_HEADERS", "1");
-    }
-    if human_readable {
-        cmd.arg("--human-readable");
-    }
-    cmd.arg("in.csv");
 
-    if human_readable {
-        use thousands::Separable;
+    // Run the command
+    let mut cmd = wrk.command("describegpt");
+    cmd.arg("in.csv")
+        .arg("--all")
+        .args(["--max-tokens", "1000"]);
 
-        let got_count: String = wrk.stdout(&mut cmd);
-        let expected_count_commas = expected_count.separate_with_commas();
-
-        rassert_eq!(got_count, expected_count_commas)
-    } else {
-        let got_count: usize = wrk.stdout(&mut cmd);
-        rassert_eq!(got_count, expected_count)
-    }
-}
-
-#[test]
-fn prop_count() {
-    fn p(rows: CsvData) -> bool {
-        prop_count_len("prop_count", rows, false, false, false, false)
-    }
-    qcheck(p as fn(CsvData) -> bool);
+    // Check that the command ran successfully
+    wrk.assert_success(&mut cmd);
 }
 
+// Valid use of describegpt with --json
 #[test]
-fn prop_count_human_readable() {
-    fn p(rows: CsvData) -> bool {
-        prop_count_len("prop_count", rows, false, false, false, true)
-    }
-    qcheck(p as fn(CsvData) -> bool);
-}
-
-#[test]
-fn prop_count_headers() {
-    fn p(rows: CsvData) -> bool {
-        prop_count_len("prop_count_headers", rows, true, false, false, false)
-    }
-    qcheck(p as fn(CsvData) -> bool);
-}
+#[ignore = "Requires environment variable to be set."]
+fn describegpt_valid_json() {
+    let wrk = Workdir::new("describegpt");
 
-#[test]
-fn prop_count_headers_human_readable() {
-    fn p(rows: CsvData) -> bool {
-        prop_count_len("prop_count_headers", rows, true, false, false, true)
-    }
-    qcheck(p as fn(CsvData) -> bool);
-}
-
-#[test]
-fn prop_count_indexed() {
-    fn p(rows: CsvData) -> bool {
-        prop_count_len("prop_count_indexed", rows, false, true, false, false)
-    }
-    qcheck(p as fn(CsvData) -> bool);
-}
-
-#[test]
-fn prop_count_indexed_headers() {
-    fn p(rows: CsvData) -> bool {
-        prop_count_len("prop_count_indexed_headers", rows, true, true, false, false)
-    }
-    qcheck(p as fn(CsvData) -> bool);
-}
+    // Create a CSV file with sample data
+    wrk.create_indexed(
+        "in.csv",
+        vec![
+            svec!["letter", "number"],
+            svec!["alpha", "13"],
+            svec!["beta", "24"],
+            svec!["gamma", "37"],
+        ],
+    );
 
-#[test]
-fn prop_count_noheaders_env() {
-    fn p(rows: CsvData) -> bool {
-        prop_count_len("prop_count_noheaders_env", rows, false, false, true, false)
+    // Run the command
+    let mut cmd = wrk.command("describegpt");
+    cmd.arg("in.csv")
+        .arg("--all")
+        .arg("--json")
+        .args(["--max-tokens", "1000"]);
+
+    // Check that the output is valid JSON
+    let got = wrk.stdout::<String>(&mut cmd);
+    match serde_json::from_str::<serde_json::Value>(&got) {
+        Ok(_) => (),
+        Err(e) => assert!(false, "Error parsing JSON: {e}"),
     }
-    qcheck(p as fn(CsvData) -> bool);
-}
 
-#[test]
-fn prop_count_noheaders_indexed_env() {
-    fn p(rows: CsvData) -> bool {
-        prop_count_len(
-            "prop_count_noheaders_indexed_env",
-            rows,
-            false,
-            true,
-            true,
-            false,
-        )
-    }
-    qcheck(p as fn(CsvData) -> bool);
+    // Check that the command ran successfully
+    wrk.assert_success(&mut cmd);
 }
diff --git a/tests/tests.rs b/tests/tests.rs
index bc613f7e0..f2ef12b7e 100644
--- a/tests/tests.rs
+++ b/tests/tests.rs
@@ -42,6 +42,7 @@ mod test_combos;
 mod test_comments;
 mod test_count;
 mod test_dedup;
+mod test_describegpt;
 mod test_diff;
 #[cfg(any(feature = "feature_capable", feature = "lite"))]
 mod test_enumerate;