diff --git a/.gitignore b/.gitignore
index 4bd6cec..670b156 100644
--- a/.gitignore
+++ b/.gitignore
@@ -7,6 +7,7 @@
 
 # Output
 /dist
+/venv
 
 pantograph/pantograph
 pantograph/lean-toolchain
diff --git a/.gitmodules b/.gitmodules
index caa4dcb..6b49ada 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -1,3 +1,4 @@
 [submodule "src"]
 	path = src
 	url = https://git.leni.sh/aniva/Pantograph.git
+	
\ No newline at end of file
diff --git a/README.md b/README.md
index f96dc56..19188f1 100644
--- a/README.md
+++ b/README.md
@@ -3,6 +3,14 @@
 Python interface to the Pantograph library
 
 ## Getting started
+Update submodule
+``` bash
+git submodule update --init
+```
+Install dependencies
+```bash
+poetry install
+```
 
 <!-- First initialize the git submodules so that git can keep track of the submodules being used do:
 ```bash
diff --git a/examples/Example/.gitignore b/examples/Example/.gitignore
new file mode 100644
index 0000000..1824d0c
--- /dev/null
+++ b/examples/Example/.gitignore
@@ -0,0 +1,3 @@
+/build
+/lakefile.olean
+/lake-packages/*
diff --git a/examples/Example/Example.lean b/examples/Example/Example.lean
new file mode 100644
index 0000000..40cb3dc
--- /dev/null
+++ b/examples/Example/Example.lean
@@ -0,0 +1,13 @@
+import Aesop
+
+-- Ensure that Aesop is running
+example : α → α :=
+  by aesop
+
+example : ∀ (p q: Prop), p ∨ q → q ∨ p := by
+  intro p q h
+  cases h
+  . apply Or.inr
+    assumption
+  . apply Or.inl
+    assumption
diff --git a/examples/Example/lake-manifest.json b/examples/Example/lake-manifest.json
new file mode 100644
index 0000000..06e1c0e
--- /dev/null
+++ b/examples/Example/lake-manifest.json
@@ -0,0 +1,23 @@
+{"version": 7,
+ "packagesDir": ".lake/packages",
+ "packages":
+ [{"url": "https://github.com/leanprover/std4",
+   "type": "git",
+   "subDir": null,
+   "rev": "3025cb124492b423070f20cf0a70636f757d117f",
+   "name": "std",
+   "manifestFile": "lake-manifest.json",
+   "inputRev": "main",
+   "inherited": true,
+   "configFile": "lakefile.lean"},
+  {"url": "https://github.com/leanprover-community/aesop.git",
+   "type": "git",
+   "subDir": null,
+   "rev": "0a21a48c286c4a4703c0be6ad2045f601f31b1d0",
+   "name": "aesop",
+   "manifestFile": "lake-manifest.json",
+   "inputRev": "v4.8.0-rc1",
+   "inherited": false,
+   "configFile": "lakefile.lean"}],
+ "name": "Example",
+ "lakeDir": ".lake"}
diff --git a/examples/Example/lakefile.lean b/examples/Example/lakefile.lean
new file mode 100644
index 0000000..52623f0
--- /dev/null
+++ b/examples/Example/lakefile.lean
@@ -0,0 +1,10 @@
+import Lake
+open Lake DSL
+
+require aesop from git
+  "https://github.com/leanprover-community/aesop.git" @ "v4.8.0-rc1"
+
+package Example
+
+@[default_target]
+lean_lib Example
diff --git a/examples/Example/lean-toolchain b/examples/Example/lean-toolchain
new file mode 100644
index 0000000..d8a6d7e
--- /dev/null
+++ b/examples/Example/lean-toolchain
@@ -0,0 +1 @@
+leanprover/lean4:v4.8.0-rc1
diff --git a/examples/README.md b/examples/README.md
new file mode 100644
index 0000000..8ccd495
--- /dev/null
+++ b/examples/README.md
@@ -0,0 +1,15 @@
+# Usage Example
+
+This example showcases how to bind library dependencies and execute the `Aesop`
+tactic in Lean. First build the example project:
+``` sh
+pushd Example
+lake build
+popd
+```
+This would generate compiled `.olean` files. Then run the example from the
+project root:
+``` sh
+poetry run examples/aesop.py
+```
+
diff --git a/examples/aesop.py b/examples/aesop.py
new file mode 100755
index 0000000..3f074b8
--- /dev/null
+++ b/examples/aesop.py
@@ -0,0 +1,19 @@
+#!/usr/bin/env python3
+
+import subprocess
+from pathlib import Path
+from pantograph.server import Server
+
+def get_project_and_lean_path():
+    cwd = Path(__file__).parent.resolve() / 'Example'
+    p = subprocess.check_output(['lake', 'env', 'printenv', 'LEAN_PATH'], cwd=cwd)
+    return cwd, p
+
+if __name__ == '__main__':
+    project_path, lean_path = get_project_and_lean_path()
+    print(f"$PWD: {project_path}")
+    print(f"$LEAN_PATH: {lean_path}")
+    server = Server(imports=['Example'], project_path=project_path, lean_path=lean_path)
+    state0 = server.goal_start("forall (p q: Prop), Or p q -> Or q p")
+    state1 = server.goal_tactic(state0, goal_id=0, tactic="aesop")
+    assert state1.is_solved
diff --git a/examples/compile-tactics.py b/examples/compile-tactics.py
new file mode 100755
index 0000000..d4845da
--- /dev/null
+++ b/examples/compile-tactics.py
@@ -0,0 +1,19 @@
+#!/usr/bin/env python3
+
+import subprocess
+from pathlib import Path
+from pantograph.server import Server
+
+def get_project_and_lean_path():
+    cwd = Path(__file__).parent.resolve() / 'Example'
+    p = subprocess.check_output(['lake', 'env', 'printenv', 'LEAN_PATH'], cwd=cwd)
+    return cwd, p
+
+if __name__ == '__main__':
+    project_path, lean_path = get_project_and_lean_path()
+    print(f"$PWD: {project_path}")
+    print(f"$LEAN_PATH: {lean_path}")
+    server = Server(imports=['Example'], project_path=project_path, lean_path=lean_path)
+    data = server.compile_tactics("Example")
+    for (before, tactic, after) in data:
+        print(f"{before}\n{tactic}\n{after}\n\n")
diff --git a/pantograph/expr.py b/pantograph/expr.py
index 71ef971..c9789a3 100644
--- a/pantograph/expr.py
+++ b/pantograph/expr.py
@@ -2,7 +2,7 @@
 Data structures for expressions and goals
 """
 from dataclasses import dataclass
-from typing import Optional, Union, List
+from typing import Optional, Union
 
 Expr = str
 
@@ -16,7 +16,7 @@ class Variable:
     name: Optional[str] = None
 
     @staticmethod
-    def parse(payload: dict) -> "Variable":  # Replace 'Self' with 'Variable'
+    def parse(payload: dict):
         name = payload.get("userName")
         t = parse_expr(payload["type"])
         v = payload.get("value")
@@ -39,11 +39,11 @@ class Goal:
     is_conversion: bool = False
 
     @staticmethod
-    def sentence(target: Expr) -> "Goal":  # Replace 'Self' with 'Goal'
+    def sentence(target: Expr):
         return Goal(variables=[], target=target)
 
     @staticmethod
-    def parse(payload: dict) -> "Goal":  # Replace 'Self' with 'Goal'
+    def parse(payload: dict) -> Self:
         name = payload.get("userName")
         variables = [Variable.parse(v) for v in payload["vars"]]
         target = parse_expr(payload["target"])
@@ -59,6 +59,16 @@ class GoalState:
     state_id: int
     goals: List[Goal]
 
+    _sentinel: list[int]
+
+    def __del__(self):
+        self._sentinel.append(self.state_id)
+
+    _sentinel: list[int]
+
+    def __del__(self):
+        self._sentinel.append(self.state_id)
+
     @property
     def is_solved(self) -> bool:
         """
@@ -67,10 +77,10 @@ class GoalState:
         return not self.goals
 
     @staticmethod
-    def parse(payload: dict) -> "GoalState":  # Replace 'Self' with 'GoalState'
-        state_id = payload.get("nextStateId", 0)  # Handle missing keys
-        goals = [Goal.parse(g) for g in payload.get("goals", [])]
-        return GoalState(state_id, goals)
+    def parse(payload: dict, _sentinel: list[int]):
+        state_id = payload["nextStateId"]
+        goals = [Goal.parse(g) for g in payload["goals"]]
+        return GoalState(state_id, goals, _sentinel)
 
 @dataclass(frozen=True)
 class TacticHave:
diff --git a/pantograph/gen_tactic.py b/pantograph/gen_tactic.py
new file mode 100644
index 0000000..68c7ccb
--- /dev/null
+++ b/pantograph/gen_tactic.py
@@ -0,0 +1,140 @@
+from pantograph.server import Server
+from pantograph.expr import Variable, Goal, TacticCalc
+import  unittest
+import sglang as sgl
+
+
+
+
+@sgl.function
+def multi_turn_question(s, question_1, question_2):
+    s += sgl.system("You are a helpful assistant.")
+    s += sgl.user(question_1)
+    s += sgl.assistant(sgl.gen("answer_1", max_tokens=256))
+    s += sgl.user(question_2)
+    s += sgl.assistant(sgl.gen("answer_2", max_tokens=256))
+
+
+@sgl.function
+def select_tactic(s, state):
+    s += sgl.system("You are an expert in Lean. Choose the next one tactic to run given the current proof state and goals.")
+    s += sgl.user("The current proof state: GoalState(state_id=0, goals=[Goal(variables=[], target='∀ (a b: Nat), (b = 2) -> 1 + a + 1 = a + b', name=None, is_conversion=False)])")
+    s += sgl.assistant("```intros a b h```")
+    s += sgl.user("The current proof state: GoalState(state_id=1, goals=[Goal(variables=[Variable(t='Nat', v=None, name='a'), Variable(t='Nat', v=None, name='b'), Variable(t='b = 2', v=None, name='h')], target='1 + a + 1 = a + b', name=None, is_conversion=False)])")
+    s += sgl.assistant('TacticCalc("1 + a + 1 = a + 1 + 1")')
+    s += sgl.user("The current proof state: " + str(state))
+    with s.copy() as tmp:
+        tmp += sgl.assistant(sgl.gen("tactic", max_tokens=64))
+        print("==tmp===")
+        print(tmp["tactic"])
+        tactic = extract_code_from_llm_output(tmp["tactic"])
+    s += sgl.assistant("```"+tactic+"```")
+    return tactic
+
+
+    
+def extract_code_from_llm_output(reply):
+    i = reply.find("```lean")
+    if i != -1:
+        reply = reply[i + 7:]
+        i = reply.find("```")
+        reply = reply[:i]
+        return reply
+    i = reply.find("```")
+    if i != -1:
+        reply = reply[i + 3:]
+        i = reply.find("```")
+        reply = reply[:i]
+        return reply
+    return reply
+
+class TestServerSGL(unittest.TestCase):
+
+    def test_conv_calc_sgl(self):
+        sgl.set_default_backend(sgl.OpenAI("gpt-4"))
+
+        server = Server()
+        state0 = server.goal_start("∀ (a b: Nat), (b = 2) -> 1 + a + 1 = a + b")
+        print("==========state0============")
+        print(state0)
+        variables = [
+            Variable(name="a", t="Nat"),
+            Variable(name="b", t="Nat"),
+            Variable(name="h", t="b = 2"),
+        ]
+
+        state1 = server.goal_tactic(state0, goal_id=0, tactic="intro a b h")
+        print("==========state1============")
+        print(state1)
+        state2 = server.goal_tactic(state1, goal_id=0, tactic=TacticCalc("1 + a + 1 = a + 1 + 1"))
+        print("==========state2============")
+        print(state2)
+        self.assertEqual(state2.goals, [
+            Goal(
+                variables,
+                target="1 + a + 1 = a + 1 + 1",
+                name='calc',
+            ),
+            Goal(
+                variables,
+                target="a + 1 + 1 = a + b",
+            ),
+        ])
+        state = select_tactic.run(str(state2))
+        tactic = state.ret_value
+        for m in state.messages():
+            print(m["role"], ":", m["content"])
+
+        print("\n-- tactic --\n", tactic)
+        
+        state3 = server.goal_tactic(state2, goal_id=1, tactic=tactic)
+        print("==========state3============")
+        print(state3)
+        # state4 = server.goal_tactic(state3, goal_id=0, tactic="rw [Nat.add_assoc]")
+        # print("==========state4============")
+        # print(state4)        
+        # self.assertTrue(state4.is_solved)
+
+
+        # print("==========state2============")
+        # print(state2)
+        # state_c1 = server.goal_conv_begin(state2, goal_id=0)
+        # print("==========state c1============")
+        # print(state_c1)
+        # state_c2 = server.goal_tactic(state_c1, goal_id=0, tactic="rhs")
+        # print("==========state c2============")
+        # print(state_c2)
+        # state_c3 = server.goal_tactic(state_c2, goal_id=0, tactic="rw [Nat.add_comm]")
+        # print("==========state c3============")
+        # print(state_c3)
+        # state_c4 = server.goal_conv_end(state_c3)
+        # print("==========state c4============")
+        # print(state_c4)
+
+        # state_c5 = server.goal_tactic(state_c4, goal_id=0, tactic="rfl")
+        # print("==========state c5============")
+        # print(state_c5)
+        # self.assertTrue(state_c5.is_solved)
+
+        # print()
+
+
+    def test_sglang_openai(self):
+        sgl.set_default_backend(sgl.OpenAI("gpt-4"))
+
+        print('\n----- Test sglang ---')
+        state = multi_turn_question.run(
+            question_1="What is the capital of the United States?",
+            question_2="List two local attractions.",
+        )
+
+        for m in state.messages():
+            print(m["role"], ":", m["content"])
+
+        print("\n-- answer_1 --\n", state["answer_1"])
+
+   
+if __name__ == '__main__':
+
+    unittest.main()
+
diff --git a/pantograph/server.py b/pantograph/server.py
index 52a7661..af8c7e3 100644
--- a/pantograph/server.py
+++ b/pantograph/server.py
@@ -2,8 +2,8 @@
 Class which manages a Pantograph instance. All calls to the kernel uses this
 interface.
 """
-import json, pexpect, pathlib, unittest
-from pantograph.expr import Variable, Goal, GoalState, \
+import json, pexpect, pathlib, unittest, os
+from pantograph.expr import parse_expr, Expr, Variable, Goal, GoalState, \
     Tactic, TacticHave, TacticCalc
 
 def _get_proc_cwd():
@@ -18,6 +18,8 @@ class Server:
 
     def __init__(self,
                  imports=["Init"],
+                 project_path=None,
+                 lean_path=None,
                  options=[],
                  timeout=20,
                  maxread=1000000):
@@ -27,8 +29,9 @@ class Server:
         """
         self.timeout = timeout
         self.imports = imports
+        self.project_path = project_path if project_path else _get_proc_cwd()
+        self.lean_path = lean_path
         self.maxread = maxread
-        self.proc_cwd = _get_proc_cwd()
         self.proc_path = _get_proc_path()
 
         self.options = options
@@ -36,35 +39,63 @@ class Server:
         self.proc = None
         self.restart()
 
+        # List of goal states that should be garbage collected
+        self.to_remove_goal_states = []
+
     def restart(self):
         if self.proc is not None:
             self.proc.close()
+        env = os.environ
+        if self.lean_path:
+            env = env | {'LEAN_PATH': self.lean_path}
+
         self.proc = pexpect.spawn(
             f"{self.proc_path} {self.args}",
             encoding="utf-8",
             maxread=self.maxread,
-            cwd=self.proc_cwd,
+            cwd=self.project_path,
+            env=env,
         )
         self.proc.setecho(False)
 
     def run(self, cmd, payload):
+        """
+        Runs a raw JSON command. Preferably use one of the commands below.
+        """
         s = json.dumps(payload)
         self.proc.sendline(f"{cmd} {s}")
         try:
             self.proc.expect("{.*}\r\n", timeout=self.timeout)
             output = self.proc.match.group()
             return json.loads(output)
-        except pexpect.exceptions.TIMEOUT:
-            raise pexpect.exceptions.TIMEOUT
+        except pexpect.exceptions.TIMEOUT as exc:
+            raise exc
 
-    def reset(self):
-        return self.run("reset", {})
+    def gc(self):
+        """
+        Garbage collect deleted goal states.
+
+        Must be called periodically.
+        """
+        if self.to_remove_goal_states:
+            self.run('goal.delete', {'stateIds': self.to_remove_goal_states})
+            self.to_remove_goal_states.clear()
+
+    def expr_type(self, expr: str) -> Expr:
+        """
+        Evaluate the type of a given expression. This gives an error if the
+        input `expr` is ill-formed.
+        """
+        result = self.run('expr.echo', {"expr": expr})
+        if "error" in result:
+            raise ServerError(result["desc"])
+        return parse_expr(result["type"])
 
     def goal_start(self, expr: str) -> GoalState:
         result = self.run('goal.start', {"expr": str(expr)})
         if "error" in result:
             raise ServerError(result["desc"])
-        return GoalState(state_id = result["stateId"], goals = [Goal.sentence(expr)])
+        return GoalState(state_id=result["stateId"], goals=[Goal.sentence(expr)], _sentinel=self.to_remove_goal_states)
 
     def goal_tactic(self, state: GoalState, goal_id: int, tactic: Tactic) -> GoalState:
         args = {"stateId": state.state_id, "goalId": goal_id}
@@ -83,7 +114,7 @@ class Server:
             raise ServerError(result["tacticErrors"])
         if "parseError" in result:
             raise ServerError(result["parseError"])
-        return GoalState.parse(result)
+        return GoalState.parse(result, self.to_remove_goal_states)
 
     def goal_conv_begin(self, state: GoalState, goal_id: int) -> GoalState:
         result = self.run('goal.tactic', {"stateId": state.state_id, "goalId": goal_id, "conv": True})
@@ -93,7 +124,7 @@ class Server:
             raise ServerError(result["tacticErrors"])
         if "parseError" in result:
             raise ServerError(result["parseError"])
-        return GoalState.parse(result)
+        return GoalState.parse(result, self.to_remove_goal_states)
 
     def goal_conv_end(self, state: GoalState) -> GoalState:
         result = self.run('goal.tactic', {"stateId": state.state_id, "goalId": 0, "conv": False})
@@ -103,7 +134,15 @@ class Server:
             raise ServerError(result["tacticErrors"])
         if "parseError" in result:
             raise ServerError(result["parseError"])
-        return GoalState.parse(result)
+        return GoalState.parse(result, self.to_remove_goal_states)
+
+
+    def compile_tactics(self, module: str) -> list[tuple[str, str, str]]:
+        result = self.run('compile.tactics', {'module': module})
+        if "error" in result:
+            raise ServerError(result["desc"])
+        return [(i['goalBefore'], i['tactic'], i['goalAfter']) for i in result['invocations']]
+
 
 
 def get_version():
@@ -117,11 +156,17 @@ def get_version():
 class TestServer(unittest.TestCase):
 
     def test_version(self):
-        self.assertEqual(get_version(), "0.2.14")
+        self.assertEqual(get_version(), "0.2.15")
+
+    def test_expr_type(self):
+        server = Server()
+        t = server.expr_type("forall (n m: Nat), n + m = m + n")
+        self.assertEqual(t, "Prop")
 
     def test_goal_start(self):
         server = Server()
         state0 = server.goal_start("forall (p q: Prop), Or p q -> Or q p")
+        self.assertEqual(len(server.to_remove_goal_states), 0)
         self.assertEqual(state0.state_id, 0)
         state1 = server.goal_tactic(state0, goal_id=0, tactic="intro a")
         self.assertEqual(state1.state_id, 1)
@@ -132,6 +177,17 @@ class TestServer(unittest.TestCase):
         )])
         self.assertEqual(str(state1.goals[0]),"a : Prop\n⊢ ∀ (q : Prop), a ∨ q → q ∨ a")
 
+        del state0
+        self.assertEqual(len(server.to_remove_goal_states), 1)
+        server.gc()
+        self.assertEqual(len(server.to_remove_goal_states), 0)
+
+        state0b = server.goal_start("forall (p: Prop), p -> p")
+        del state0b
+        self.assertEqual(len(server.to_remove_goal_states), 1)
+        server.gc()
+        self.assertEqual(len(server.to_remove_goal_states), 0)
+
     def test_conv_calc(self):
         server = Server()
         state0 = server.goal_start("∀ (a b: Nat), (b = 2) -> 1 + a + 1 = a + b")
diff --git a/poetry.lock b/poetry.lock
index fcd2d0c..de51915 100644
--- a/poetry.lock
+++ b/poetry.lock
@@ -2636,4 +2636,4 @@ torch = "2.2.1"
 [metadata]
 lock-version = "2.0"
 python-versions = "^3.11"
-content-hash = "3b1551795c385fa89fc314894360b50d284ec86c66f5fdd7347442511d714dab"
+content-hash = "aaae6d4832f97d9ad1b145776be94a2c44a7e77068bb1b3a05241a81dde23909"
diff --git a/pyproject.toml b/pyproject.toml
index b3c7d65..70c006a 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -7,7 +7,7 @@ license = "GPL-3"
 readme = "README.md"
 
 [tool.poetry.dependencies]
-python = "^3.11"
+python = "^3.10"
 pexpect = "^4.9.0"
 torch = "2.2.1"
 #vllm = "0.4.1"
diff --git a/src b/src
index f20ee8d..b9b16ba 160000
--- a/src
+++ b/src
@@ -1 +1 @@
-Subproject commit f20ee8dc87ae3fa9c0f13040e1b4f097f1a40503
+Subproject commit b9b16ba0e9d99279837527bcb40176277d11e725