Skip to content

Examples

This page provides practical examples of using LeanInteract in different scenarios. You can find a few full example scripts in the examples directory of the repository.

Basic Theorem Proving

This example demonstrates how to define a simple theorem with a partial proof in Lean using LeanInteract:

from lean_interact import LeanREPLConfig, LeanServer, Command

# Initialize configuration and server
config = LeanREPLConfig()
server = LeanServer(config)

# Define a simple theorem
print(server.run(Command(cmd="""
theorem add_comm (a b : Nat) : a + b = b + a := by
  induction a with
  | zero => simp
  | succ a ih => sorry
""")))

CommandResponse(messages=[Message(start_pos=Pos(line=2, column=8), severity='warning', data="declaration uses 'sorry'", end_pos=Pos(line=2, column=16))], env=0, sorries=[Sorry(start_pos=Pos(line=5, column=17), end_pos=Pos(line=5, column=22), proof_state=0, goal='case succ\nb a : Nat\nih : a + b = b + a\n⊢ a + 1 + b = b + (a + 1)')])

Working with Mathlib

This example shows how to use Mathlib to work with more advanced mathematical concepts:

from lean_interact import LeanREPLConfig, LeanServer, Command, TempRequireProject

# Create configuration with Mathlib
config = LeanREPLConfig(project=TempRequireProject(lean_version="v4.19.0", require="mathlib"))
server = LeanServer(config)

# Define a theorem using Mathlib's real numbers
print(server.run(Command(cmd="""
import Mathlib

theorem irrational_plus_rational 
  (x : ℝ) (y : ℚ) : Irrational x → Irrational (x + y) := by
  intro h
  simp
  assumption
""")))

CommandResponse(messages=[Message(start_pos=Pos(line=4, column=0), severity='info', data='Goals accomplished!', end_pos=Pos(line=8, column=12))], env=0)

Using Custom REPL Versions

This example demonstrates how to use a specific REPL version from a custom repository:

from lean_interact import LeanREPLConfig, LeanServer, Command

# Use a specific REPL version from the official Lean repository
config = LeanREPLConfig(
    repl_rev="v4.21.0-rc3", 
    repl_git="https://github.com/leanprover-community/repl"
)
server = LeanServer(config)

# Check the Lean version
response = server.run(Command(cmd="#eval Lean.versionString"))
print(response.messages[0].data)  # Output: "4.21.0-rc3"

# If you encounter interface compatibility issues with custom REPLs,
# you can use run_dict to communicate directly with the REPL:
result = server.run_dict({"cmd": "#eval Lean.versionString"})
print(result)  # Example raw output from the REPL

"4.21.0-rc3"
{'messages': [{'severity': 'info', 'pos': {'line': 1, 'column': 0}, 'endPos': {'line': 1, 'column': 5}, 'data': '"4.21.0-rc3"'}], 'env': 1}

Real-World Examples

For more comprehensive examples, check out the following scripts in the examples directory:

  1. parallelization.py
    Shows how to parallelize calls to LeanInteract for faster processing.

  2. proof_generation_and_autoformalization.py
    Shows how to use models like DeepSeek-Prover-V1.5 and Goedel-Prover on MiniF2F and ProofNet# benchmarks.

  3. beq_plus.py
    Demonstrates how to run the autoformalization BEq+ metric on the ProofNetVerif benchmark.

  4. type_check.py
    Shows how to optimize type checking using environment states.