certora-cli-beta-mirror 8.2.0__py3-none-any.whl → 8.2.2__py3-none-any.whl

certora_cli/CertoraProver/certoraBuild.py

@@ -203,11 +203,18 @@ class PresetImmutableReference(ImmutableReference):
 # this function is Solidity specific.
 # todo: create certoraBuildUtilsSol.py file, where such solidity specific functions will be.
 def generate_finder_body(f: Func, internal_id: int, sym: int, compiler_collector: CompilerCollectorSol,
-                         compressed: bool = False) -> Optional[Tuple[List[int], str]]:
+                         compressed: bool = False,
+                         should_generate_inlining: bool = True) -> Optional[Tuple[List[int], str]]:
     if compressed:
         return generate_compressed_finder(
             f, internal_id, sym, compiler_collector
         )
+    elif not should_generate_inlining:
+        # We should not generate inlining as reported in CERT-9399, only when compressed=False.
+        # used_symbols is being generated the exact same way as in generate_full_finder
+        used_symbols = [i for i in range(len(f.fullArgs))]
+        return used_symbols, ''
+
     else:
         return generate_full_finder(
             f, internal_id, sym, compiler_collector
@@ -403,8 +410,9 @@ def get_modifier_param_type_name(ind: int, def_node: Dict[str, Any], f: Func) ->
 
 def generate_modifier_finder(f: Func, internal_id: int, sym: int,
                              compiler_collector: CompilerCollectorSol, def_node: Dict[str, Any],
-                             compress: bool) -> Optional[Tuple[str, str]]:
-    compressed = generate_finder_body(f, internal_id, sym, compiler_collector, compressed=compress)
+                             compress: bool, should_generate_inlining: bool) -> Optional[Tuple[str, str]]:
+    compressed = generate_finder_body(f, internal_id, sym, compiler_collector, compressed=compress,
+                                      should_generate_inlining=should_generate_inlining)
     if compressed is None:
         return None
     modifier_name = f"logInternal{internal_id}"
@@ -427,9 +435,10 @@ def generate_modifier_finder(f: Func, internal_id: int, sym: int,
         return f'{modifier_name}({",".join(arg_strings)})', modifier_body
 
 
-def generate_inline_finder(f: Func, internal_id: int, sym: int,
-                           compiler_collector: CompilerCollectorSol, should_compress: bool) -> Optional[str]:
-    finder = generate_finder_body(f, internal_id, sym, compiler_collector, compressed=should_compress)
+def generate_inline_finder(f: Func, internal_id: int, sym: int, compiler_collector: CompilerCollectorSol,
+                           should_compress: bool, should_generate_inlining: bool) -> Optional[str]:
+    finder = generate_finder_body(f, internal_id, sym, compiler_collector, compressed=should_compress,
+                                  should_generate_inlining=should_generate_inlining)
     if finder is None:
         return None
     return finder[1]
@@ -2448,6 +2457,19 @@ class CertoraBuildGenerator:
                     # (deprecate this option later)
                     should_compress = self.context.function_finder_mode == Vf.FunctionFinderMode.DEFAULT.name
 
+                if self.context.solc_optimize and self.get_solc_via_ir_value(Path(contract_file)) and \
+                   any(param_name == "" for param_name in f.paramNames):
+                    # As reported in CERT-9399, we should not generate function finder for this edge case
+
+                    instrumentation_logger.warning(f"Failed to generate auto finder for {f.name} @ {f.where()}.")
+                    instrumentation_logger.warning(
+                        "Cannot apply summaries for internal functions with unnamed argument when compiling "
+                        "using solc_optimize and solc_via_ir"
+                    )
+                    should_generate_inlining = False
+                else:
+                    should_generate_inlining = True
+
                 if len(mods) > 0:
                     # we need to add the instrumentation in a modifer because solidity modifiers will (potentially)
                     # appear before any instrumentation we add to the literal source body, which will tank the detection
@@ -2459,8 +2481,10 @@ class CertoraBuildGenerator:
                     # where in the source such modifiers will go. In order to insert a modifier, we have to have at
                     # least one modifier already present, and then insert before the first modifier's location in the
                     # source code
-                    mod_inst = generate_modifier_finder(f, internal_id, function_symbol, sdc.compiler_collector,
-                                                        def_node, compress=should_compress)
+                    mod_inst = generate_modifier_finder(
+                        f, internal_id, function_symbol, sdc.compiler_collector, def_node,
+                        compress=should_compress, should_generate_inlining=should_generate_inlining
+                    )
                     if mod_inst is None:
                         instrumentation_logger.debug(f"Modifier generation for {f.name} @ {f.where()} failed")
                         return None
@@ -2494,8 +2518,8 @@ class CertoraBuildGenerator:
                     per_file_inst[first_mod_offset] = Instrumentation(expected=bytes(modifier_name[0:1], "utf-8"),
                                                                       to_ins=modifier_invocation, mut=InsertBefore())
                 else:
-                    finder_res = generate_inline_finder(f, internal_id, function_symbol,
-                                                        sdc.compiler_collector, should_compress)
+                    finder_res = generate_inline_finder(f, internal_id, function_symbol, sdc.compiler_collector,
+                                                        should_compress, should_generate_inlining)
                     if finder_res is None:
                         instrumentation_logger.debug(f"Generating auto finder for {f.name} @ {f.where()}"
                                                      f" failed, giving up generation")
@@ -2676,7 +2700,7 @@ class CertoraBuildGenerator:
                     Ctx.run_local_spec_check(False, context, ["-listCalls",  tmp_file.name], print_errors=False)
                     spec_calls = tmp_file.read().split("\n")
                 except Exception as e:
-                    instrumentation_logger.warning(f"Failed to get calls from spec\n{e}")
+                    instrumentation_logger.debug(f"Failed to get calls from spec\n{e}")
 
         self.context.remappings = []
         for i, build_arg_contract_file in enumerate(sorted(self.input_config.sorted_files)):
@@ -3938,7 +3962,7 @@ def build_from_cache_or_scratch(context: CertoraContext,
                 if output:
                     internal_calls = output.split("\n")
             except Exception as e:
-                instrumentation_logger.warning(f"Failed to get calls from spec\n{e}")
+                instrumentation_logger.debug(f"Failed to get calls from spec\n{e}")
 
             if internal_calls:
                 build_logger.info("Found new internal calls in the spec file, need to recompile anyway")
@@ -3989,9 +4013,6 @@ def build(context: CertoraContext, ignore_spec_syntax_check: bool = False) -> No
         if Cv.mode_has_spec_file(context) and not context.build_only and not ignore_spec_syntax_check:
             if Ctx.should_run_local_speck_check(context):
                 Ctx.run_local_spec_check(False, context)
-            else:
-                build_logger.warning(
-                    "Local checks of CVL specification files disabled. It is recommended to enable the checks.")
 
         should_save_cache, cached_files = build_from_cache_or_scratch(context,
                                                                       certora_build_generator,

certora_cli/CertoraProver/certoraContext.py

@@ -523,7 +523,15 @@ def __rename_key(context: CertoraContext, old_key: str, new_key: str) -> None:
 
 
 def should_run_local_speck_check(context: CertoraContext) -> bool:
-    return not (context.disable_local_typechecking or Util.is_ci_or_git_action())
+    output = context.compilation_steps_only or not (context.disable_local_typechecking or Util.is_ci_or_git_action())
+    if not output:
+        if context.disable_local_typechecking:
+            context_logger.warning(
+                "Local checks of CVL specification files disabled. It is recommended to enable the checks."
+            )
+        else:
+            context_logger.info("Local checks of CVL specification files skipped in CI and will run remotely.")
+    return output
 
 
 def run_typechecker(typechecker_name: str, with_typechecking: bool, args: List[str], print_errors: bool) -> None:
@@ -550,18 +558,13 @@ def run_typechecker(typechecker_name: str, with_typechecking: bool, args: List[s
     context_logger.debug(f"typechecking cmd: {' '.join(cmd_str_list)}")
 
     exit_code = Util.run_jar_cmd(cmd_str_list, False,
-                                 custom_error_message="Failed to run Certora Prover locally. Please check the errors "
-                                                      "below for problems in the specifications (.spec files) or the "
-                                                      "prover_args defined in the .conf file.",
+                                 custom_error_message="Failed to run Certora Prover typechecker locally.\n"
+                                                      "Please check the errors below for problems in the specifications"
+                                                      " (.spec files) or the prover_args defined in the .conf file.",
                                  logger_topic="type_check",
                                  print_err=print_errors)
     if exit_code != 0:
-        raise Util.CertoraUserInputError(
-            "CVL syntax or type check failed.\n Please fix the issue. "
-            "Using --disable_local_typechecking to skip this check is strongly discouraged, "
-            "as simple syntax errors will only be detected during the cloud run."
-            "with --disable_local_typechecking is not recommended is not recommended as "
-            "simple syntax failures will be visible only on the cloud run.")
+        raise Util.CertoraUserInputError("CVL syntax or type check failed, please fix the issue.")
 
 
 def run_local_spec_check(with_typechecking: bool, context: CertoraContext, extra_args: List[str] = list(), print_errors: bool = True) -> None:

certora_cli/Mutate/mutateValidate.py

@@ -69,10 +69,18 @@ class MutateValidator:
                 except Exception as e:
                     raise Util.CertoraUserInputError(f"Invalid file_to_mutate in manual mutant: {mutant[Constants.FILE_TO_MUTATE]}", e)
 
-                try:
-                    Vf.validate_dir(mutant[Constants.MUTANTS_LOCATION])
-                except Exception as e:
-                    raise Util.CertoraUserInputError(f"Invalid mutants location {mutant[Constants.MUTANTS_LOCATION]}", e)
+                mutants_location = mutant[Constants.MUTANTS_LOCATION]
+                if Path(mutants_location).is_dir():
+                    try:
+                        Vf.validate_dir(mutants_location)
+                    except Exception as e:
+                        raise Util.CertoraUserInputError(f"Invalid directory for mutants location {mutants_location}",
+                                                         e)
+                else:
+                    try:
+                        Vf.validate_readable_file(mutants_location, Util.SOL_EXT)
+                    except Exception as e:
+                        raise Util.CertoraUserInputError(f"Invalid file for mutants location {mutants_location}", e)
 
     def mutation_attribute_in_prover(self) -> None:
         gambit_attrs = ['filename', 'contract', 'functions', 'seed', 'num_mutants']

certora_cli/certoraRun.py

@@ -132,10 +132,6 @@ def run_certora(args: List[str], app: Type[App.CertoraApp] = App.EvmApp,
                 Ctx.run_local_spec_check(True, context)
                 typechecking_end = time.perf_counter()
                 timings['typecheckingTime'] = round(typechecking_end - typechecking_start, 4)
-            else:
-                run_logger.warning(
-                    "Local checks of CVL specification files disabled. It is recommended to enable "
-                    "the checks.")
 
         # Remove debug logger and run remote verification
         logging_manager.remove_debug_logger()

{certora_cli_beta_mirror-8.2.0.dist-info → certora_cli_beta_mirror-8.2.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: certora-cli-beta-mirror
-Version: 8.2.0
+Version: 8.2.2
 Summary: Runner for the Certora Prover
 Home-page: https://pypi.org/project/certora-cli-beta-mirror
 Author: Certora
@@ -39,4 +39,4 @@ Dynamic: requires-dist
 Dynamic: requires-python
 Dynamic: summary
 
-Commit 03cec31.                    Build and Run scripts for executing the Certora Prover on Solidity smart contracts.
+Commit 04cd39e.                    Build and Run scripts for executing the Certora Prover on Solidity smart contracts.

{certora_cli_beta_mirror-8.2.0.dist-info → certora_cli_beta_mirror-8.2.2.dist-info}/RECORD

@@ -6,13 +6,13 @@ certora_cli/certoraEVMProver.py,sha256=AH8ZeWXAs7wK6nNKFANCt6seLf9EaaFXtfpjc3xdI
 certora_cli/certoraEqCheck.py,sha256=qfZq7bpU1kbAIezC66W61VfKNZz7Uywg2Ygup62qYeo,1069
 certora_cli/certoraMutate.py,sha256=XhFHyNVP_sk-3XkY6AAV5fVliEFAVRq-JeDGsqE5IQQ,3333
 certora_cli/certoraRanger.py,sha256=thJ4EKo-MFHklCk-7zJADZ-9SO6eCg1AFv88-QssLj0,1289
-certora_cli/certoraRun.py,sha256=11tc3-l_G8cidFNSQfzCmuPbqfF-n_UEIZC4yASLz8g,6561
+certora_cli/certoraRun.py,sha256=3v3o_nMfdVHT-MSqTW0EYqyu2IjKsZvMcXwmfkOQaU8,6371
 certora_cli/certoraSolanaProver.py,sha256=1R1YnGHCofb05GqFgpxRh3ZmHkmwMm1hPM7rfeiEu7o,3250
 certora_cli/certoraSorobanProver.py,sha256=SYJKz5Sw-N0bJrSa1njRCE53R9_PMz7IWLhfamOjisk,2840
 certora_cli/rustMutator.py,sha256=6AvOGU8Ijz89zW_ZJCWlfXkeobJsk7EsqZhK7Eqwn-Y,14544
 certora_cli/CertoraProver/__init__.py,sha256=QHNr-PJQAoyuPgTkO7gg23GRchiWSXglWNG7yLSQZvs,849
 certora_cli/CertoraProver/certoraApp.py,sha256=RKJ2Krb_CzbRUvczbdE6FhUDrFcvrR8j0JS8MNWXX7s,1469
-certora_cli/CertoraProver/certoraBuild.py,sha256=CEavCUO0mBZ74dxR23RRInyDskHAddoX8Ox8-gaqnJk,222188
+certora_cli/CertoraProver/certoraBuild.py,sha256=Yn-J7QKxQvtp1uvs03mOfj2w7N_9XixnDmCgoxM-dWM,223495
 certora_cli/CertoraProver/certoraBuildCacheManager.py,sha256=YnZmBZ_gCIbLwExgK5oxFlVDQGe4_YuGIpDLMy589E0,13318
 certora_cli/CertoraProver/certoraBuildDataClasses.py,sha256=hO0w3YK9V9gZsTbh4gxxlnEAaOiubUwfzNEw6uL1HaE,14841
 certora_cli/CertoraProver/certoraBuildRust.py,sha256=ZPbNp4ttRmzcKhFsgHSiHDRExNPaLOzgxTRqu23o1D0,6061
@@ -22,7 +22,7 @@ certora_cli/CertoraProver/certoraCollectConfigurationLayout.py,sha256=Rln6LsqMp-
 certora_cli/CertoraProver/certoraCollectRunMetadata.py,sha256=i31dkYt8kwlX44SHZtV_b8QI1Egi6cbB5-uuh5beYN0,12630
 certora_cli/CertoraProver/certoraCompilerParameters.py,sha256=r35y03IRwWIoz1GCNC7PuW3n8JPz9J1NGwhwUYKdYtI,1452
 certora_cli/CertoraProver/certoraConfigIO.py,sha256=J12jAZ8aKdOQhnMtGUtWmvibjzm_2qlvH6YwI8I4LAc,7483
-certora_cli/CertoraProver/certoraContext.py,sha256=l6aXnfe-fdaDacIVFlhwGqfWb5dne69AnqBrck7-3bw,28758
+certora_cli/CertoraProver/certoraContext.py,sha256=OVNACUNj9fGb9He8sGdod2waC0eWE8ZgXNPWjRC1lFg,28817
 certora_cli/CertoraProver/certoraContextAttributes.py,sha256=O7beTjvk0uBTFyxUaSgdV9Tcg33--G4XA6nnIdX00GU,66778
 certora_cli/CertoraProver/certoraContextClass.py,sha256=d7HDqM72K7YnswR7kEcAHGwkFNrTqRz5-_0m7cl2Mso,900
 certora_cli/CertoraProver/certoraContextValidator.py,sha256=clXwmYHIBUmOJ8ADRUmu9U6_pNe_tXRNawvUPBLlYVY,46490
@@ -36,7 +36,6 @@ certora_cli/CertoraProver/certoraProjectScanner.py,sha256=jT7FeWzcy8o83LrZRwsg_L
 certora_cli/CertoraProver/certoraSourceFinders.py,sha256=qwJtwrQq3NUNYmdmn1UmANN4lmJFIUh4M-St2x1FJ2Y,19038
 certora_cli/CertoraProver/certoraType.py,sha256=inwaLkMVwtJnwkyQhDJs-wRxoyytu2Xa_BJ5MdGlZqY,29737
 certora_cli/CertoraProver/certoraVerifyGenerator.py,sha256=YMuzGj2RNOnADOx8UnV2ys1ptw_-2mermgC9ZLMWceo,11052
-certora_cli/CertoraProver/concordance.py,sha256=CCDN7hT9rnlU-ZKJ7hVxHR2fS8ZXwKVdE59jtMTtY0M,38666
 certora_cli/CertoraProver/erc7201.py,sha256=BME5kBZsDx6lgqLn7EE91I1cEOZtsnZ8BlRVF62eEBE,1660
 certora_cli/CertoraProver/splitRules.py,sha256=dNhy05ShB_-rWYTnJH5m-Xc5A4HGStAvwLRs1BTu1GA,7627
 certora_cli/CertoraProver/storageExtension.py,sha256=nrCrbH8ne-yCYSDFzh3J9A7Q6h96WxhEfLbfxGSUCSc,14363
@@ -60,7 +59,7 @@ certora_cli/Mutate/mutateApp.py,sha256=UY4TWn9f318b2upVAFUL0cqcgzlyTDZ5XeNW3Mjof
 certora_cli/Mutate/mutateAttributes.py,sha256=2onGaPmztwmHg5V_X7BUG4HcQCThhqYzGYKBy695Izc,10587
 certora_cli/Mutate/mutateConstants.py,sha256=LRrz3wMM8WpPYSshkc-PLYqT0nexcWQeBNsehip-LOE,3945
 certora_cli/Mutate/mutateUtil.py,sha256=B7MCIFtZBetjR4MMxU6F5ikYsaot1wTG7XYMjgVXl4k,2287
-certora_cli/Mutate/mutateValidate.py,sha256=vL3YxVkHIGoxolAcvPhK2xitUV0frQ4jTdTN270yeQ4,9298
+certora_cli/Mutate/mutateValidate.py,sha256=6DRYfnE-HCMvMxbb0v1dx2BTSd-x8YeFRVp35KplVPc,9760
 certora_cli/Shared/ExpectedComparator.py,sha256=eyRR-jni4WJoa6j2TK2lnZ89Tyb8U99wT2PNdu4se8w,18457
 certora_cli/Shared/__init__.py,sha256=s0dhvolFtsS4sRNzPVhC_rlw8mm194rCZ0WhOxInY40,1025
 certora_cli/Shared/certoraAttrUtil.py,sha256=Nw8ban5Axp6c6dT-KJfCD9i9tKnGk1DbvRDDNH3--DU,8574
@@ -68,13 +67,13 @@ certora_cli/Shared/certoraLogging.py,sha256=cV2UQMhQ5j8crGXgeq9CEamI-Lk4HgdiA3HC
 certora_cli/Shared/certoraUtils.py,sha256=buUE95PnbpDEbYEZss_jcP7pWCRJhXnfV-eOA6-zvYM,58407
 certora_cli/Shared/certoraValidateFuncs.py,sha256=BPLuVsS3yAcYIuCvkXtDuFQKf2qaT74TIddB0lM84yM,42508
 certora_cli/Shared/proverCommon.py,sha256=uZkl9PDLPj81kKRnBnlPUmvhMZovNP25_74No_7jaQ4,11215
-certora_jars/ASTExtraction.jar,sha256=yGLGQYChi03jYAh5xcEdCDtJNrKJUaUvnKCvqPNEIoI,17614869
-certora_jars/CERTORA-CLI-VERSION-METADATA.json,sha256=7erl5giUb6YUO0NSTkPTRLlt1Ji4iJDxlLrnsCrPlzo,144
-certora_jars/Typechecker.jar,sha256=-EbNEH1eheFLgb1vVSlG5o1g1bAmK2Q6iBKFkKMaND0,17577026
+certora_jars/ASTExtraction.jar,sha256=HAx5yR4VWRf89cTNLqbTvhWBasvlJ6-QVq8oOmsJTcE,17621131
+certora_jars/CERTORA-CLI-VERSION-METADATA.json,sha256=ojcDYBFNRvg4kTdYiKJ3aIrL4qP6pe_5B65U207pd-M,144
+certora_jars/Typechecker.jar,sha256=vt_FUhZTnO5aRHXmi46jh0Cb6XUJIEUIMlqoY7Kdgvo,17583288
 certora_jars/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-certora_cli_beta_mirror-8.2.0.dist-info/LICENSE,sha256=UGKSKIJSetF8m906JLKqNLkUS2CL60XfQdNvxBvpQXo,620
-certora_cli_beta_mirror-8.2.0.dist-info/METADATA,sha256=bU4nWKle__qGpvqqT8bsiiLCehI_qEtlpBFNi-2tcX4,1286
-certora_cli_beta_mirror-8.2.0.dist-info/WHEEL,sha256=jB7zZ3N9hIM9adW7qlTAyycLYW9npaWKLRzaoVcLKcM,91
-certora_cli_beta_mirror-8.2.0.dist-info/entry_points.txt,sha256=ClZiFkCYDdK25_ufxZvnE2Rx_kNk1_4vj7KpgYUKxGM,509
-certora_cli_beta_mirror-8.2.0.dist-info/top_level.txt,sha256=8C77w3JLanY0-NW45vpJsjRssyCqVP-qmPiN9FjWiX4,38
-certora_cli_beta_mirror-8.2.0.dist-info/RECORD,,
+certora_cli_beta_mirror-8.2.2.dist-info/LICENSE,sha256=UGKSKIJSetF8m906JLKqNLkUS2CL60XfQdNvxBvpQXo,620
+certora_cli_beta_mirror-8.2.2.dist-info/METADATA,sha256=KGPhyStTLcrr-_ku2jSW-NkytzN8ymPm4UwMH-0zv5c,1286
+certora_cli_beta_mirror-8.2.2.dist-info/WHEEL,sha256=jB7zZ3N9hIM9adW7qlTAyycLYW9npaWKLRzaoVcLKcM,91
+certora_cli_beta_mirror-8.2.2.dist-info/entry_points.txt,sha256=ClZiFkCYDdK25_ufxZvnE2Rx_kNk1_4vj7KpgYUKxGM,509
+certora_cli_beta_mirror-8.2.2.dist-info/top_level.txt,sha256=8C77w3JLanY0-NW45vpJsjRssyCqVP-qmPiN9FjWiX4,38
+certora_cli_beta_mirror-8.2.2.dist-info/RECORD,,

certora_jars/CERTORA-CLI-VERSION-METADATA.json

@@ -1 +1 @@
-{"name": "certora-cli-beta-mirror", "tag": "8.2.0", "branch": "", "commit": "03cec31", "timestamp": "20250821.16.28.619968", "version": "8.2.0"}
+{"name": "certora-cli-beta-mirror", "tag": "8.2.2", "branch": "", "commit": "04cd39e", "timestamp": "20250904.12.49.982490", "version": "8.2.2"}

certora_cli/CertoraProver/concordance.py

@@ -1,939 +0,0 @@
-#      The Certora Prover
-#      Copyright (C) 2025  Certora Ltd.
-#
-#      This program is free software: you can redistribute it and/or modify
-#      it under the terms of the GNU General Public License as published by
-#      the Free Software Foundation, version 3 of the License.
-#
-#      This program is distributed in the hope that it will be useful,
-#      but WITHOUT ANY WARRANTY; without even the implied warranty of
-#      MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
-#      GNU General Public License for more details.
-#
-#      You should have received a copy of the GNU General Public License
-#      along with this program.  If not, see <https://www.gnu.org/licenses/>.
-
-# ============================================================================
-# IMPORTS
-# ============================================================================
-
-from langchain_anthropic import ChatAnthropic
-from typing import Optional, List, TypedDict, Annotated, Literal, Required, TypeVar, Type, Protocol, Union, Any
-from langchain_core.messages import ToolMessage, AnyMessage, SystemMessage, HumanMessage, BaseMessage
-from langchain_core.tools import tool, InjectedToolCallId, BaseTool
-from langchain_core.language_models.base import LanguageModelInput
-from langchain_core.language_models.chat_models import BaseChatModel
-from langchain_core.runnables import Runnable, RunnableConfig
-from langgraph.graph import StateGraph, START, MessagesState
-from langgraph.graph.state import CompiledStateGraph
-from langgraph._internal._typing import StateLike
-from langgraph.graph.message import add_messages
-from langgraph.types import Command, interrupt
-from langgraph.prebuilt import ToolNode
-from langgraph.checkpoint.memory import MemorySaver
-from pydantic import BaseModel, Field
-import os
-import tempfile
-import json
-import subprocess
-import sys
-import logging
-import argparse
-
-# ============================================================================
-# LOGGING SETUP
-# ============================================================================
-
-logger = logging.getLogger("concordance")
-response_logger = logger.getChild("response")
-tool_logger = logger.getChild("tools")
-
-
-# ============================================================================
-# SHARED UTILITIES
-# ============================================================================
-
-
-class GraphInput(TypedDict):
-    code_input: str
-
-
-class WithToolCallId(BaseModel):
-    tool_call_id: Annotated[str, InjectedToolCallId]
-
-
-def tool_return(
-    tool_call_id: str,
-    content: str
-) -> Command:
-    """
-    Create a LangGraph Command for tool responses that need to continue processing.
-
-    Used by tools that want to return a result and continue the workflow by routing
-    back to the tool_result node for LLM processing.
-
-    Args:
-        tool_call_id: The ID of the tool call being responded to
-        content: The response content from the tool execution
-
-    Returns:
-        Command that updates messages and continues workflow
-    """
-    return Command(
-        update={
-            "messages": [ToolMessage(tool_call_id=tool_call_id, content=content)]
-        }
-    )
-
-
-def tool_output(tool_call_id: str, res: dict) -> Command:
-    """
-    Create a LangGraph Command for final tool outputs that update workflow state.
-
-    Used by completion tools (like harness_output, rewrite_output) to set final
-    results in the workflow state. The workflow's conditional edge will detect
-    these state updates and route to completion.
-
-    Args:
-        tool_call_id: The ID of the tool call being responded to
-        res: Dictionary containing the final workflow results to merge into state
-
-    Returns:
-        Command that updates state with final results and a success message
-    """
-    return Command(update={
-        **res,
-        "messages": [ToolMessage(
-            tool_call_id=tool_call_id,
-            content="Success"
-        )]
-    })
-
-def pretty_print_messages(messages: list[AnyMessage]) -> str:
-    """Format a list of AnyMessage objects for readable debug output."""
-    formatted_lines = []
-    for i, msg in enumerate(messages):
-        msg_type = type(msg).__name__
-
-        # Get message role if available
-        role = getattr(msg, 'role', 'unknown')
-
-        role = getattr(msg, 'type', 'unknown')
-        # Get content preview (handle both string and list content)
-        if hasattr(msg, 'content') and msg.content:
-            if isinstance(msg.content, list):
-                # For list content, show count and first item preview
-                content_preview = \
-                    f"[{len(msg.content)} items: {str(msg.content[0])[:50] if msg.content else 'empty'}...]"
-            else:
-                content_preview = str(msg.content)[:100]
-                if len(str(msg.content)) > 100:
-                    content_preview += "..."
-        else:
-            content_preview = "<empty>"
-
-        # Format tool calls if present
-        tool_info = ""
-        if tool_calls := getattr(msg, 'tool_calls', None):
-            tool_names = [tc.get('name', 'unknown') for tc in tool_calls]
-            tool_info = f" | Tools: {', '.join(tool_names)}"
-
-        formatted_lines.append(f"  [{i}] {msg_type} (role: {role}): {content_preview}{tool_info}")
-
-    return "\n" + "\n".join(formatted_lines) if formatted_lines else " <no messages>"
-
-
-class InitNodeFunction(Protocol):
-    """Protocol defining the signature for LangGraph node functions."""
-    def __call__(self, state: GraphInput) -> dict[str, List[BaseMessage]]:
-        ...
-
-
-class ChatNodeFunction(Protocol):
-    def __call__(self, state: MessagesState) -> dict[str, List[BaseMessage]]:
-        ...
-
-
-def tool_result_generator(llm: Runnable[LanguageModelInput, BaseMessage]) -> ChatNodeFunction:
-    """
-    Create a LangGraph node function that processes tool results by sending
-    the current message history to the LLM for the next response.
-
-    Args:
-        llm: The LLM bound with tools to invoke for generating responses
-
-    Returns:
-        A node function that takes MessagesState and returns updated messages
-    """
-    def tool_result(state: MessagesState) -> dict[str, List[BaseMessage]]:
-        logger.debug("Tool result state messages:%s", pretty_print_messages(state["messages"]))
-        return {"messages": [llm.invoke(state["messages"])]}
-    return tool_result
-
-def initial_node(sys_prompt: str, initial_prompt: str, llm: Runnable[LanguageModelInput, BaseMessage]) -> InitNodeFunction:
-    """
-    Create a LangGraph node function that initializes a workflow with system and human messages,
-    then gets the first LLM response.
-
-    Args:
-        sys_prompt: System message content to set the LLM's role and context
-        initial_prompt: Human message template to start the conversation
-        llm: The LLM bound with tools to invoke for generating the initial response
-
-    Returns:
-        A node function that takes GraphInput and returns initial message history
-    """
-    def to_return(state: GraphInput) -> dict[str, List[BaseMessage]]:
-        initial_messages : List[BaseMessage] = [
-            SystemMessage(
-                sys_prompt
-            ),
-            HumanMessage(
-                content=[initial_prompt, state["code_input"]]
-            )
-        ]
-        initial_messages.append(
-            llm.invoke(initial_messages)
-        )
-        return {"messages": initial_messages}
-    return to_return
-
-
-# TypeVars for generic typing
-StateT = TypeVar('StateT', bound=StateLike)
-OutputT = TypeVar('OutputT', bound=StateLike)
-
-
-def build_workflow(
-    state_class: Type[StateT],
-    tools_list: List[BaseTool],
-    sys_prompt: str,
-    initial_prompt: str,
-    output_key: str,
-    unbound_llm: BaseChatModel,
-    output_schema: Optional[Type[OutputT]] = None,
-) -> StateGraph[StateT, None, GraphInput, OutputT]:
-    """
-    Build a standard workflow with initial node -> tools -> tool_result pattern.
-    Uses fixed GraphInput schema and explicit LLM currying.
-    """
-    # Node name constants
-    INITIAL_NODE = "initial"
-    TOOLS_NODE = "tools"
-    TOOL_RESULT_NODE = "tool_result"
-
-    def should_end(state: StateT) -> Literal["__end__", "tool_result"]:
-        """Check if workflow should end based on output key being defined."""
-        assert isinstance(state, dict)
-        if state.get(output_key, None) is not None:
-            return "__end__"
-        return TOOL_RESULT_NODE
-
-    llm = unbound_llm.bind_tools(tools_list)
-
-    # Create initial node and tool node with curried LLM
-    init_node = initial_node(sys_prompt=sys_prompt, initial_prompt=initial_prompt, llm=llm)
-    tool_node = ToolNode(tools_list)
-    tool_result_node = tool_result_generator(llm)
-
-    # Build the graph with fixed input schema, no context
-    builder = StateGraph(
-        state_class,
-        input_schema=GraphInput,
-        output_schema=output_schema
-    )
-    builder.add_node(INITIAL_NODE, init_node)
-    builder.add_edge(START, INITIAL_NODE)
-    builder.add_node(TOOLS_NODE, tool_node)
-    builder.add_edge(INITIAL_NODE, TOOLS_NODE)
-    builder.add_node(TOOL_RESULT_NODE, tool_result_node)
-    builder.add_edge(TOOL_RESULT_NODE, TOOLS_NODE)
-
-    # Add conditional edge from tools
-    builder.add_conditional_edges(
-        TOOLS_NODE,
-        should_end
-    )
-
-    return builder
-
-
-# ============================================================================
-# SOLIDITY COMPILER TOOL (SHARED)
-# ============================================================================
-
-
-class SolidityCompilerInput(BaseModel):
-    """
-    A Solidity compiler capable of compiling a single, Solidity file into EVM bytecode. The compiler
-    also performs typechecking and will flag any syntax errors. The compiler comes from the official
-    distribution channels for Solidity and understands all the Solidity language and features.
-    """
-    compiler_version: str = \
-        Field(description=
-              "The compiler version string to use for compilation. Compiler versions are taken from the known compiler "
-              "releases (e.g., 0.8.2), but with the leading '0.' dropped (e.g., 8.2)."
-              )
-
-    source: str = Field(description="The Solidity source to be compiled")
-
-
-@tool(args_schema=SolidityCompilerInput)
-def solidity_compiler(source: str, compiler_version: str) -> str:
-    compiler_input = {
-        "language": "Solidity",
-        "sources": {
-            "harness.sol": {
-                "content": source
-            }
-        },
-        "settings": {
-            "outputSelection": {
-                "*": {
-                    "*": []
-                }
-            }
-        }
-    }
-    compile_result = subprocess.run(
-        [f'solc{compiler_version}', "--standard-json"],
-        input=json.dumps(compiler_input),
-        text=True,
-        encoding="utf-8",
-        capture_output=True
-    )
-    res = f"Return code was: {compile_result.returncode}\nStdout:\n{compile_result.stdout}"
-    return res
-
-
-# ============================================================================
-# HARNESSING WORKFLOW
-# ============================================================================
-
-
-harness_system_prompt = """
-You are an expert Solidity developer, with several years of experience writing smart contracts. You also
-have a deep understanding of the EVM and how the Solidity language is ultimately compiled to the EVM bytecode.
-This lets you understand why certain programs written in Solidity language are invalid and rejected by the
-compiler. For example, you know that an `external` function in a `contract` cannot accept a reference type
-marked as `storage`: you know that this is only allowed in a `library` which is always accessed with a delegatecall.
-
-You also understand the subtleties around ABI encoding and decoding, and the translation of high-level types to
-an ABI signature. For example, you know that a struct with two uint fields is represented in an ABI signature as
-`(uint256,uint256)`.
-"""
-
-harnessing_prompt = """
-Create an external 'harness contract' which provides a minimal way to execute the given 'internal' Solidity function
-via an external function wrapper.
-The external function wrapper should simply pass its arguments to the internal function, and return the result back
-to the caller. The internal function being harnessed
-should be included in the contract. You MAY include extra type definitions, but only if absolutely necessary for the
-code to compile; definitions solely for documentation or explanation
-purposes should NOT be included."
-The external harness should be type correct and syntax correct. To ensure this, use the Solidity compiler and
-incorporate its feedback to fix any type/syntax errors.
-"""
-
-
-class HarnessedOutput(TypedDict):
-    harness_definition: Optional[str]
-
-
-class HarnessingState(TypedDict):
-    GraphInput: str
-    harness_definition: Optional[str]
-    messages: Annotated[list[AnyMessage], add_messages]
-
-
-class HarnessOutputSchema(WithToolCallId):
-    """
-    Used to communicate the results of harness generation, which is the minimal contract to exercise an internal
-    function, along with the ABI signature of the method which is the external entry point and the name of the contract.
-    Used only for successfully validated (type correct, syntax correct) harnesses.
-    """
-    source_code: str = \
-        Field(description=
-              "The self-contained Solidity source code which wraps the provided internal function"
-              )
-
-    contract_name: str = \
-        Field(description=
-              "The name of the Solidity contract containing the external method that wraps the internal function"
-              )
-
-    abi_signature: str = \
-        Field(description=
-              "The ABI signature of the external function generated as the internal function wrapper. "
-              "Includes parameter types (but not return types)"
-              )
-
-
-@tool(args_schema=HarnessOutputSchema)
-def harness_output(source_code: str, tool_call_id: Annotated[str, InjectedToolCallId], contract_name: str, abi_signature: str) -> Command:
-    return tool_output(tool_call_id=tool_call_id, res={"harness_definition": source_code})
-
-
-# Harness workflow setup
-HARNESS_TOOLS = [harness_output, solidity_compiler]
-
-# ============================================================================
-# REWRITE WORKFLOW
-# ============================================================================
-
-simplification_system_prompt = """
-You are an expert Solidity developer with several years of experience writing smart contracts and
-optimizing them for gas costs. You know all about low-level assembly and how to use it to directly
-access EVM internal representations to optimize execution. This means you are also familiar with how
-the EVM works on a deep, fundamental level. You are also intimately familiar with how Solidity
-lays out memory, and its allocation pattern. Among other facts, you know that it uses a bump allocator
-with a free pointer whose value resides in memory at slot 0x40. You also know that memory in the range 0x0 - 0x40
-is considered "scratch" and is freely usable, as is all memory after the current value of the free pointer. You
-know that arrays are allocated to include an extra word at the beginning of the allocated block to hold the length
-of the memory, followed by elements of the array. `bytes` and `string` arrays pack their elements tightly (1 byte
-per element), whereas all other arrays use 32 bytes per element.
-
-You also hold a PhD in static analysis, and are an expert in the field of points to analyses and memory
-safety analyses. You help maintain a static analysis which attempts to recover the pointer relationships
-between stack values and memory locations in compiled EVM bytecode. For soundness, this analysis
-must be able to prove that every access to memory is either in the scratch areas OR said access can be
-attributed to a field of some object. Accesses to memory which cannot be proven to satisfy one of these two conditions
-cause the entire analysis to fail. The analysis is partially path sensitive, and can understand that
-`i < array.length` means that `i` is a valid index into `array`. The analysis uses these facts to prove
-accesses are safe AND which object's fields are being accessed by each memory operation.
-"""
-
-rewriting_prompt = """
-<context>
-The following contract "harnesses" a problematic internal function which causes a pointer analysis on EVM bytecode to fail.
-This may be due to non-standard operations on memory that occcurs in memory blocks, or due to imprecision in the
-pointer analysis.
-</context>
-
-<task>
-Rewrite the *internal* function so that it is semantically equivalent but is more amenable to static analysis.
-Common problems include:
-  - Inline assembly with direct memory manipulation
-  - Unchecked array/memory access
-  - Pointer arithmetic that the analyzer cannot track
-  - Non-standard memory layout assumptions
-Your rewrite should satisfy the following constraints:
- - It must be semantically equivalent to the original function.
- - Wherever possible, eschew the use of inline assembly in favor of more straightforward, standard Solidity
- - You may ignore the gas implications of any code you write: code that is accepted by the pointer analysis is
-   preferable to gas efficient code. However, you should consider that the original code may by optimized for gas
-   consumption, which should inform your understanding of its intent
-"Semantic equivalence" means the following:
-- Functions produce the same return value
-- The functions have exactly the same observable effects. These external effects are:
-  - Reverting (including the revert data)
-  - EVM level returns (that is, the return opcode)
-  - External calls
-  - Changes to storage
-  - emitted logs/events
-
-In other words, if the original function reverts, the rewritten function must also revert with
-the same buffer.
-For the purposes of this rewrite, you can ignore the possibility of out-of-gas exceptions.
-Similarly, the rewrite must emit the same log events (if any) and in the same order.
-The rewrite must also make the same external calls, and make the same modifications to storage.
-However, remember that if both functions revert, any other side effects (external calls, storage changes, etc.)
-are mooted.
-</task>
-
-<algorithm>
-   <input>An "original harness" around a "problematic internal function"</input>
-   <output>The rewritten "better function"</output>
-   <steps>
-   1. Analyze the "problematic internal function" in the "original harness" to understand its behavior.
-      Pay close attention to its revert conditions and side effects
-   2. Generate a rewrite of the internal function called the "better function", which uses straight-forward
-      solidity while preserving equivalence to the "problematic internal function"
-      a. Keep track of and remember any extra definitions required for this "better function" rewrite.
-   3. Adapt the "original harness" into a "rewrite harness" by replacing the "problematic internal function" with the
-      "better function" generated in step 2 and changing the name of the "original harness" contract.
-      Incorporate any definitions generated by step 2.a
-   4. Check that the "rewrite harness" is type correct and syntax correct using the solidity compiler
-   5. Check that the "rewrite harness" and "original harness" are semantically equivalent using the equivalence checker.
-   6. Interpret the results of the equivalence checker:
-      a. If the result is 'Equivalent', then go to step 7
-      b. Otherwise, examine the explanation provided by the equivalence checker for why the two functions are not
-         equivalent. Incorporating this feedback, adjust the definition of "better function" within the
-         "rewrite harness", and go to step 5.
-   7. Output the definition of the "better function" along with any of the extra definitions that are necessary.
-   </steps>
-</algorithm>
-
-<guidance>
-   <important>
-     When invoking the equivalence checker, you *may not* change the external entry point of
-     either the "original harness" or the "rewrite harness"
-   </important>
-   <important>
-     You *MAY NOT* change the "original harness" in any way: you must pass it to the equivalence checker without
-     modification.
-   </important>
-   <important>
-     The task is complete only when the equivalence checker says the implementations are 'Equivalent'
-   </important>
-   <soft_requirement>4
-     You should *not* add additional error/interface declarations unless absolutely necessary
-     for your rewrite to compile.
-   </soft_requirement>
-   <soft_requirement>
-     Inline assembly should be absolutely avoided unless you have no other option to preserve semantic equivalence.
-     If you have no choice but to use inline assembly, the inline assembly should hew as closely as possible to
-     standard Solidity memory access patterns; array accesses should be "guarded" by length checks, and so on.
-   </soft_requirement>
-   <reminder>
-     When adapting the "original harness" to check equivalence, you **should** change the name of the harnessing
-     contract.
-   </reminder>
-   <tool_advice>
-      You **should** check that your rewrite harness is type and syntax correct using the solidity compiler.
-   </tool_advice>
-   <tool_advice>
-      You are an automated tool, and should only use the the human_in_the_loop tool as a last resort to get "unstuck".
-      Be sure to iterate on a particular issue a few times before asking the user for help.
-    </tool_advice>
-</guidance>
-"""
-
-
-class EquivalenceCheckerSchema(BaseModel):
-    """
-    A formal verification tool that is able to compare the behavior of two external methods in two different contracts
-    on all possible inputs, and judges whether they have the same side effects.
-    A side effect includes: changes to storage, external calls, logs, and returns/reverts.
-
-    If the equivalence checker thinks the external contracts exhibit different behaviors, it will respond with
-    a concrete example demonstrating the difference in behaviors. Otherwise it will respond with just 'Equivalent'.
-
-    IMPORTANT: The name of the two contracts containing the external methods *must* be different and the external
-    methods *must* have the same ABI signature.
-    """
-
-    contract1: str = \
-        Field(description=
-              "Solidity source code of the first contract to compare for equivalence. This source code must be s"
-              "elf-contained, and must be compilable with a standard solidity compiler. It must be type correct and "
-              "syntactically correct."
-              )
-
-    contract1_name: str = \
-        Field(description=
-              "The name of the contract defined in the `contract1` param. For example, if `contract1` contains the "
-              "source `contract Foo { ... }` this parameter should be `Foo`"
-              )
-
-    contract2: str = \
-        Field(description=
-              "Solidity source code of the second contract to compare for equivalence. The source code must be "
-              "self-contained, and must be compilable with a standard solidity compiler. It must therefore be type "
-              "correct and syntactically correct."
-              )
-
-    contract2_name: str = \
-        Field(description=
-              "The name of the contract defined in the `contract2` param. MUST be different from the value of "
-              "`contract1-name`. For example, if `contract2` contains the source code "
-              "`contract Bar { ... }` this parameter should be `Bar`."
-              )
-
-    abi_signature: str = \
-        Field(description=
-              "The ABI signature (name and parameter types) of the external method to compare between "
-              "contract1 and contract2"
-              )
-
-    compiler_version: str = \
-        Field(description=
-              "The compiler version string to use for compiling contract1 and contract2. Compiler versions are taken "
-              "from the known compiler releases (e.g., 0.8.2), but with the leading '0.' dropped (e.g., 8.2)."
-              )
-
-    loop_bound: int = \
-        Field(description=
-              "When verifying equivalence of looping code, how many times to unroll the loop for bounded verification. "
-              "For performance reasons, this should be set as small as possible while still demonstrating non-trivial "
-              "behavior. While values above 3 are supported, performance gets exponentially worse above these values, "
-              "and they should be avoided if possible."
-              )
-
-
-@tool(args_schema=EquivalenceCheckerSchema)
-def equivalence_check(
-    contract1: str,
-    contract1_name: str,
-    contract2: str,
-    contract2_name: str,
-    abi_signature: str,
-    loop_bound: int,
-    compiler_version: str
-) -> str:
-    print("Running the equivalence checker...")
-
-    # Create temporary files - result in current directory, trace anywhere
-    with tempfile.NamedTemporaryFile(mode='w', dir=".", suffix='.sol') as f1, \
-            tempfile.NamedTemporaryFile(mode='w', dir=".", suffix='.sol') as f2, \
-            tempfile.NamedTemporaryFile(mode='w') as trace, \
-            tempfile.NamedTemporaryFile(mode='w', dir='.', suffix=".json") as result:
-
-        # Write contract bodies to files
-        f1.write(contract1)
-        f1.flush()
-
-        f2.write(contract2)
-        f2.flush()
-
-        # Build the command
-        command = [
-            'certoraRun.py',
-            f'{f1.name}:{contract1_name}',
-            f'{f2.name}:{contract2_name}',
-            '--equivalence_contracts', f'{contract1_name}={contract2_name}',
-            '--method', abi_signature,
-            '--prover_args', f'-equivalenceCheck true -maxHeuristicFoldingDepth 5 -equivTraceFile {trace.name}',
-            '--tool_output', os.path.basename(result.name),
-            '--loop_iter', str(loop_bound),
-            "--optimistic_hashing",
-            "--optimistic_loop",
-            '--solc', 'solc8.29'
-        ]
-
-        # Run the command without assuming success
-        result_process = subprocess.run(command,
-                                        capture_output=True,
-                                        text=True,
-                                        env={**os.environ, "DONT_USE_VERIFICATION_RESULTS_FOR_EXITCODE": "1"}
-                                        )
-
-        # If non-zero exit, just return
-        if result_process.returncode != 0:
-            return f"The equivalence checker failed with returncode {result_process.returncode}. " \
-                   "It's possible something in your code wasn't handled. " \
-                   "Try a few more times, and then ask for assistance"
-
-        # Load and parse result JSON
-        with open(result.name, 'r') as result_file:
-            result_data = json.load(result_file)
-
-        # Extract the rules dictionary
-        rules_dict = result_data['rules']
-
-        # Get the single key-value pair (since it's a singleton)
-        _, rule_value = next(iter(rules_dict.items()))
-
-        # Check if SUCCESS
-        if rule_value == "SUCCESS":
-            print("Equivalence check passed")
-            return "Equivalent"
-        else:
-            print("Divergent behavior found; returning for refinement")
-            # Read and return trace contents
-            with open(trace.name, 'r') as trace_file:
-                to_return = trace_file.read()
-                tool_logger.info("Trace was:\n%s", to_return)
-                return to_return
-
-
-class ExtraDefinition(BaseModel):
-    definition: str = \
-        Field(description=
-              "A snippet of Solidity that defines some type/error/interface etc. that is needed for the rewrite to work"
-              )
-
-    where: str = \
-        Field(description=
-              "Human readable description of where this definition should be placed. If there is no strong "
-              "guidance/requirement for where the definition lives, 'Nearby' is an acceptable answer"
-              )
-
-    justification: str = \
-        Field(description=
-              "Explanation for why this additional definition is necessary."
-              )
-
-
-class RewriteResultSchema(WithToolCallId):
-    """
-    Used to communicate the successful rewrite to the client. Should only be invoked once the problematic rewritten function has been
-    successfully validated using the equivalence checker; that is, it has returned "Equivalent".
-    """
-    rewrite: str = \
-        Field(description=
-              "The validated; rewritten function. Should consist only of the internal function definition; "
-              "the surrounding external harness should NOT be included."
-              )
-
-    extra_definitions: List[ExtraDefinition] = \
-        Field(description="Any extra definitions that are necessary for the rewrite.")
-
-    remarks: str = \
-        Field(description=
-              "Any explanation of the rewrite. In particular, be sure to justify the use of any inline assembly or "
-              "extra type definitions included"
-              )
-
-
-@tool(args_schema=RewriteResultSchema)
-def rewrite_output(rewrite: str, extra_definitions: List[ExtraDefinition], remarks: str,
-                   tool_call_id: Annotated[str, InjectedToolCallId]) -> Command:
-    return tool_output(
-        tool_call_id=tool_call_id,
-        res={
-            "result": RewriteResultSchema(
-                tool_call_id=tool_call_id,
-                extra_definitions=extra_definitions,
-                remarks=remarks,
-                rewrite=rewrite
-            )
-        }
-    )
-
-
-class HumanInTheLoopSchema(WithToolCallId):
-    """
-    A tool that allows the LLM agent to request human assistance when encountering divergent behaviors
-    during the rewriting process. This tool should be used when the equivalence checker reports
-    differences between the original and rewritten functions that the agent cannot resolve automatically.
-    """
-    question: str = Field(description="The specific question or problem the agent needs help with")
-
-    context: str = \
-        Field(description=
-              "Relevant context about the divergent behavior, including equivalence checker output, "
-              "and what has been tried before (and what didn't work)"
-              )
-
-    original_function: str = Field(description="The original problematic function being rewritten")
-    attempted_rewrite: str = Field(description="The current attempted rewrite that shows divergent behavior")
-
-
-@tool(args_schema=HumanInTheLoopSchema)
-def human_in_the_loop(
-    question: str,
-    context: str,
-    original_function: str,
-    attempted_rewrite: str,
-    tool_call_id: Annotated[str, InjectedToolCallId]
-) -> Command[Literal["tool_result", "error"]]:
-    """
-    Request human assistance to resolve divergent behaviors during rewriting.
-    """
-    # Use LangGraph's interrupt mechanism to pause execution and request human input
-    human_guidance = interrupt({
-        "question": question,
-        "context": context,
-        "original_function": original_function,
-        "attempted_rewrite": attempted_rewrite
-    })
-
-    return tool_return(
-        tool_call_id=tool_call_id,
-        content=f"Human guidance: {human_guidance}"
-    )
-
-
-class ToolError(TypedDict, total=False):
-    error_message: Required[str]
-    tool_stdout: str
-    tool_stderr: str
-
-
-class RewriterState(TypedDict):
-    messages: Annotated[list[AnyMessage], add_messages]
-    code_input: str
-    error: Optional[ToolError]
-    result: Optional[RewriteResultSchema]
-
-
-# Rewrite workflow setup
-rewrite_tools = [
-    rewrite_output,
-    solidity_compiler,
-    equivalence_check,
-    human_in_the_loop
-]
-
-
-# ============================================================================
-# APPLICATION ORCHESTRATION AND CLI INTERFACE
-# ============================================================================
-
-def setup_argument_parser() -> argparse.ArgumentParser:
-    """Configure command line argument parser."""
-    parser = argparse.ArgumentParser(description="Certora Concordance Tool for Solidity Function Rewriting")
-    parser.add_argument("input_file", help="Input Solidity file containing the function to process")
-    parser.add_argument("--harness-model", default="claude-sonnet-4-20250514",
-                        help="Model to use for harness generation (default: claude-sonnet-4-20250514)")
-    parser.add_argument("--rewrite-model", default="claude-opus-4-20250514",
-                        help="Model to use for function rewriting (default: claude-opus-4-20250514)")
-    parser.add_argument("--harness-tokens", type=int, default=1024,
-                        help="Token budget for harness generation (default: 1024)")
-    parser.add_argument("--rewrite-tokens", type=int, default=4096,
-                        help="Token budget for function rewriting (default: 4096)")
-    parser.add_argument("--thinking-tokens", type=int, default=2048,
-                        help="Token budget for thinking in rewriting (default: 2048)")
-    parser.add_argument("--debug", action="store_true",
-                        help="Enable debug logging output")
-    return parser
-
-
-def setup_logging(debug: bool) -> None:
-    """Configure logging based on debug flag."""
-    if debug:
-        logger.setLevel(logging.DEBUG)
-        if not logger.handlers:
-            handler = logging.StreamHandler()
-            handler.setFormatter(logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s'))
-            logger.addHandler(handler)
-
-
-def create_harness_llm(args: argparse.Namespace) -> BaseChatModel:
-    """Create and configure the harness generation LLM."""
-    return ChatAnthropic(
-        model_name=args.harness_model,
-        max_tokens_to_sample=args.harness_tokens,
-        temperature=0,
-        timeout=None,
-        max_retries=2,
-        stop=None
-    )
-
-
-def create_rewrite_llm(args: argparse.Namespace) -> BaseChatModel:
-    """Create and configure the rewrite LLM."""
-    return ChatAnthropic(
-        model_name=args.rewrite_model,
-        max_tokens_to_sample=args.rewrite_tokens,
-        temperature=1,
-        timeout=None,
-        max_retries=2,
-        stop=None,
-        thinking={"type": "enabled", "budget_tokens": args.thinking_tokens}
-    )
-
-
-def generate_harness(harness_llm: BaseChatModel, input_file: str) -> str:
-    """Generate harness for the input function."""
-    runner = build_workflow(
-        state_class=HarnessingState,
-        tools_list=HARNESS_TOOLS,
-        sys_prompt=harness_system_prompt,
-        initial_prompt=harnessing_prompt,
-        output_key="harness_definition",
-        output_schema=HarnessedOutput,
-        unbound_llm=harness_llm
-    ).compile()
-
-    # Read input file
-    with open(input_file, "r") as f:
-        f_def = f.read()
-
-    # Generate harness
-    return runner.invoke(
-        input=GraphInput(code_input=f_def),
-    )["harness_definition"]
-
-def handle_human_interrupt(interrupt_data: dict) -> str:
-    """Handle human-in-the-loop interrupts and get user input."""
-    print("\n" + "=" * 80)
-    print("HUMAN ASSISTANCE REQUESTED")
-    print("=" * 80)
-    print(f"Question: {interrupt_data.get('question', 'N/A')}")
-    print(f"Context: {interrupt_data.get('context', 'N/A')}")
-    print(f"Original Function:\n{interrupt_data.get('original_function', 'N/A')}")
-    print(f"Attempted Rewrite:\n{interrupt_data.get('attempted_rewrite', 'N/A')}")
-    print("-" * 80)
-    return input("Please provide guidance: ")
-
-def display_rewrite_result(result: RewriteResultSchema) -> None:
-    """Display the final rewrite results to the user."""
-    print("\n" + "=" * 80)
-    print("REWRITE COMPLETED")
-    print("=" * 80)
-    print(f"Rewritten Function:\n{result.rewrite}")
-
-    # Format extra definitions nicely
-    if result.extra_definitions:
-        print("\nExtra Definitions:")
-        for i, extra_def in enumerate(result.extra_definitions, 1):
-            print(f"  {i}. {extra_def.definition}")
-            print(f"     Where: {extra_def.where}")
-            print(f"     Justification: {extra_def.justification}")
-            if i < len(result.extra_definitions):  # Add spacing between definitions
-                print()
-
-    print(f"\nRemarks: {result.remarks}")
-
-def execute_rewrite_workflow(rewrite_llm: BaseChatModel, harness: str) -> int:
-    """Execute the rewrite workflow with interrupt handling."""
-    # Add checkpointer for interrupt functionality
-    checkpointer = MemorySaver()
-    rewriter_exec: CompiledStateGraph[RewriterState, None, GraphInput, Any] = build_workflow(
-        state_class=RewriterState,
-        tools_list=rewrite_tools,
-        sys_prompt=simplification_system_prompt,
-        initial_prompt=rewriting_prompt,
-        output_key="result",
-        unbound_llm=rewrite_llm
-    ).compile(checkpointer=checkpointer)
-
-    # Execute rewrite workflow with interrupt handling
-    thread_id = "rewrite_session"
-    config: RunnableConfig = {"configurable": {"thread_id": thread_id}}
-
-    # Start with initial input
-    current_input: Union[None, Command, GraphInput] = GraphInput(code_input=harness)
-
-    while True:
-        assert current_input is not None
-        # Stream execution
-        interrupted = False
-        r = current_input
-        current_input = None
-        for event in rewriter_exec.stream(input=r, config=config):
-            logger.debug("Stream event: %s", event)
-
-            # Check if we hit an interrupt
-            if "__interrupt__" in event:
-                interrupt_data = event["__interrupt__"][0].value
-                human_response = handle_human_interrupt(interrupt_data)
-
-                # Set up for resumption
-                current_input = Command(resume=human_response)
-                interrupted = True
-                break
-
-        # If we were interrupted, continue the loop to resume
-        if interrupted:
-            continue
-
-        state = rewriter_exec.get_state(config)
-        result = state.values.get("result", None)
-        if result is None or not isinstance(result, RewriteResultSchema):
-            return 1
-
-        display_rewrite_result(result)
-        return 0  # Success
-
-def main() -> int:
-    """Main entry point for the concordance tool."""
-    parser = setup_argument_parser()
-    args = parser.parse_args()
-
-    setup_logging(args.debug)
-
-    # Create configured LLMs
-    harness_llm = create_harness_llm(args)
-    rewrite_llm = create_rewrite_llm(args)
-
-    # Generate harness
-    harness = generate_harness(harness_llm, args.input_file)
-
-    # Execute rewrite workflow
-    return execute_rewrite_workflow(rewrite_llm, harness)
-
-if __name__ == "__main__":
-    sys.exit(main())