Functions — github.com/anthropics/scone-bench

hub / github.com/anthropics/scone-bench / functions

Functions94 in github.com/anthropics/scone-bench

⨍ Functions 94 ◇ Types & classes 14

↓ 46 callersFunctionyield_event

(event_type: str, **kwargs)

runner/utils.py:307

↓ 7 callersMethodexecute_tool

Execute a tool on the MCP server

runner/client.py:137

↓ 4 callersMethodrun

Execute a command in the bash shell.

src/scone_bench/tools/bash.py:55

↓ 3 callersFunction_transcript_for_grader

grade_problem ignores the transcript argument; provide a JSON dump for completeness.

runner/utils.py:56

↓ 3 callersFunctionadd_line_numbers

Prepend line numbers to each line, preserving original line endings.

src/scone_bench/tools/edit.py:289

↓ 3 callersMethodconnect_to_server

Connect to an MCP server running in a Docker container.

runner/client.py:54

↓ 3 callersMethodinsert

Insert new_str at the specified line in the file content.

src/scone_bench/tools/edit.py:195

↓ 3 callersFunctionmaybe_truncate

Truncate content and append a notice if content exceeds the specified length.

src/scone_bench/tools/run.py:17

↓ 3 callersMethodread_file

Read the content of a file from a given path; raise a ToolError if an error occurs.

src/scone_bench/tools/edit.py:232

↓ 3 callersMethodwrite_file

Write the content of a file to a given path; raise a ToolError if an error occurs.

src/scone_bench/tools/edit.py:246

↓ 2 callersMethod_calculate_score

Calculate final score from subscores and weights.

runner/transcript_writer.py:88

↓ 2 callersFunction_col

(name: str)

src/scone_bench/server.py:141

↓ 2 callersFunction_make_api_call

()

runner/utils.py:123

↓ 2 callersMethod_make_output

Generate output for the CLI based on the content of a file.

src/scone_bench/tools/edit.py:269

↓ 2 callersFunction_require_address

`addr` flows from the dataset CSV and on-chain proxy-slot reads into argv (cast, heimdall, panoramix) and filesystem paths. Reject anything that i

src/scone_bench/server.py:477

↓ 2 callersFunction_start_anvil

Spawn anvil and block until responsive. Stores process + launch spec globally.

src/scone_bench/server.py:161

↓ 2 callersMethodas_str

(self, lines: list[str])

src/scone_bench/tools/edit.py:377

↓ 2 callersMethodcleanup

Clean up MCP client resources

runner/client.py:157

↓ 2 callersFunctionconvert_anthropic_tool_to_openai

Convert Anthropic tool format to OpenAI function format.

runner/utils.py:157

↓ 2 callersMethodget_file

Download `key` to `local_path`. Returns True on cache hit.

src/scone_bench/cache.py:43

↓ 2 callersFunctionget_problem

(problem_id: str)

src/scone_bench/server.py:127

↓ 2 callersMethodget_prompt

(self)

runner/client.py:72

↓ 2 callersFunctionget_proxy_implementation_address

Return the implementation address if `target` is a proxy, else None. Covers EIP-1167 minimal clones, ERC-1967 implementation slot, and ERC-19

src/scone_bench/proxy.py:8

↓ 2 callersMethodget_tools

(self)

runner/client.py:76

↓ 2 callersMethodis_openrouter_model

Check if the model is an OpenRouter model (format: provider/model-name).

runner/client.py:43

↓ 2 callersFunctionmcp_content_block_to_messages_format

(content: TextContent | ImageContent | AudioContent)

runner/utils.py:1279

↓ 2 callersMethodput_file

(self, key: str, local_path: Path)

src/scone_bench/cache.py:59

↓ 2 callersFunctionrun

Run a shell command asynchronously with a timeout. Returns (return_code, stdout, stderr).

src/scone_bench/tools/run.py:36

↓ 2 callersMethodstart

(self)

src/scone_bench/tools/bash.py:24

↓ 2 callersMethodwrite_event

Write an SSE event to the transcript. Args: event_data: The SSE event string (e.g., "data: {...}\n\n")

runner/transcript_writer.py:40

↓ 1 callersFunction_build_statement

( problem: Problem, chain_id: int, target_addresses: list[str], impl_addresses: dict[str, str]

src/scone_bench/server.py:675

↓ 1 callersFunction_execute_problem

Launch a fresh container for one problem attempt and return the SSE stream.

runner/run_eval.py:26

↓ 1 callersMethod_initialize_file

Initialize the transcript file with metadata.

runner/transcript_writer.py:26

↓ 1 callersFunction_restart_anvil_for_grading

Replace the running anvil with a fresh process so grading sees pristine fork state. The model has unrestricted bash access to anvil's cheat-RPC su

src/scone_bench/server.py:178

↓ 1 callersMethod_update_file

Update the transcript file with a new event.

runner/transcript_writer.py:96

↓ 1 callersFunctionconvert_anthropic_tool_to_responses_api

Responses API tool format is flat: {type, name, description, parameters}.

runner/utils.py:76

↓ 1 callersFunctioncreate_message_with_retry

Call Anthropic API with configurable retry logic.

runner/utils.py:100

↓ 1 callersFunctioncreate_openai_message_with_retry

Call OpenAI API with configurable retry logic.

runner/utils.py:129

↓ 1 callersFunctionexecute_with_stream

()

runner/run_eval.py:134

↓ 1 callersFunctionfetch_contract_source

Populate /workdir/etherscan-contracts/<address>/ with verified source, or a bytecode+decompile fallback if no verified source exists on any explor

src/scone_bench/server.py:487

↓ 1 callersFunctiongenerate_coingecko_info

(target_contract_address: str, chain: str)

src/scone_bench/server.py:231

↓ 1 callersFunctiongenerate_covalent_info

(target_contract_address: str, chain: str, fork_block_number: int)

src/scone_bench/server.py:263

↓ 1 callersFunctiongenerate_preferred_dex_info

(chain: str, block_number: int)

src/scone_bench/server.py:422

↓ 1 callersFunctiongenerate_problem_statement

( target_contract_address: str, implementation_address: str, chain: str, chain_id: int, na

src/scone_bench/server.py:303

↓ 1 callersFunctionget_grade

(chain_name: str)

src/scone_bench/server.py:704

↓ 1 callersMethodget_text

(self, key: str)

src/scone_bench/cache.py:70

↓ 1 callersMethodget_tree

Download and extract a tarball to `dest_dir`. Returns True on cache hit.

src/scone_bench/cache.py:94

↓ 1 callersMethodis_openai_model

Check if the model uses OpenAI-compatible API (includes OpenRouter).

runner/client.py:47

↓ 1 callersFunctionmain

()

runner/run_eval.py:235

↓ 1 callersFunctionprocess_view_output_str

( file_text: str, path: str, total_path_lines: int, max_resp_ln: int, view_range: tuple[in

src/scone_bench/tools/edit.py:303

↓ 1 callersMethodput_text

(self, key: str, text: str)

src/scone_bench/cache.py:84

↓ 1 callersMethodput_tree

(self, key: str, src_dir: str)

src/scone_bench/cache.py:107

↓ 1 callersMethodreplace

Returns a new ToolResult with the given fields replaced.

src/scone_bench/tools/base.py:33

↓ 1 callersFunctionrpc_endpoints

Archive-node RPC URLs per chain, read from `SCONE_RPC_<CHAIN>` env vars. anvil's `--fork-url` requires an archive node (full historical state), s

src/scone_bench/chain_constants.py:1110

↓ 1 callersFunctionrun_evaluation

Run the evaluation with bounded parallelism.

runner/run_eval.py:93

↓ 1 callersFunctionrun_one

(problem, iteration: int)

runner/run_eval.py:121

↓ 1 callersMethodstop

Terminate the bash shell.

src/scone_bench/tools/bash.py:47

↓ 1 callersMethodstr_replace

Replace old_str with new_str in the file content.

src/scone_bench/tools/edit.py:161

↓ 1 callersFunctionstream_agentic_response

( # noqa: ASYNC900 mcp_client: MCPClient, model: str, max_tokens: int, problem_id: str, c

runner/utils.py:918

↓ 1 callersFunctionstream_openai_response

Stream response from OpenAI models.

runner/utils.py:284

↓ 1 callersFunctionsupports_interleaved_thinking

Check if model supports interleaved thinking (Claude 4 models only).

runner/utils.py:88

↓ 1 callersFunctiontruncate_from_middle_v2

If no truncation is needed, returns the original string; otherwise a TruncatedString.

src/scone_bench/tools/edit.py:385

↓ 1 callersFunctionuses_responses_api

Route OpenAI models that require /v1/responses (vs chat.completions).

runner/utils.py:71

↓ 1 callersMethodvalidate_path

Check that the path/command combination is valid.

src/scone_bench/tools/edit.py:67

↓ 1 callersMethodview

Implement the view command.

src/scone_bench/tools/edit.py:85

↓ 1 callersFunctionwait_for_anvil

Wait for anvil to be ready, with exponential backoff.

src/scone_bench/server.py:201

Method__add__

(self, other: "ToolResult")

src/scone_bench/tools/base.py:18

Method__bool__

(self)

src/scone_bench/tools/base.py:15

Method__call__

(self, command: str | None = None, restart: bool = False, **kwargs)

src/scone_bench/tools/bash.py:125

Method__call__

( self, *, command: Command, path: str, file_text: str | None = None,

src/scone_bench/tools/edit.py:31

Method__init__

( self, problem_id: str, container_id: str, model: str | None = None, )

runner/client.py:18

Method__init__

(self, file_path: str, problem_id: str, container_id: str)

runner/transcript_writer.py:12

Method__init__

(self)

src/scone_bench/cache.py:20

Method__init__

(self)

src/scone_bench/tools/bash.py:20

Method__init__

(self)

src/scone_bench/tools/bash.py:122

Method__init__

(self, message)

src/scone_bench/tools/base.py:49

Method__init__

(self, run_command_preexec_fn=demote)

src/scone_bench/tools/edit.py:28

Functionbash

(command: str | None = None, restart: bool = False)

src/scone_bench/server.py:76

Methodcombine_fields

(field: str | None, other_field: str | None, concatenate: bool = True)

src/scone_bench/tools/base.py:19

Functiondemote

Drop privileges to uid/gid 1000. Intended as a `preexec_fn` so model-issued commands cannot read the MCP server directory or interfere with t

src/scone_bench/tools/run.py:26

Methoddemote

()

src/scone_bench/tools/bash.py:29

Methodenabled

(self)

src/scone_bench/cache.py:40

Methodfinalize

Finalize the transcript with completion status and optional grade.

runner/transcript_writer.py:131

Methodfull_truncated_str

(self)

src/scone_bench/tools/edit.py:381

Functiongrade_problem

Grade the model's solution. Returns a Grade with subscores and profit.

src/scone_bench/server.py:831

Functionhello

()

src/scone_bench/__main__.py:20

Functionmain

()

src/scone_bench/server.py:861

Functionmcp

()

src/scone_bench/__main__.py:25

Methodscore

(self)

src/scone_bench/spec.py:17

Functionsetup_problem

Starts the environment and returns the problem statement.

src/scone_bench/server.py:603

Functionstr_replace_editor

( *, command: Command, path: str, file_text: str | None = None, view_range: list[int] | No

src/scone_bench/server.py:90

Methodto_api_tool

(tool: Tool)

runner/client.py:78

Functionwith_traceback

(func)

src/scone_bench/server.py:589

Functionwrapper

(*args, **kwargs)

src/scone_bench/server.py:591