Send a JSON-RPC notification to the LSP server

Collect output files from output directory.

Create a new execution record.

Ensure AutoSkills is initialized (lazy initialization). Returns: True if AutoSkills is available and initialized.

Execute search using the specified engine. The search pipeline with optimization: 1. Execute search query 2. (Optiona

Load outline.json.

Add a message to a session

Get LLM configuration from environment variables.

Get session by ID

Get the path to claude_skills directory.

Get work directory from env or create temp directory.

Trigger a specific callback hook across all registered callbacks. Args: point (str): Name of the callback method to call

Thread-safe progress update.

Assert common DAG result validations.

Calls the OpenAI chat completion API with the provided messages and tools. Args: messages (`List[Message]`): Formatted message hi

创建多模态配置 Args: stream: 是否启用流式输出 Returns: Config: 配置好的 Config 对象，如果 API Key 未设置则返回 None

Execute code in EnclaveSandbox.

Heuristic report locator: - v2 reporter tool: <workdir>/reports/report.md - legacy workflows: <workdir>/report.md - fallback: <workdi

Get skills configuration from agent config.

Generate LLM response from prompt.

Run subprocess locally with security restrictions. Cross-platform support with timeout and resource limits. Args:

Parse XML fragment and return list of imported nodes. Args: xml_content: String containing XML fragment Returns

Parse JSON from LLM response with robust extraction.

Parse sandbox execution results into stdout, stderr, exit_code.

Process HTML inline elements and add formatted runs to paragraph

Save index.json.

Save outline.json and render outline.md.

Check code for potentially dangerous patterns. Args: code: Code string to check. is_local: If True, use stri

Best-effort terminate/kill a subprocess without raising. We intentionally keep this conservative: first SIGTERM, then SIGKILL if needed.

Call a tool. Args: server_name(`str`): The server name of the tool. tool_name: The tool name. tool_args:

Cleanup resources.

Initialize search engines and content fetcher.

Estimate token count from text. Simple heuristic: ~4 chars per token for mixed content.

Get the tool name for this engine.

Initialize the optimizer (only summarizer needs initialization).

Initialize the LSP server and wait for it to be ready

Insert before with automatic attribute injection.

Checks if a package is installed by attempting to import it. Args: package_or_import_name: The name of the package or import as a string

Load .env into the process env, then merge with ``envs`` and return.

Parse the package dependencies listed in a requirements file but strips specific versioning information. Args: fname (str): path

Stop the agent

Encodes a given text using SHA256 and returns the last 8 characters of the hexadecimal representation. Args: text (str): The inp

Write a new evidence note.

Get path string for a resource file. Tries relative path first, falls back to absolute path. Args: file_path: P

Generate a unique key for a skill based on its ID and version. Args: skill: SkillSchema object Returns:

Initialize the sirchmunk AgenticSearch instance.

Only accept top-level final report files in `workdir`: - <workdir>/final_report.md - <workdir>/report.md We intentionally ignore fil

Check if a directory is a valid skill directory. Args: path: Path to check Returns: True if directo

Load conflict.json.

Load a single skill from directory. Args: skill_dir: Path to skill directory Returns: SkillSchema o

Merges a partial message into the last message in the message list. Args: messages (`List[Message]`): Current list of messages.

Read the content of a file. Args: file_path: Path to the file Returns: Content of the file as a str

Read a JSON-RPC message from the LSP server

Update file content and check for errors

Aggregate the streaming chunks into a single response dictionary within current round of chat. stream_chunks: A list of

Checks whether a specified Python package is available in the current environment. If the package is not found, an AssertionError is raised

Get chat response from OpenAI API using streaming. messages: A list of dictionaries representing the chat messages.

Disconnect and clean up the tool. Returns: None Raises: Exceptions if anything goes wrong.

Initialize directory structure.