Return an existing provider or create it lazily.

Format one Anthropic-style SSE event.

Apply a parsed CLI transcript event.

Create and configure the FastAPI application.

Emit one structured TRACE row (merged into JSON by the log sink).

Create a minimal IncomingMessage for testing.

Get the tree containing a given node.

Create a new tree with a root node. Args: node_id: ID for the root node incoming: The incoming message

Parse a CLI event and return a structured result. Args: event: Raw event dictionary from CLI log_raw_cli: When True, log ful

Get the singleton instance of the limiter. ``rate_limit`` and ``rate_window`` apply only when the singleton is first created. Call :m

Create a provider instance for testing.

Create a mock streaming chunk.

Count tokens for a request after applying configured model routing.

Feed text and return safe text plus detected tool calls.

Reset singleton (for testing).

Add a reply as a child node to an existing tree. Args: parent_node_id: ID of the parent message node_id: ID

Extract the command prefix for fast prefix detection. Parses a shell command safely, handling environment variables and command injection att

Get an attribute from a Pydantic model, lightweight object, or dict.

Main entry point for handling an incoming platform message.

Create a mock request with all fields build_request_body needs.

Render transcript + optional status line and edit the platform message.

Feed content and yield parsed chunks.

Estimate token count for a request.

Parse an SSE event string into its data payload.

Extract concatenated text from message content.

Get cached settings instance.

Collect all SSE events from a stream.

Escape text for Telegram MarkdownV2.

Flush any remaining tool call in the buffer.

Extract provider type from any 'provider/model' string.

Create a tree with a single root node.

Enqueue a node for processing. If the tree is not busy, processing starts immediately. If busy, the message is queued.

Process a single tool-call delta and yield Anthropic SSE events.

Return whether thinking should be enabled for this request.

Check minimal Anthropic-style SSE invariants: start/stop, block nesting. Does *not* assert strict event ordering (e.g. :class:`message_delta` vs

Build one report outcome from a CLI run and its server log delta.

Map OpenAI or HTTPX exception to specific ProviderError. Streaming transports should pass their scoped limiter (``self._global_rate_limiter``)

Format an SSE event from its event name and data payload.

Record a native Anthropic SSE event and re-emit its normalized shape.

Configure loguru with JSON output to log_file and intercept stdlib logging. Idempotent: skips if already configured (e.g. hot reload). Use fo

Create an Anthropic-compatible message response.

Extract a Claude Code session ID from supported stream-json event shapes.

Format one OpenAI Responses SSE event.

Extract model name from any 'provider/model' string.

Create a MessagesRequest with a single user message.

Core task processor for a single CLI interaction.

Register a node ID to a tree (for external mapping).

Render common Markdown into Discord-compatible format.

Start a new task or continue an existing session. Args: prompt: The user's message/prompt session_id: Option

Minimal settings namespace for AppRuntime (matches typed :class:`Settings` fields used).

Create runtime/outbound components for the configured messaging platform.

Escape text for Discord markdown (bold, italic, etc.).

Return exception type and optionally ``str(exc)`` for operator diagnostics.

Return a readable, non-empty error message for users. Known transport and OpenAI SDK exception types are mapped to stable wording before fall