Skip to main content

Enumeration: ELLMClientEventKind

Defined in: internal/llm-client/types.ts:343

Event kinds a client can render as progress.

Enumeration Members​

ANSWER_WITHDRAWN​

ANSWER_WITHDRAWN: "ANSWER_WITHDRAWN"

Defined in: internal/llm-client/types.ts:372

The answer streamed by this call is not kept (D74): it failed the output checks, or the call was aborted while they ran (D90). The last event of the call, after DONE. A client drops or greys the text it rendered for the call and shows issues; the response is FAILED with the same issues. An abort during the model call itself sends nothing after the deltas: the response says the call failed.

DONE​

DONE: "DONE"

Defined in: internal/llm-client/types.ts:363

The call is complete; the response follows. Sent once per call, after the last step.

ERROR​

ERROR: "ERROR"

Defined in: internal/llm-client/types.ts:365

The call failed; the response says why.

RUN_ENDED_BY_TOOL​

RUN_ENDED_BY_TOOL: "RUN_ENDED_BY_TOOL"

Defined in: internal/llm-client/types.ts:359

A tool with endsRun ran (D86): sent once, after the STEP_END of its step and before the final step that phrases the answer. A client can show "handed over" before the closing text streams.

RUN_PAUSED​

RUN_PAUSED: "RUN_PAUSED"

Defined in: internal/llm-client/types.ts:354

The run waits for a human: the pending calls in toolCalls, or none when the answer waits for a review (D99).

STEP_END​

STEP_END: "STEP_END"

Defined in: internal/llm-client/types.ts:361

A step of the run is complete: its model call and its tools.

TEXT_DELTA​

TEXT_DELTA: "TEXT_DELTA"

Defined in: internal/llm-client/types.ts:345

A chunk of the model's text, in text.

TOOL_REQUEST​

TOOL_REQUEST: "TOOL_REQUEST"

Defined in: internal/llm-client/types.ts:350

The model asked for a tool call; its TOOL_RESULT follows once the tool answered. A call the abort cut (D90) gets no TOOL_RESULT event: the response says the call failed, and the thread holds the error result.

TOOL_RESULT​

TOOL_RESULT: "TOOL_RESULT"

Defined in: internal/llm-client/types.ts:352

A tool answered (or was denied): the call in toolCall, what the model reads in toolOutput.