Context Management

Token Tracking

ContextWindowTracker (in agent-sessions) accumulates input token counts from provider response metadata after each run() call.

const state = session.getContextState();
// { maxTokens: 200000, usedTokens: 85000, usedPercentage: 42.5, remainingPercentage: 57.5 }

Model Context Sizes

Model	Context Window
Claude Sonnet 4.6 / Haiku 4.5	200,000 tokens
Claude Opus 4.6	1,000,000 tokens

Compaction

When the context window fills up, Robota compresses the conversation by generating an LLM summary.

Auto-Compaction

Triggers at ~83.5% of the model's context window. The sequence:

1. PreCompact hook fires
2. Full conversation history sent to LLM with summarization prompt
3. History cleared and replaced with [Context Summary] message
4. Token tracking resets
5. PostCompact hook fires with the summary
6. onCompact callback notifies the UI

Manual Compaction

await session.compact('Focus on the API design decisions');

CLI: /compact focus on API changes

Compact Instructions

CLAUDE.md can define a "Compact Instructions" section. These instructions are automatically extracted during context loading and included in the compaction prompt to preserve project-specific context.

Streaming

The onTextDelta callback provides real-time text as the model generates it:

const session = createSession({
  config,
  context,
  terminal,
  onTextDelta: (delta) => process.stdout.write(delta),
});

The completed response is returned by session.run() after streaming finishes.

Web Search

The Anthropic provider supports server-side web search (web_search_20250305). When enableWebTools is set on the provider instance, the model can search the web during response generation. The onServerToolUse callback fires when search executes.

Abort

Cancel a running session.run():

session.abort(); // Rejects run() with AbortError

Note: The underlying provider API call is not cancelled — the abort only rejects the Session-level promise.