agentfootprint
Version:
The explainable agent framework — backtrack a wrong answer to the exact context that caused it (evidence, not guesses). Built on footprintjs.
36 lines (35 loc) • 1.63 kB
TypeScript
/**
* agentfootprint/observe — everything that watches.
*
* One door for the whole watching story, which used to be nine:
*
* • Recorders — the Tier 1/2/3 observers, `recordRun`, the
* offline `Trace`, the context ledger.
* • Diagnosis — influence scoring, trace toolpack, context-bisect,
* tool lint, and the pluggable context-error finders.
* • Vendor sinks — AgentCore / CloudWatch / X-Ray / OTel strategies,
* plus the tamper-evident audit export.
* • Strategy ports — the observability / cost / live-status / lens
* interfaces and their `attach*` helpers.
* • Delivery — `toSSE` and friends, for streaming a run to a
* browser.
* • Status + words — the chat-bubble state machine AND the message
* catalogs it renders with. The logic and the prose
* are one story, so they are one door.
*
* The typed event stream those observers consume is its own door,
* `agentfootprint/events` — it is the wire vocabulary, not a watching tool.
*
* @example
* ```ts
* import { recordRun, toolChoiceRecorder, selectStatus } from 'agentfootprint/observe';
* ```
*/
export * from '../observe.js';
export * from '../observability-providers.js';
export * from '../strategies/index.js';
export * from '../hosting/stream.js';
export * from '../status.js';
export * from '../debug.js';
export * from '../debug/finders.js';
export * from '../locales/index.js';