Merge branch 'codex/goal-session' into codex/commands

# Conflicts:
#	docs/config-catalog.md
#	docs/module-graph.md
#	examples/acp-agent/tests/snapshots/error-finish/stdout.expected.jsonl
#	packages/ui/acp/README.md
#	packages/ui/tui/README.md

docs/core-data-structures/core.md

@@ -226,7 +226,7 @@ interface GenerateOptions {
 }
 ```
 
-Why a model response stopped is a merge-extensible reason:
+Why a model response stopped is a merge-extensible reason. Terminal provider failures carry the streaming contract's [`LlmFailure`](llm-streaming.md#llmfailure):
 
 ```ts type-equiv
 /**
@@ -237,8 +237,8 @@ interface FinishReasonMap {
   'stop': { kind: 'stop' }
   'tool-calls': { kind: 'tool-calls' }
   'max-tokens': { kind: 'max-tokens' }
-  'aborted': { kind: 'aborted' }
-  'error': { kind: 'error'; message: string; code?: string }
+  'aborted': { kind: 'aborted'; failure: LlmFailure }
+  'error': { kind: 'error'; failure: LlmFailure }
 }
 ```
 
@@ -450,14 +450,14 @@ type ContinuationDecision =
   | { action: 'continue'; reason?: { content: ContentBlock[]; source: MessageSource } }
 ```
 
-`agent/request-error` receives the original `RequestError`, whose optional provider-neutral `code` supports stable routing without message parsing:
+`agent/request-error` receives the exact original `RequestError` beside its immutable `LlmFailure`, an immutable list of failures that already authorized another request in the consecutive sequence, the turn signal, and `next()`. Recovery plugins route on `failure.code`, not the live error's message; each policy counts only its own codes, and a successful request clears the history:
 
 ```ts type-equiv
 /** Model-request failure with an optional machine-routable provider code. */
 type RequestError = Error & { code?: string }
 ```
 
-It returns a `RequestErrorDecision`; `retry` opens a new numbered step after the recovery listener's durable mutation, while `fail` preserves that error:
+It returns a `RequestErrorDecision`; `retry` opens a new numbered step after the recovery listener's durable mutation, while `fail` retains the structured failure on `turn/end`:
 
 ```ts type-equiv
 /** Failed-request recovery decision; `retry` opens another numbered step while listeners delegate by calling `next()`. */

docs/core-data-structures/llm-streaming.md

@@ -31,18 +31,40 @@ type StreamChunk =
   }
 ```
 
+## `LlmFailure`
+
+Every thrown or in-band final-adapter failure normalizes to one serializable provider-neutral payload. `providerRetryAfterMs` is a validated positive delay requested by the provider, not a retry decision; `ProviderRequestId` is an opaque branded string for diagnostics.
+
+```ts type-equiv
+/** Serializable provider-boundary facts; policy decides whether they are retryable. */
+interface LlmFailure {
+  /** Human-readable provider or transport failure. */
+  readonly message: string
+  /** Stable provider-neutral machine-routing code. */
+  readonly code: string
+  /** HTTP status observed at the provider boundary, when available. */
+  readonly status?: number
+  /** Provider-requested delay in milliseconds, when valid and available. */
+  readonly providerRetryAfterMs?: number
+  /** Opaque provider-issued request identifier for diagnostics. */
+  readonly requestId?: ProviderRequestId
+}
+```
+
 ## The adapter contract
 
 Every adapter MUST obey these, and every consumer may rely on them:
 
 - **`usage` before `finish`, nothing after `finish`.** Defer both to the provider's end-of-stream marker so a trailing usage-only chunk can't violate the ordering.
 - **Tool-call `arguments` stay raw JSON strings end-to-end.** Partial fragments stream via `argumentsDelta`; a provider that hands back parsed objects re-stringifies at `block-end`.
-- **Two sanctioned error paths.** A failure may either THROW from `stream()` (transport/protocol errors) **or** end the stream with `finish {kind:'error'|'aborted'}` (provider in-band errors, for adapters that can't throw mid-stream). Consumers must handle *both*. The agent loop closes the failed step and offers either form to `agent/request-error`; absent recovery it becomes a turn error, and no normal completed assistant message is logged for that request.
+- **Two sanctioned error paths, one fact shape.** A failure may either THROW from `stream()` (transport/protocol errors) **or** end the stream with `finish {kind:'error'|'aborted', failure}` (provider in-band errors, for adapters that can't throw mid-stream). `LlmError.failure` carries the same `LlmFailure`. The final adapter boundary preserves the exact thrown `Error` object and associates immutable facts with that call; the agent loop closes the failed step and offers the error, facts, and immutable prior-retried facts to `agent/request-error`. Absent recovery the structured failure becomes the turn error, and no normal assistant message or tool side effect is committed for that attempt.
+- **One adapter call is one provider attempt.** Adapters disable library retries. Agent-level recovery opens another durable numbered step; direct `ctx.llm.stream()` callers remain single-attempt.
+- **Provider stalls are bounded at the transport.** Both shipping remote adapters expose positive finite `streamIdleTimeoutMs` with a five-minute default. The watchdog arms only while iterator `next()` is outstanding, uses one stable signal for the whole request, maps its own expiry to `TIMEOUT`, and keeps an earlier caller abort as `ABORTED`.
 - **Context overflow has one canonical code.** Both DeepSeek adapters classify explicit provider detail through `isContextWindowExceededError()` and surface `CONTEXT_WINDOW_EXCEEDED`, whether the failure arrives as a thrown HTTP `LlmError` or an in-band finish error. Consumers route on the code, never provider text.
 - **Every provider HTTP request carries the app-attribution header.** Adapters send `attributionHeaders()` (below) - the `User-Agent` baseline - and prove it with a wire-level test (mock server asserting the received header, or the library's header hook for a library-backed adapter).
 - **Replay state is adapter-owned.** A successful `finish` may carry lossless-JSON state needed to reconstruct a native provider response. The loop stores it with the assembled assistant message unless an `agent/step-result` listener rewrote the content. On a later request, `LlmService` passes the state only when the historical provider and target provider are currently registered to the exact same adapter instance. That adapter validates the state and owns any cross-model or cross-provider conversion; other adapters receive the provider-neutral content and provenance without the private state.
 
-This contract was pinned down by two deliberately independent implementations: `dsh-llm-deepseek` (hand-rolled fetch/SSE) and `dsh-llm-pi-ai` (a generic multi-provider adapter through `@earendil-works/pi-ai`). The library-backed adapter cannot throw mid-stream, so it exercises the finish-chunk error path the hand-rolled one might not.
+This contract is pinned down by two deliberately independent implementations: `dsh-llm-deepseek` (hand-rolled fetch/SSE) and `dsh-llm-pi-ai` (a generic multi-provider adapter through `@earendil-works/pi-ai`). The library-backed adapter exercises the finish-chunk error path, while transport-boundary tests prove each idle watchdog stops its actual request.
 
 ## `AppIdentity` — app attribution
 

docs/core-data-structures/session.md

@@ -430,7 +430,7 @@ declare class Session {
 - `context/message` → a user-role message carrying its `content` verbatim at its chronological position. Optional JSON `meta` remains in the event log and is never rendered.
 - `steering/message` → a user-role message carrying its content verbatim at its chronological position.
 
-Everything else (`turn/*`, `step/*`) is structural and does not project into a message. Token usage is observed on `assistant/message.usage` (the step that produced it); an operational error's step number is on `turn/end.reason` for `kind: 'error'`. Because this unreleased format intentionally has no compatibility promise, seed/load validation rejects request headers without provider+model and assistant messages without provider/model provenance instead of guessing a route for historical data.
+Everything else (`turn/*`, `step/*`, plugin-owned `llm/retry`) is structural and does not project into a message. Token accounting reads per-step `assistant/chunk { type: 'usage' }` records and treats `assistant/message.usage` as the committed-step fallback when no usage chunk exists; failed model-request attempts have no assistant message, so their usage chunk is the durable accounting record. An operational error's step number is on `turn/end.reason` for `kind: 'error'`, with normalized `LlmFailure` facts for a final model-request failure and message/code for other live errors. Because this unreleased format intentionally has no compatibility promise, seed/load validation rejects request headers without provider+model and assistant messages without provider/model provenance instead of guessing a route for historical data.
 
 ## Live-session fork API
 
@@ -474,9 +474,13 @@ interface TurnEndReasonMap {
    * The turn failed: a step threw or the model reported a failure. `step` is the
    * step number the failure occurred on (the operational error's location — the
    * single durable record of an in-turn failure; live diagnostics also fire via
-   * `agent/error`). `code` is the error's code when one was attached.
+   * `agent/error`). Final model-request failures retain their normalized facts
+   * as one `failure`; other turn failures retain their live Error message/code.
    */
-  error: { kind: 'error'; step: number; message: string; code?: string }
+  error: { kind: 'error'; step: number } & (
+    | { failure: LlmFailure; message?: never; code?: never }
+    | { message: string; code?: string; failure?: never }
+  )
   disposed: { kind: 'disposed' }
   /** At least one step reached its output-token ceiling, even if a plugin continued the turn. */
   'max-tokens': { kind: 'max-tokens' }