- overview: new paragraph framing session.out as bounded (~one turn at
  steady state), trim record after each turn-complete, durable snapshot
  for full history, eventually-consistent trim window for resume
- new "Records on session.out" subsection naming the three record kinds
  (data, trigger control, S2 command) and the uniform header-based
  filter rule custom transports need
- trigger:turn-complete and trigger:upgrade-required rewritten as
  header-form control records under the trigger-control namespace;
  legacy chunk.type shape called out as the prior wire
- self-contained parser example updated to route the three record kinds
- refreshing-the-token and validation-error blocks updated for the
  public-access-token header on turn-complete
- end-to-end curl recipe updated to grep on trigger-control
- resuming-a-stream block spells out single-turn vs multi-turn-away
  resume contract
- backend / chat history: cross-reference to records taxonomy and a note
  that the agent's accumulator (not session.out) is the source of truth
  for full history
- changelog: headline entry for the wire change + dashboard snapshot
  read on the Sessions detail page

A conceptual deep dive sitting between Overview and Backend. Covers the
chat session as a pair of S2 channels and a long-lived task, the run's
lifecycle states (Cold start, Streaming, Idle, Suspended, Resuming,
Continuation, Closed), a step-by-step trace of one turn, and the three
persistence layers that survive idle gaps, deploys, refreshes, and crashes:
the engine checkpoint (CRIU today, full Firecracker VM snapshots on the new
microVM compute), the chat-level S3 snapshot, and the browser's
lastEventId cursor. Closes with warm-vs-resumed-vs-continuation timings,
a hooks pointer table, a good-fit subsection, and a single-turn Mermaid
sequence diagram.

…cords

Document the new sibling header that rides on every turn-complete
record carrying the agent's committed-consume cursor on .in. The
header is what lets the next worker boot seed its .in SSE subscribe
past already-processed user messages without a wall-clock-derived
dedup cutoff.

- client-protocol: new row in the turn-complete header table; updated
  records-on-session-out summary; example block shows the third
  header
- changelog: short paragraph under "what changed on the wire" framing
  the cursor as the seq-based replacement for the prior timestamp
  cutoff, with an explicit "custom transports should ignore" note

Document the new string-or-function baseURL and fetch override on
TriggerChatTransport, AgentChat, and chat.createStartSessionAction.

- New patterns/trusted-edge-signals page covering the trusted-proxy
  pattern: how to route .in/append + session-create through an edge
  worker for server-trusted metadata injection while keeping .out SSE
  direct, plus a complete Cloudflare Worker reference implementation,
  agent-side clientDataSchema example, and threat model.
- reference: TriggerChatTransport options table now lists baseURL with
  the function shape, adds a fetch row, and drops streamBaseURL. New
  AgentChat options and createStartSessionAction options tables with
  the same shape (different endpoint enum: "sessions" | "auth").
- frontend: self-hosting section shows the function form alongside the
  string form, plus a one-line pointer to trusted-edge-signals.
- server-chat: AgentChat options table picks up baseURL and fetch rows.
- docs.json: nav entry for the new patterns page.

…release-20260519091352

…eserving default

…s caught

…(no graceful exits)

Comprehensive pass on the AI Agents + Prompts docs ahead of cutting
v4.5.0-rc.0:

- RC banner across every AI chat + Prompts page (single snippet) with
  links to a new Compatibility table covering ai SDK v5/v6, React 18/19,
  and Node 18.20+ peer ranges.

- Restructured the Agents sidebar into four sub-groups (Building agents,
  Features, Patterns, Reference) and hoisted Agents to the top level of
  the docs nav.

- New Sessions reference page documenting sessions.start / retrieve /
  list / open, SessionHandle.in / .out, control records, and the
  authorization model.

- Split the catch-all Features page into proper homes: chat.local moves
  to its own page under Building agents, chat.defer joins background
  injection, custom data parts + raw chat.stream move into Backend, and
  the target:"root" + ai.toolExecute walkthroughs move into the
  Sub-agents pattern.

- Fixed broken code examples and cross-link anchors: auth.createPublicToken
  scopes shape, useTriggerChatTransport import path, AgentChatOptions
  table, runInMockTaskContext re-export, and anchor slugs that didn't
  match Mintlify's slugifier.

The overview was carrying internals (sequence diagrams, suspend timing,
inbox patterns, "what the backend accumulates") before a reader saw a
runnable line of code. Rewritten to match how the product is actually
pitched:

- Opener leads with "An AI chat isn't a request — it's a session"
- Minimal example (chat.agent + useChat) appears in the first H2, not
  after three screens of mechanics
- Outcome bullets cover durability, native AI SDK support, multi-turn,
  Head Start, production primitives, and observability
- "How it fits together" names the three primitives (chat agents,
  Sessions, sub-agents via AgentChat) without explaining internals
- CardGroup CTAs to Quick Start / How it works / Backend / Patterns

Length: ~520 words (was ~1180). Internals that were duplicated here
already live on how-it-works, sessions, and backend.

…t guidance

Sweep across the AI chat docs against the v4.5 release candidate:

- Every `streamText` example uses Anthropic (`anthropic("claude-sonnet-4-5")`)
  instead of OpenAI, with imports rewritten. AI SDK v5+ defaults to a
  single step, so every example also now sets `stopWhen: stepCountIs(15)`
  to avoid the "tool calls fire but no follow-up text" gotcha.

- `startSession` destructure drops the unused `taskId` — the customer's
  wrapped action is task-bound at creation time, and the SDK-side
  `chat.createStartSessionAction` now accepts a typed `clientData` field
  via a `<typeof yourChat>` generic.

- Quick start gains a "before you start" paragraph linking to manual
  setup, and a note that the `@/*` path alias is Next.js-specific.

- Frontend page opens with a new "How the transport works" section
  explaining the no-API-route architecture — the browser uses a session-
  scoped PAT to call `.in/append` and subscribe to `.out` directly,
  bypassing any route handler `useChat` would normally need.

- Backend page now has an explicit "Which form to call" table for
  `chat.toStreamTextOptions()` — when to call with no args, when to pass
  `{ registry }`, when to pass `{ tools }`.

- `auth.createPublicToken` is called out as resolving to `Promise<string>`
  (the JWT), not an object — clears up a small ambiguity in the docs.

First release candidate cut of v4.5. Everything covered by the
0.0.0-chat-prerelease-* entries below now ships under a stable
semver tag.