@cubicecho/agent-core 4.4.0 → 4.6.0

package/README.md

@@ -481,6 +481,34 @@ await runAgentLoop({
 });
 ```
 
+### What the calls made
+
+A tool that wrote a document or saved a file can say so, and `dispatch` can pass that on: it
+resolves to a `ToolAnswer`, `{ text, links }`, in place of the string. `text` is the whole of
+what goes into the transcript, so the request, the prompt cache, pruning and compaction never see
+the links. `links` are `ResourceLink`s, MCP's own shape declared structurally, so what
+agent-mcp-pool's `call` answers with `links: true` is a dispatcher's answer as it stands.
+
+```ts
+const { artifacts } = await runAgentLoop({
+  // ...
+  dispatch: ({ name, args }, signal) => pool.call(name, args, { signal, links: true }),
+});
+// [{ link: { uri, name, title?, ... }, callId, tool, step, count }]
+```
+
+The loop collects them into the run's `artifacts`: one per uri in the order first seen, each with
+the call, tool and step that first linked to it and the number of calls that did. A link is what
+one call answered with; an artifact is what the run made. `onToolResult` hears a call's `links`
+beside its `content`, an `artifact` event goes out the first time each uri appears — `name` the
+tool, `text` the title or else the uri, and `link` the link — and `runMetrics` counts those events
+as `artifacts`. A deduplicated repeat shares the first call's answer and links but adds nothing to
+the count, since nothing was made twice. A call that threw made nothing. A failed run carries what
+its calls made on the error too, read off with `failedRun`.
+
+A link meant for the model alone — one whose MCP `audience` leaves out `"user"` — is the
+dispatcher's to drop, since that is MCP's rule and this package knows nothing of MCP.
+
 Arguments go through `parseToolArguments`, which is lenient where the model's meaning is plain:
 JSON held in a string is opened, and the almost-JSON local models write — single quotes, Python's
 `True` and `None`, bare keys, a trailing comma — is repaired, without touching what is inside a
@@ -815,9 +843,15 @@ once `used` (the last turn's prompt tokens, or the estimate) is past three quart
 The kept tail fills at most 35% of it and starts on a user message, since a transcript resuming
 mid-exchange is one servers refuse; leading system prompts are never folded, and an earlier
 summary is continued rather than summarised. `compactTranscript` writes the summary and tells
-`beforeCompact` hooks what is going while it does. Because the cut lands on
-a user message, one long tool run under a single question has nothing to fold — pruning is what
-keeps that one going.
+`beforeCompact` hooks what is going while it does.
+
+A run that is one question and a long tool loop has no user message to cut at, so there the plan
+cuts inside the loop instead: the question is kept whole (the plan's `question`), the calls and
+results after it are folded up to an assistant message, and the request becomes the summary, the
+question, then the rest of the loop — every tool result still after its call. A transcript with a
+user message to cut at is planned as before. `runAgentLoop({ compact: true })` folds such a run on
+its own; a host that numbers messages across steps passes `hooks.offset`, the stored index of the
+run's first message, so the ids `beforeCompact` is told match its own.
 
 ```ts
 beforeStep: async (messages, step) => {
@@ -950,8 +984,8 @@ leading `system` messages, and reads an earlier summary back out of a `SUMMARY_L
 them. A host whose system prompt is a separate argument and whose summary is a column has neither
 in the array, and knows both exactly. Given them, nothing is scanned.
 
-`applyCompaction(messages, fold)` is the way back — the summary as a `system` message, then
-everything from `through` — and it writes the same `SUMMARY_LEAD` `planCompaction` looks for, so
+`applyCompaction(messages, fold)` is the way back — the summary as a `system` message, then the
+fold's `question` if it was made inside one, then everything from `through` — and it writes the same `SUMMARY_LEAD` `planCompaction` looks for, so
 the next fold continues those notes rather than summarising them a second time. No fold yet hands
 back the messages themselves.
 

package/dist/context/compaction.d.ts

@@ -125,9 +125,17 @@ export interface CompactionOptions {
 export interface CompactionPlan {
     /** The first message folded away. Everything before it is a system prompt and stays. */
     from: number;
-    /** The first message kept whole, always a user message. */
+    /**
+     * The first message kept whole: a user message, or an assistant message on a fold made inside
+     * one question's tool loop (see `question`).
+     */
     cut: number;
-    /** The messages from `from` to `cut`, the ones the summary replaces. */
+    /**
+     * The user message kept whole between the summary and `cut`, on a fold made inside its tool
+     * loop. Absent on a fold that cuts at a user message.
+     */
+    question?: number;
+    /** The messages from `from` to `cut`, `question` left out: the ones the summary replaces. */
     toSummarise: Message[];
     /** The summary an earlier compaction left, which this one continues. */
     previous?: string;
@@ -155,6 +163,13 @@ export interface CompactionPlan {
  * than summarised as if it were conversation. No plan comes back when the window is not full
  * enough, or when the only legal cut folds too little to pay for the summary.
  *
+ * A transcript that is one question and a long tool loop has no user message to cut at, so where
+ * no user cut folds anything the plan cuts inside the last question's loop instead: the question
+ * is kept whole as `question`, the assistant and tool messages after it are folded up to an
+ * assistant message, and the tail resumes there — every tool result in it still follows its call.
+ * `applyCompaction` sends the question after the summary. A transcript with a user cut is planned
+ * exactly as it was.
+ *
  * Both of those are recovered by reading the transcript, which is what a host whose array holds
  * everything it sends has to do. A host that keeps its fold as a record beside an append-only
  * transcript — its system prompt a separate argument, no summary message in the array at all —
@@ -206,6 +221,11 @@ export interface CompactionRecord {
     summary: string;
     /** Index into the transcript the plan was made for: the first message still sent whole. */
     through: number;
+    /**
+     * The user message sent whole after the summary and ahead of `through`, on a fold made inside
+     * that question's tool loop. Absent on a fold that cut at a user message.
+     */
+    question?: number;
     /** ISO 8601, so the chat can show where history was folded and how stale the notes are. */
     at: string;
 }
@@ -213,7 +233,8 @@ export interface CompactionRecord {
 export interface CompactionRunOptions {
     /**
      * Hooks to tell. `context` is extended with `compacting` and `range`, whose indexes are the
-     * plan's — and so the host's own, for a plan made over a stored transcript. `honourVeto` waits
+     * plan's — and so the host's own, for a plan made over a stored transcript. `compacting` leaves
+     * out a `question` the fold keeps. `honourVeto` waits
      * for the hooks and lets one stop the compaction; without it they add no latency.
      */
     hooks?: {
@@ -221,6 +242,14 @@ export interface CompactionRunOptions {
         context: HookContext;
         onNote?: OnNote;
         honourVeto?: boolean;
+        /**
+         * What the transcript's first message is numbered as in the `compacting` uuids: the host's
+         * stored index of `messages[0]`, for a host that numbers messages across steps or requests.
+         * See `turnMessages`.
+         *
+         * @defaultValue `0`
+         */
+        offset?: number;
     };
     /**
      * The window is already exceeded — the caller caught a `ContextOverflow`, or is compacting to
@@ -256,7 +285,8 @@ export declare function runCompaction(messages: Message[], plan: CompactionPlan,
  * Absent, the leading `system` messages are found by scanning, and zero of them is the ordinary
  * case for a host whose system prompt is a separate argument.
  * @returns `messages` itself when there is no record or its summary is blank, otherwise a new
- * array: the kept head, one summary message, then everything from the record's `through` on.
+ * array: the kept head, one summary message, the record's `question` when it has one and the head
+ * does not already hold it, then everything from the record's `through` on.
  *
  * @remarks
  * The inverse of storing a `CompactionRecord`, and the shape `planCompaction` expects to meet
@@ -264,7 +294,7 @@ export declare function runCompaction(messages: Message[], plan: CompactionPlan,
  * continues these notes rather than summarising them a second time. Any earlier summary message in
  * the kept head is dropped, since the record's already contains it.
  */
-export declare function applyCompaction(messages: Message[], record?: Pick<CompactionRecord, 'summary' | 'through'>, { from }?: {
+export declare function applyCompaction(messages: Message[], record?: Pick<CompactionRecord, 'summary' | 'through' | 'question'>, { from }?: {
     from?: number;
 }): Message[];
 /**
@@ -278,7 +308,8 @@ export declare function applyCompaction(messages: Message[], record?: Pick<Compa
  * prompts, when the host keeps them in the array. Zero is the stored-fold case,
  * where the summary is the request's first message.
  * @returns The index in the request. One before the record's `through` gives `head`, where the
- * summary sits — an index among the kept head messages themselves included.
+ * summary sits — an index among the kept head messages themselves included — except the record's
+ * `question`, which is sent whole right after the summary.
  *
  * @remarks
  * A transcript that stays append-only and a request that does not are two numberings of the same
@@ -286,7 +317,7 @@ export declare function applyCompaction(messages: Message[], record?: Pick<Compa
  * hook — has to say which it is in. An index inside the folded stretch answers with the summary
  * message that now stands for it.
  */
-export declare const requestIndex: (index: number, record?: Pick<CompactionRecord, 'through'>, head?: number) => number;
+export declare const requestIndex: (index: number, record?: Pick<CompactionRecord, 'through' | 'question'>, head?: number) => number;
 /**
  * The transcript with the plan's stretch replaced by one system message holding its summary.
  *

package/dist/context/compaction.js

@@ -142,6 +142,56 @@ export function pruneToolResults(messages, { keepLast = 5, maxChars = 256 } = {}
  * @returns `index` where it lies from zero to `most`, otherwise the nearer of the two.
  */
 const clamp = (index, most) => Math.min(Math.max(index, 0), most);
+/**
+ * Where to cut, by `target` when there is one and by `keepRatio` otherwise, among the indexes
+ * `isCut` allows.
+ *
+ * @param messages - The transcript. Not written to.
+ * @param search - The weights, the bounds and which indexes are legal cuts. See `CutSearch`.
+ * @returns `cut`, the length of `messages` when no legal cut was found, and on a plan made to a
+ * `target` the cost it leaves as `after`.
+ */
+const findCut = (messages, { from, cost, limit, keepRatio, target, summaryTokens, estimate, isCut, kept }) => {
+    let cut = messages.length;
+    let after;
+    if (target !== undefined && target > 0) {
+        // Forward from the head, so the first cut that reaches the target is the one that folds least.
+        let folded = 0;
+        for (let at = from; at < messages.length; at++) {
+            if (isCut(at)) {
+                cut = at;
+                after = cost - folded + summaryTokens;
+                if (after <= limit * target) {
+                    break;
+                }
+            }
+            if (at !== kept) {
+                folded += estimate(messages[at]);
+            }
+        }
+        return { cut, after };
+    }
+    const budget = limit * keepRatio;
+    let tail = 0;
+    for (let at = messages.length - 1; at > from; at--) {
+        tail += estimate(messages[at]);
+        if (tail > budget) {
+            break;
+        }
+        cut = at;
+    }
+    while (cut < messages.length && isCut(cut) === false) {
+        cut++;
+    }
+    return { cut };
+};
+/**
+ * The index of the last user message in a transcript.
+ *
+ * @param messages - The transcript.
+ * @returns The index, or `-1` when there is none.
+ */
+const lastUser = (messages) => messages.findLastIndex((message) => message.role === Role.User);
 /**
  * Where to fold a transcript that has grown into its window, or `undefined` when it should not be.
  *
@@ -159,6 +209,13 @@ const clamp = (index, most) => Math.min(Math.max(index, 0), most);
  * than summarised as if it were conversation. No plan comes back when the window is not full
  * enough, or when the only legal cut folds too little to pay for the summary.
  *
+ * A transcript that is one question and a long tool loop has no user message to cut at, so where
+ * no user cut folds anything the plan cuts inside the last question's loop instead: the question
+ * is kept whole as `question`, the assistant and tool messages after it are folded up to an
+ * assistant message, and the tail resumes there — every tool result in it still follows its call.
+ * `applyCompaction` sends the question after the summary. A transcript with a user cut is planned
+ * exactly as it was.
+ *
  * Both of those are recovered by reading the transcript, which is what a host whose array holds
  * everything it sends has to do. A host that keeps its fold as a record beside an append-only
  * transcript — its system prompt a separate argument, no summary message in the array at all —
@@ -184,47 +241,53 @@ export function planCompaction(messages, { limit, used, compactAt = COMPACT_AT,
     const head = systemHead(messages);
     const from = clamp(givenFrom ?? head.from, messages.length);
     const previous = givenPrevious ?? head.previous;
-    let cut = messages.length;
-    let after;
-    if (target !== undefined && target > 0) {
-        // Forward from the head, so the first cut that reaches the target is the one that folds least.
-        let folded = 0;
-        for (let at = from; at < messages.length; at++) {
-            if (messages[at].role === Role.User) {
-                cut = at;
-                after = cost - folded + summaryTokens;
-                if (after <= limit * target) {
-                    break;
-                }
-            }
-            folded += estimate(messages[at]);
-        }
+    const search = { from, cost, limit, keepRatio, target, summaryTokens, estimate };
+    const atUser = findCut(messages, { ...search, isCut: (at) => messages[at].role === Role.User });
+    const foldsAtUser = atUser.cut < messages.length && atUser.cut - from >= FEWEST_FOLDED;
+    if (foldsAtUser) {
+        return planFor(messages, from, atUser.cut, previous, atUser.after);
     }
-    else {
-        const budget = limit * keepRatio;
-        let kept = 0;
-        for (let at = messages.length - 1; at > from; at--) {
-            kept += estimate(messages[at]);
-            if (kept > budget) {
-                break;
-            }
-            cut = at;
-        }
-        while (cut < messages.length && messages[cut].role !== Role.User) {
-            cut++;
-        }
+    // A cut that folds too little but already reaches the target means there is nothing to fold.
+    const reachesTarget = target !== undefined && atUser.after !== undefined && atUser.after <= limit * target;
+    if (reachesTarget) {
+        return undefined;
     }
-    if (cut >= messages.length || cut - from < FEWEST_FOLDED) {
+    // No user message to cut at: fold inside the last question's tool loop, keeping the question.
+    const question = lastUser(messages);
+    if (question < 0) {
         return undefined;
     }
-    return {
-        from,
-        cut,
-        toSummarise: messages.slice(from, cut),
-        ...(previous ? { previous } : {}),
-        ...(after === undefined ? {} : { after: Math.round(after) }),
-    };
+    const inLoop = findCut(messages, {
+        ...search,
+        isCut: (at) => at > question && messages[at].role === Role.Assistant,
+        kept: question,
+    });
+    if (inLoop.cut >= messages.length) {
+        return undefined;
+    }
+    const folded = planFor(messages, from, inLoop.cut, previous, inLoop.after, question);
+    return folded.toSummarise.length >= FEWEST_FOLDED ? folded : undefined;
 }
+/**
+ * A plan for a cut `planCompaction` has chosen.
+ *
+ * @param messages - The transcript. Not written to.
+ * @param from - The first message folded.
+ * @param cut - The first message kept whole.
+ * @param [previous] - The summary this one continues.
+ * @param [after] - What the request is expected to cost once folded, on a plan made to a target.
+ * @param [question] - The user message kept whole ahead of `cut`, on a fold inside its tool loop.
+ * Left out of `toSummarise` where it lies between `from` and `cut`.
+ * @returns The plan, `toSummarise` a new array of the same message objects.
+ */
+const planFor = (messages, from, cut, previous, after, question) => ({
+    from,
+    cut,
+    ...(question === undefined ? {} : { question }),
+    toSummarise: messages.slice(from, cut).filter((_, at) => at + from !== question),
+    ...(previous ? { previous } : {}),
+    ...(after === undefined ? {} : { after: Math.round(after) }),
+});
 /**
  * What the summariser is handed for a plan: the earlier summary if there was one, then each
  * message as its role and at most 4000 characters of its text.
@@ -256,6 +319,22 @@ export function summaryInput(plan) {
  * resolves with an empty string where the model wrote nothing.
  */
 export const summariser = (config, model, { system = SUMMARY_PROMPT, maxTokens = SUMMARY_TOKENS, ...options } = {}) => (text) => ask(config, model, system, text, { maxTokens, ...options });
+/**
+ * What `beforeCompact` hooks are told is going: the plan's stretch as `turnMessages` reads it,
+ * less the question the fold keeps.
+ *
+ * @param sessionId - Prefixes every uuid.
+ * @param messages - The transcript the plan was made for.
+ * @param plan - What `planCompaction` returned for it.
+ * @param [offset] - What `messages[0]` is numbered as in the uuids.
+ * @returns The user and assistant text from `plan.from` to `plan.cut`, `plan.question` left out.
+ */
+const foldedMessages = (sessionId, messages, plan, offset = 0) => {
+    const { from, cut, question } = plan;
+    const read = (start, end) => turnMessages(sessionId, messages, start, end, { offset });
+    const splitsAtQuestion = question !== undefined && question >= from && question < cut;
+    return splitsAtQuestion ? [...read(from, question), ...read(question + 1, cut)] : read(from, cut);
+};
 /**
  * The hooks and the summariser for a plan, as a record to store rather than a transcript to send.
  *
@@ -275,7 +354,7 @@ export const summariser = (config, model, { system = SUMMARY_PROMPT, maxTokens =
 export async function runCompaction(messages, plan, summarise, { hooks, forced = false } = {}) {
     const context = hooks && {
         ...hooks.context,
-        compacting: turnMessages(hooks.context.session.id, messages, plan.from, plan.cut),
+        compacting: foldedMessages(hooks.context.session.id, messages, plan, hooks.offset),
         range: { from: plan.from, through: plan.cut },
     };
     let summary;
@@ -295,7 +374,12 @@ export async function runCompaction(messages, plan, summarise, { hooks, forced =
     if (!summary.trim()) {
         return undefined;
     }
-    return { summary: summary.trim(), through: plan.cut, at: new Date().toISOString() };
+    return {
+        summary: summary.trim(),
+        through: plan.cut,
+        ...(plan.question === undefined ? {} : { question: plan.question }),
+        at: new Date().toISOString(),
+    };
 }
 /**
  * The transcript as the server should see it: the folded head replaced by its summary.
@@ -308,7 +392,8 @@ export async function runCompaction(messages, plan, summarise, { hooks, forced =
  * Absent, the leading `system` messages are found by scanning, and zero of them is the ordinary
  * case for a host whose system prompt is a separate argument.
  * @returns `messages` itself when there is no record or its summary is blank, otherwise a new
- * array: the kept head, one summary message, then everything from the record's `through` on.
+ * array: the kept head, one summary message, the record's `question` when it has one and the head
+ * does not already hold it, then everything from the record's `through` on.
  *
  * @remarks
  * The inverse of storing a `CompactionRecord`, and the shape `planCompaction` expects to meet
@@ -321,12 +406,22 @@ export function applyCompaction(messages, record, { from } = {}) {
         return messages;
     }
     const head = clamp(from ?? systemHead(messages).from, record.through);
+    const question = keptQuestion(record, head);
     return [
         ...messages.slice(0, head).filter((message) => !isSummary(message)),
         summaryMessage(record.summary),
+        ...(question === undefined ? [] : [messages[question]]),
         ...messages.slice(record.through),
     ];
 }
+/**
+ * The question a fold sends after its summary, unless the kept head already holds it.
+ *
+ * @param record - The fold.
+ * @param head - How many messages are kept ahead of the summary.
+ * @returns `record.question` where it lies from `head` up to `through`, otherwise `undefined`.
+ */
+const keptQuestion = ({ question, through }, head) => question !== undefined && question >= head && question < through ? question : undefined;
 /**
  * Where a stored index sits in the request `applyCompaction` builds, once a fold has shifted
  * everything after it.
@@ -338,7 +433,8 @@ export function applyCompaction(messages, record, { from } = {}) {
  * prompts, when the host keeps them in the array. Zero is the stored-fold case,
  * where the summary is the request's first message.
  * @returns The index in the request. One before the record's `through` gives `head`, where the
- * summary sits — an index among the kept head messages themselves included.
+ * summary sits — an index among the kept head messages themselves included — except the record's
+ * `question`, which is sent whole right after the summary.
  *
  * @remarks
  * A transcript that stays append-only and a request that does not are two numberings of the same
@@ -350,7 +446,12 @@ export const requestIndex = (index, record, head = 0) => {
     if (!record) {
         return index;
     }
-    return index < record.through ? head : index - record.through + head + 1;
+    const question = keptQuestion(record, head);
+    if (index === question) {
+        return head + 1;
+    }
+    const afterSummary = question === undefined ? 1 : 2;
+    return index < record.through ? head : index - record.through + head + afterSummary;
 };
 /**
  * The transcript with the plan's stretch replaced by one system message holding its summary.

package/dist/index.d.ts

@@ -21,7 +21,7 @@ export { assembleContext, configureHooks, consult, type Gathered, gather, HOOK_C
 export { type AgentLoopHooks, type AgentLoopOptions, type AgentLoopRequest, type AgentLoopResult, PRESELECT_APPEND, PRESELECT_EXCLUSIVE, PRESELECTED_NOTICE, runAgentLoop, type StepWindow, } from './run/agent-loop.ts';
 export { configureEvents, type EventBusOptions, emit, endRun, fold, history, type RunEvent, type RunEventInput, RunEventKind, type RunMetrics, type RunMetricsOptions, RunOutcome, type RunUsage, resetEvents, runMetrics, type TurnReport, watch, } from './run/events.ts';
 export { buildBody } from './run/request-body.ts';
-export { preview } from './run/run-calls.ts';
+export { preview, type ToolDispatch } from './run/run-calls.ts';
 export { resetAll } from './runtime/reset.ts';
 export { createRuntime, defaultRuntime, type Runtime, type RuntimeOptions } from './runtime/runtime.ts';
 export { CAPABILITY_SNAPSHOT_VERSION, type CapabilitySnapshot, type EndpointSnapshot, exportCapabilities, importCapabilities, type ModelSnapshot, } from './runtime/snapshot.ts';
@@ -37,5 +37,5 @@ export { backoffMs, ContextOverflow, EndpointSilent, isModelLoading, isOverflow,
 export { CacheBreakReason, type Produced, type StreamTurnOptions, streamTurn, type Turn, type TurnUsage, } from './wire/stream.ts';
 export { ALL_FENCES, DEFAULT_FENCES, type Fence, FenceSplitter, type FenceSplitterOptions, SPLIT_OUTPUT, SPLIT_REASONING, type Split, stripThinking, THINK_FENCE, } from './wire/thinking.ts';
 export { type ContextBreakdown, type ContextBreakdownOptions, compact, contextChars, contextTokens, estimateTokens, messageTokens, requestChars, requestTokens, type TokenEstimateOptions, } from './wire/tokens.ts';
-export { ARGUMENTS_MALFORMED, ARGUMENTS_TRUNCATED, parseToolArguments, recoverToolCalls, ToolArgumentsError, type ToolCall, type ToolCallOutcome, type ToolCallRequest, type ToolCallResult, } from './wire/tool-calls.ts';
+export { ARGUMENTS_MALFORMED, ARGUMENTS_TRUNCATED, parseToolArguments, type ResourceLink, type RunArtifact, recoverToolCalls, type ToolAnswer, ToolArgumentsError, type ToolCall, type ToolCallOutcome, type ToolCallRequest, type ToolCallResult, } from './wire/tool-calls.ts';
 export { FinishReason, FUNCTION_TOOL, JSON_SCHEMA_FORMAT, PartType, Role, SchemaType } from './wire/wire.ts';

package/dist/run/agent-loop.d.ts

@@ -9,7 +9,7 @@ import { type KeywordPreselectOptions } from '../tools/preselect.ts';
 import { type ToolOrder } from '../tools/tool-loading.ts';
 import { type SideTask } from '../turn/side-task.ts';
 import { type Turn, type TurnUsage } from '../wire/stream.ts';
-import { type ToolCallOutcome, type ToolCallRequest, type ToolCallResult } from '../wire/tool-calls.ts';
+import { type RunArtifact, type ToolCallOutcome, type ToolCallRequest, type ToolCallResult } from '../wire/tool-calls.ts';
 import { type RunEventInput, type RunMetrics } from './events.ts';
 import { type ToolDispatch } from './run-calls.ts';
 /**
@@ -70,6 +70,14 @@ export interface AgentLoopHooks {
     preface?: string;
     /** Hears each note, from before the request and from `afterTurn`. */
     onNote?: OnNote;
+    /**
+     * What the run's first message is numbered as in the uuids hooks are told — the host's stored
+     * index of `messages[0]`, for a host that numbers messages across steps or runs. Read for
+     * `beforeCompact`'s `compacting` and `afterTurn`'s `turn`. See `turnMessages`.
+     *
+     * @defaultValue `0`
+     */
+    offset?: number;
 }
 /** How full the window is as a step is about to be sent, as `beforeStep` is told it. */
 export interface StepWindow {
@@ -212,7 +220,14 @@ export interface AgentLoopOptions {
      * mode, where a tool is loaded only while its definition is in the history.
      */
     loaded?: Iterable<string>;
-    /** Runs one tool call and returns what the model reads. What it throws, the model reads too. */
+    /**
+     * Runs one tool call and returns what the model reads. What it throws, the model reads too.
+     *
+     * @remarks
+     * A `ToolAnswer` in place of the string adds what the call produced: its `text` is the whole of
+     * what goes into the transcript, and its `links` become the result's `artifacts`, an `artifact`
+     * event the first time each uri is seen, and `links` on what `onToolResult` is handed.
+     */
     dispatch: ToolDispatch;
     /**
      * Runs a step's calls together rather than one after another. Results still go into the
@@ -346,6 +361,12 @@ export interface AgentLoopResult {
      * proxied preselection's among them, first.
      */
     toolCalls: ToolCallOutcome[];
+    /**
+     * What the run's calls linked to, one per uri in the order first seen, each with the call and
+     * step that first linked to it and how many calls did. Empty where `dispatch` answered with
+     * strings only.
+     */
+    artifacts: RunArtifact[];
     /**
      * What is loaded at the end, for `carryOver`. Empty in eager mode, and in proxied mode: a
      * definition loaded there is already in the history, and one a compaction folded away has to be

package/dist/run/agent-loop.js

@@ -116,6 +116,7 @@ export async function runAgentLoop(options) {
             messages: [...options.messages],
             usage: noUsage(),
             toolCalls: [],
+            artifacts: [],
             loaded: [],
             used: [],
             preselected: [],
@@ -235,10 +236,10 @@ async function foldBeforeStep(transcript, reported, summarise, { options, charsP
         return transcript;
     }
     const folded = await compactTranscript(transcript, plan, summarise, {
-        ...(hooks ? { hooks: { run: hooks.run, context: hooks.context, onNote: hooks.onNote } } : {}),
+        ...(hooks ? { hooks: { run: hooks.run, context: hooks.context, onNote: hooks.onNote, offset: hooks.offset } } : {}),
     });
     if (folded !== transcript) {
-        notice(`compacted ${plan.cut - plan.from} messages into a summary`);
+        notice(`compacted ${plan.toSummarise.length} messages into a summary`);
     }
     return folded;
 }
@@ -260,7 +261,7 @@ const tellAfterTurn = (hooks, messages, at, reply) => notify(hooks.run, HookEven
         ? {
             turn: {
                 index: turnIndex(messages, at),
-                messages: turnMessages(hooks.context.session.id, messages, at),
+                messages: turnMessages(hooks.context.session.id, messages, at, undefined, { offset: hooks.offset }),
             },
         }
         : {}),
@@ -450,10 +451,12 @@ async function runSteps(options, standing) {
     let messages = [...options.messages];
     let usage = noUsage();
     const toolCalls = [];
+    const artifacts = new Map();
     standing.read = () => ({
         messages,
         usage,
         toolCalls,
+        artifacts: [...artifacts.values()],
         // As the result would have said it: proxied, nothing is loaded that the history does not hold.
         loaded: proxied ? [] : [...loaded],
         used: [...used],
@@ -498,6 +501,7 @@ async function runSteps(options, standing) {
         loads,
         used,
         toolCalls,
+        artifacts,
         dedupable,
         dispatch,
         parallel,
@@ -645,6 +649,7 @@ async function runSteps(options, standing) {
                 messages,
                 usage,
                 toolCalls,
+                artifacts: [...artifacts.values()],
                 loaded: proxied ? [] : [...loaded],
                 used: [...used],
                 preselected: [...preselected],

package/dist/run/events.d.ts

@@ -1,6 +1,7 @@
 import type { TokenLedger } from '../context/ledger.ts';
 import type { TurnUsage } from '../wire/stream.ts';
 import { CacheBreakReason } from '../wire/stream.ts';
+import type { ResourceLink } from '../wire/tool-calls.ts';
 /**
  * What a run is doing, while it is doing it.
  *
@@ -89,6 +90,12 @@ export declare const RunEventKind: {
     readonly ToolResult: 'tool-result';
     /** Something the runner did that is not the model's doing — a preselection, a retry. */
     readonly Notice: 'notice';
+    /**
+     * A tool call's answer linked to something the run had not produced before. `id` is the call's,
+     * `name` the tool, `text` the link's title or else its uri, and `link` the link itself. Once
+     * per uri per run: a later call that links the same uri again only adds to its count.
+     */
+    readonly Artifact: 'artifact';
     /**
      * What the run has cost so far, at the end of every turn, and what that one turn did. Sent
      * whether or not the endpoint reported tokens, since the timings and the cache comparison are
@@ -136,8 +143,8 @@ export interface TurnReport extends TurnUsage {
  * One thing that happened in a run, as a watcher receives it.
  *
  * @remarks
- * Every field but `id` is always present — the empty ones are `""` or `null` rather than
- * missing — so a client reads it without guarding each key.
+ * Every field but `id` and `link` is always present — the empty ones are `""` or `null` rather
+ * than missing — so a client reads it without guarding each key.
  */
 export interface RunEvent {
     /** The run it belongs to. `emit` fills this in; a caller does not pass it. */
@@ -175,6 +182,11 @@ export interface RunEvent {
     ok: boolean | null;
     /** Running totals on `usage`, otherwise null. */
     usage: RunUsage | null;
+    /**
+     * What the call produced, on `artifact` and on no other kind. Optional for the reason `id` is:
+     * an event built before it existed still compiles.
+     */
+    link?: ResourceLink;
 }
 /**
  * What `emit` is given: the run and the sequence are the bus's to assign.
@@ -300,6 +312,8 @@ export interface RunMetrics {
     toolErrors: Record<string, number>;
     /** `load_tools` calls. */
     loadCalls: number;
+    /** `artifact` events: the distinct things the run's tool calls linked to. */
+    artifacts: number;
     /**
      * Tools `load_tools` loaded that were not loaded already. Filled by `runAgentLoop`, which sees
      * the resolution; the events carry only its text.

package/dist/run/events.js

@@ -56,6 +56,12 @@ export const RunEventKind = {
     ToolResult: 'tool-result',
     /** Something the runner did that is not the model's doing — a preselection, a retry. */
     Notice: 'notice',
+    /**
+     * A tool call's answer linked to something the run had not produced before. `id` is the call's,
+     * `name` the tool, `text` the link's title or else its uri, and `link` the link itself. Once
+     * per uri per run: a later call that links the same uri again only adds to its count.
+     */
+    Artifact: 'artifact',
     /**
      * What the run has cost so far, at the end of every turn, and what that one turn did. Sent
      * whether or not the endpoint reported tokens, since the timings and the cache comparison are
@@ -500,6 +506,7 @@ export function runMetrics(events, { contextLength } = {}) {
         toolCalls: 0,
         toolErrors: {},
         loadCalls: 0,
+        artifacts: 0,
         promptTokens: 0,
         completionTokens: 0,
         cachedTokens: 0,
@@ -540,6 +547,9 @@ export function runMetrics(events, { contextLength } = {}) {
                 add('toolMs', event.at - called);
             }
         }
+        else if (event.kind === RunEventKind.Artifact) {
+            metrics.artifacts++;
+        }
         else if (event.kind === RunEventKind.Usage && event.usage?.turn) {
             const turn = event.usage.turn;
             last = turn;

package/dist/run/run-calls.d.ts

@@ -1,7 +1,7 @@
 import type OpenAI from 'openai';
 import type { CatalogServer } from '../tools/catalog.ts';
 import type { Turn } from '../wire/stream.ts';
-import { type ToolCall, type ToolCallOutcome, type ToolCallRequest, type ToolCallResult } from '../wire/tool-calls.ts';
+import { type RunArtifact, type ToolAnswer, type ToolCall, type ToolCallOutcome, type ToolCallRequest, type ToolCallResult } from '../wire/tool-calls.ts';
 import type { RunEventInput } from './events.ts';
 /**
  * The tools between two turns: a step's calls read, run and written into the transcript.
@@ -63,8 +63,11 @@ export declare const assistantMessage: (content: string, parsed: ReadCall[]) =>
  * @returns A new array of the definitions themselves, not copies.
  */
 export declare const definedAs: (definitions: ReadonlyMap<string, OpenAI.ChatCompletionTool>, names: Iterable<string>) => (OpenAI.ChatCompletionCustomTool | OpenAI.ChatCompletionFunctionTool)[];
-/** Runs one tool call and resolves to what the model reads; what it throws, the model reads too. */
-export type ToolDispatch = (call: ToolCallRequest, signal?: AbortSignal) => Promise<string>;
+/**
+ * Runs one tool call and resolves to what the model reads, alone or with what the call produced;
+ * what it throws, the model reads too.
+ */
+export type ToolDispatch = (call: ToolCallRequest, signal?: AbortSignal) => Promise<string | ToolAnswer>;
 /**
  * What running a step's calls reads of the run and writes back into it. One per run, shared by
  * every step: the sets and counts here are the run's own, not copies.
@@ -86,6 +89,8 @@ export interface Calling {
     };
     used: Set<string>;
     toolCalls: ToolCallOutcome[];
+    /** What the run's calls linked to, by uri, in the order first seen. */
+    artifacts: Map<string, RunArtifact>;
     dedupable: (call: ToolCallRequest) => boolean;
     dispatch: ToolDispatch;
     parallel: boolean;

package/dist/run/run-calls.js

@@ -84,6 +84,13 @@ export const assistantMessage = (content, parsed) => ({
  * @returns A new array of the definitions themselves, not copies.
  */
 export const definedAs = (definitions, names) => [...names].flatMap((name) => definitions.get(name) ?? []);
+/**
+ * A dispatcher's answer in the one shape, whichever it gave.
+ *
+ * @param answer - What `dispatch` resolved to.
+ * @returns The text and the links, a string answer's being none.
+ */
+const answerOf = (answer) => typeof answer === 'string' ? { text: answer, links: [] } : { text: answer.text, links: answer.links ?? [] };
 /**
  * What a call is shown, counted and dispatched as: a `call_tool` as the tool it names.
  *
@@ -129,6 +136,7 @@ function answerLoad(run, args) {
  * Makes a call at most once per key, sharing the in-flight promise so two identical calls in one
  * step make one request between them whether they run together or one after the other.
  *
+ * @typeParam T - What a call answers with.
  * @param answered - The step's calls so far, by key. Written to.
  * @param key - What makes two calls the same one.
  * @param make - Makes the call. Not called where the key has an answer, or one on its way.
@@ -156,6 +164,30 @@ async function once(answered, key, make) {
         throw error;
     }
 }
+/**
+ * Adds one call's links to the run's artifacts, telling watchers of each uri the first time.
+ *
+ * @param run - The run. Written to: a new uri is added to `artifacts`, and a known one counted.
+ * @param links - What the call's answer linked to.
+ * @param from - The call they came from.
+ */
+function collect(run, links, from) {
+    // Per call, so a tool that lists one uri twice in an answer is counted as making it once.
+    const seen = new Set();
+    for (const link of links) {
+        if (seen.has(link.uri)) {
+            continue;
+        }
+        seen.add(link.uri);
+        const known = run.artifacts.get(link.uri);
+        if (known) {
+            known.count++;
+            continue;
+        }
+        run.artifacts.set(link.uri, { link, ...from, count: 1 });
+        run.onEvent({ kind: RunEventKind.Artifact, id: from.callId, name: from.tool, text: link.title || link.uri, link });
+    }
+}
 /**
  * Runs one call to its result, telling the host of both. A tool that throws is that call's answer
  * rather than the run's end, unless the run was stopped.
@@ -163,10 +195,12 @@ async function once(answered, key, make) {
  * @param run - The run the call belongs to.
  * @param entry - The call, with its arguments read.
  * @param answered - The step's calls already made, by tool and arguments, for `dedupeToolCalls`.
- * @returns The call's id, the name it was shown under, whether it worked, and the whole of what
- * the model reads. Rejects where the run was stopped or a host callback threw.
+ * @param step - Which step this is, for the artifacts the call's links become.
+ * @returns The call's id, the name it was shown under, whether it worked, the whole of what the
+ * model reads, and the links its answer carried, where it carried any. Rejects where the run was
+ * stopped or a host callback threw.
  */
-async function runCall(run, entry, answered) {
+async function runCall(run, entry, answered, step) {
     const { catalog, onDemand, proxied, signal } = run;
     const { call, args, error: unreadable, normal } = entry;
     const proxy = run.proxies && call.function.name === CALL_TOOL;
@@ -191,7 +225,10 @@ async function runCall(run, entry, answered) {
     run.onEvent({ kind: RunEventKind.ToolCall, id: call.id, name, text: preview(raw) });
     run.onToolCall?.(request);
     let content;
+    let links = [];
     let ok = true;
+    // Answered from an earlier identical call of the step, whose links are already counted.
+    let shared = false;
     try {
         if (!args) {
             throw unreadable;
@@ -214,9 +251,13 @@ async function runCall(run, entry, answered) {
             run.used.add(name);
             // Keyed on the inner call, so one made through `call_tool` and one made natively share.
             const key = `${name}\0${proxy ? JSON.stringify(inner) : normal}`;
-            content = run.dedupable(request)
+            const dedupable = run.dedupable(request);
+            // Read before `once` looks, which it does before it awaits anything.
+            shared = dedupable && answered.has(key);
+            const answer = dedupable
                 ? await once(answered, key, () => run.dispatch(request, signal))
                 : await run.dispatch(request, signal);
+            ({ text: content, links } = answerOf(answer));
         }
     }
     catch (error) {
@@ -227,7 +268,11 @@ async function runCall(run, entry, answered) {
         ok = false;
     }
     run.onEvent({ kind: RunEventKind.ToolResult, id: call.id, name, ok, text: preview(content) });
-    const result = { id: call.id, name, ok, content };
+    const isFirstAnswer = shared === false;
+    if (isFirstAnswer) {
+        collect(run, links, { callId: call.id, tool: name, step });
+    }
+    const result = { id: call.id, name, ok, content, ...(links.length ? { links } : {}) };
     run.onToolResult?.(result);
     return result;
 }
@@ -270,7 +315,7 @@ export async function runCalls(run, parsed, messages, step) {
             signal?.throwIfAborted();
             begun = parsed.length;
             await Promise.all(parsed.map(async (call, at) => {
-                outcomes[at] = await runCall(run, call, answered);
+                outcomes[at] = await runCall(run, call, answered, step);
             }));
             for (const { id, name, ok, content } of outcomes.flatMap((outcome) => outcome ?? [])) {
                 toolCalls.push({ id, name, ok });
@@ -281,7 +326,7 @@ export async function runCalls(run, parsed, messages, step) {
             for (const [at, call] of parsed.entries()) {
                 signal?.throwIfAborted();
                 begun = at + 1;
-                const { id, name, ok, content } = await runCall(run, call, answered);
+                const { id, name, ok, content } = await runCall(run, call, answered, step);
                 toolCalls.push({ id, name, ok });
                 await keep(id, content);
             }

package/dist/wire/errors.d.ts

@@ -1,7 +1,7 @@
 import type OpenAI from 'openai';
 import { ContextOverflow } from './retry.ts';
 import type { TurnUsage } from './stream.ts';
-import type { ToolCallOutcome } from './tool-calls.ts';
+import type { RunArtifact, ToolCallOutcome } from './tool-calls.ts';
 /**
  * What went wrong, as a string.
  *
@@ -34,6 +34,12 @@ export interface AgentLoopFailure {
     usage: TurnUsage;
     /** Every call that was made, in order. One the run was stopped during is `ok: false`. */
     toolCalls: ToolCallOutcome[];
+    /**
+     * What the calls that came back linked to, as the result's `artifacts` would have said it: a
+     * failed run's writes were still made. `runAgentLoop` always sets it; optional so a failure
+     * built before it existed still compiles.
+     */
+    artifacts?: RunArtifact[];
     /** What was loaded when the run failed, for `carryOver`. Empty in eager and proxied modes. */
     loaded: string[];
     /** The tools the model had called, `load_tools` excluded. */
@@ -55,6 +61,7 @@ export declare class AgentLoopError extends Error implements AgentLoopFailure {
     readonly messages: OpenAI.ChatCompletionMessageParam[];
     readonly usage: TurnUsage;
     readonly toolCalls: ToolCallOutcome[];
+    readonly artifacts: RunArtifact[];
     readonly loaded: string[];
     readonly used: string[];
     readonly preselected: string[];
@@ -88,6 +95,7 @@ export declare class AgentLoopOverflow extends ContextOverflow implements AgentL
     readonly messages: OpenAI.ChatCompletionMessageParam[];
     readonly usage: TurnUsage;
     readonly toolCalls: ToolCallOutcome[];
+    readonly artifacts: RunArtifact[];
     readonly loaded: string[];
     readonly used: string[];
     readonly preselected: string[];

package/dist/wire/errors.js

@@ -26,6 +26,7 @@ export class AgentLoopError extends Error {
     messages;
     usage;
     toolCalls;
+    artifacts;
     loaded;
     used;
     preselected;
@@ -39,6 +40,7 @@ export class AgentLoopError extends Error {
         this.messages = run.messages;
         this.usage = run.usage;
         this.toolCalls = run.toolCalls;
+        this.artifacts = run.artifacts ?? [];
         this.loaded = run.loaded;
         this.used = run.used;
         this.preselected = run.preselected;
@@ -67,6 +69,7 @@ export class AgentLoopOverflow extends ContextOverflow {
     messages;
     usage;
     toolCalls;
+    artifacts;
     loaded;
     used;
     preselected;
@@ -80,6 +83,7 @@ export class AgentLoopOverflow extends ContextOverflow {
         this.messages = run.messages;
         this.usage = run.usage;
         this.toolCalls = run.toolCalls;
+        this.artifacts = run.artifacts ?? [];
         this.loaded = run.loaded;
         this.used = run.used;
         this.preselected = run.preselected;

package/dist/wire/tool-calls.d.ts

@@ -146,4 +146,68 @@ export interface ToolCallResult extends ToolCallOutcome {
      * `load_tools` answer, or why the arguments could not be read.
      */
     content: string;
+    /**
+     * What the dispatcher's answer said the call produced, as it said it. Absent where it said
+     * nothing, and always for a call that threw. A deduplicated repeat carries the first call's.
+     */
+    links?: ResourceLink[];
+}
+/**
+ * A handle to something a tool call made or found — a document it wrote, a file it saved.
+ *
+ * @remarks
+ * MCP's `ResourceLink` less its `type`, declared structurally rather than imported, the way
+ * agent-mcp-pool declares `ToolDefinition` against `ChatCompletionTool`: the pool's own
+ * `ResourceLink` has these field names exactly, so its answer is assignable here and neither
+ * package depends on the other.
+ */
+export interface ResourceLink {
+    uri: string;
+    name: string;
+    title?: string;
+    mimeType?: string;
+    description?: string;
+    /** Who the link is for, how much it matters, and when it last changed — the server's word. */
+    annotations?: {
+        audience?: ('user' | 'assistant')[];
+        priority?: number;
+        lastModified?: string;
+    };
+    _meta?: Record<string, unknown>;
+}
+/**
+ * A dispatcher's answer when it has more to say than the text: what the model reads, and what the
+ * call produced for the host.
+ *
+ * @remarks
+ * A link meant for the model alone — one whose MCP `audience` leaves out `"user"` — is the
+ * dispatcher's to drop, since that is MCP's rule and this package knows nothing of MCP.
+ */
+export interface ToolAnswer {
+    /** What goes into the transcript, exactly as a string answer does. */
+    text: string;
+    /** What the call produced, in the order the tool gave them. Never sent to the model. */
+    links?: ResourceLink[];
+}
+/**
+ * One thing a run's tool calls produced: the first link to a uri, and where it came from.
+ *
+ * @remarks
+ * A link is what one call answered with; an artifact is what the run made, one per uri however
+ * many calls linked to it.
+ */
+export interface RunArtifact {
+    /** The link as the first call to give its uri gave it. A later call's says nothing new here. */
+    link: ResourceLink;
+    /** The call that first linked to it. */
+    callId: string;
+    /** That call's tool. For a `call_tool`, the tool it named. */
+    tool: string;
+    /** The loop's step that call was made in, from 0. */
+    step: number;
+    /**
+     * How many calls linked to it. A deduplicated repeat does not count, since nothing was made
+     * twice, and a call that links one uri twice counts once.
+     */
+    count: number;
 }

package/llms.txt

@@ -185,6 +185,7 @@ What a run is doing, while it is doing it.
 The tools between two turns: a step's calls read, run and written into the transcript.
 
 - `preview` — A long tool argument or result cut to what a watcher needs, with the full length said.
+- `ToolDispatch` (type) — Runs one tool call and resolves to what the model reads, alone or with what the call produced; what it throws, the model reads too.
 
 ### reset
 
@@ -384,7 +385,10 @@ Reading what a model meant by a tool call when it did not write one cleanly.
 - `ARGUMENTS_MALFORMED` — The `ToolArgumentsError` kind for a call the model wrote wrongly.
 - `ARGUMENTS_TRUNCATED` — The `ToolArgumentsError` kind for a call cut off at the reply ceiling.
 - `parseToolArguments` — A tool call's arguments as the object the tool is handed.
+- `ResourceLink` (type) — A handle to something a tool call made or found — a document it wrote, a file it saved.
+- `RunArtifact` (type) — One thing a run's tool calls produced: the first link to a uri, and where it came from.
 - `recoverToolCalls` — Tool calls a model wrote into its reply as text, taken out of it and made into calls.
+- `ToolAnswer` (type) — A dispatcher's answer when it has more to say than the text: what the model reads, and what the call produced for the host.
 - `ToolArgumentsError` — Tool arguments that could not be read as an object, with why.
 - `ToolCall` (type) — A tool call as the loop handles it, recovered or streamed.
 - `ToolCallOutcome` (type) — What one tool call did, in the order the model asked.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@cubicecho/agent-core",
-  "version": "4.4.0",
+  "version": "4.6.0",
   "description": "The endpoint-agnostic half of an OpenAI-compatible agent loop: tool-schema compatibility, on-demand tool loading, one-shot side tasks, run events, and a pooled client.",
   "keywords": [
     "openai",