mirror of
https://github.com/rowboatlabs/rowboat.git
synced 2026-07-12 21:02:17 +02:00
feat(x): reference-based model requests + wire-form composer
model_call_requested carried three duplications that dominated turn-file
size (measured on a real 647KB two-step turn): the system prompt (28KB)
and the 38-tool snapshot (40KB) repeated per call despite being byte-
identical to turn_created.agent.resolved, and messages re-inlining the
whole current-turn transcript (230KB re-stating a 207KB tool result).
ModelRequest is now a list of references into the turn's own events —
call 0: [{context}?, {input}]; call N: [{assistant: N-1}, ...that
batch's toolResults in source order]; re-issue after an interruption:
[]. Every referenced byte exists exactly once in the file; the measured
turn drops to ~285KB and each further model call costs ~200 bytes. The
reducer's ordering invariant got stricter: reference lists are matched
exactly against the transcript.
Debuggability is now byte-for-byte at the wire level: ResolvedModel
gains encodeMessages (the structural->wire conversion — user-message
context weaving, attachment rendering, tool-result enveloping, i.e.
convertFromMessages), and composeModelRequest rebuilds the exact
provider payload (resolved system prompt + wrapped tools + materialized
prefix + resolved refs, encoded). The loop transmits exactly the
composer's output, so the durable file plus the composer reproduce what
the model received — pinned by a property test asserting composed ==
sent for every call, and a 2KB size guard on request events. A bridge
test demonstrates the woven wire form (no raw userMessageContext).
Breaking for existing dev turn files (same schemaVersion, pre-release):
wipe ~/.rowboat/storage.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Ramnique Singh
parent
a58493b545
commit
3822bf53da
18 changed files with 524 additions and 227 deletions
|
|
@ -118,9 +118,7 @@ function requested(
|
|||
ts: TS,
|
||||
modelCallIndex: index,
|
||||
request: {
|
||||
systemPrompt: "SYS",
|
||||
messages,
|
||||
tools: [echoTool, fetchTool],
|
||||
parameters: {},
|
||||
...requestOverrides,
|
||||
},
|
||||
|
|
@ -348,13 +346,13 @@ function syncToolSequence(): TEvent[] {
|
|||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
return [
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
permResolved("tc1", "allow"),
|
||||
invocation("tc1"),
|
||||
result("tc1", "echo"),
|
||||
requested(1, [user("hello"), call0, toolMsg("tc1", "echo")]),
|
||||
requested(1, [{ kind: "assistant", modelCallIndex: 0 }, { kind: "toolResult", toolCallId: "tc1" }]),
|
||||
completed(1, assistantText("done")),
|
||||
turnCompletedEv(),
|
||||
];
|
||||
|
|
@ -395,7 +393,7 @@ describe("plain completion", () => {
|
|||
it("reduces a plain model response to a completed turn", () => {
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
stepEvent(0),
|
||||
completed(0, assistantText("done")),
|
||||
turnCompletedEv(),
|
||||
|
|
@ -420,7 +418,7 @@ describe("plain completion", () => {
|
|||
it("leaves usage fields undefined when never reported", () => {
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, assistantText("a"), { inputTokens: 5 }),
|
||||
]);
|
||||
expect(state.usage).toEqual({ inputTokens: 5 });
|
||||
|
|
@ -436,7 +434,7 @@ describe("tool execution", () => {
|
|||
);
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
]);
|
||||
expect(state.toolCalls).toHaveLength(2);
|
||||
|
|
@ -459,7 +457,7 @@ describe("tool execution", () => {
|
|||
it("leaves identity undefined for tools missing from the agent snapshot", () => {
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, assistantCalls(toolCallPart("x1", "unknown-tool"))),
|
||||
result("x1", "unknown-tool", "runtime", "No such tool", true),
|
||||
]);
|
||||
|
|
@ -482,7 +480,7 @@ describe("tool execution", () => {
|
|||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
const state = reduceTurn([
|
||||
created({ config: { autoPermission: true, humanAvailable: true, maxModelCalls: 20 } }),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
permClassified("tc1", "allow"),
|
||||
|
|
@ -499,7 +497,7 @@ describe("tool execution", () => {
|
|||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
permClassificationFailed(["tc1"]),
|
||||
|
|
@ -514,7 +512,7 @@ describe("tool execution", () => {
|
|||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
progress("tc1"),
|
||||
|
|
@ -528,7 +526,7 @@ describe("tool execution", () => {
|
|||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
permResolved("tc1", "deny"),
|
||||
|
|
@ -543,7 +541,7 @@ describe("context references", () => {
|
|||
it("accepts a referenced context with matching contextRef on requests", () => {
|
||||
const state = reduceTurn([
|
||||
created({ context: { previousTurnId: PREV_TURN_ID } }),
|
||||
requested(0, [user("hello")], {
|
||||
requested(0, [{ kind: "input" }], {
|
||||
contextRef: { previousTurnId: PREV_TURN_ID },
|
||||
}),
|
||||
completed(0, assistantText("done")),
|
||||
|
|
@ -556,7 +554,7 @@ describe("context references", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created({ context: { previousTurnId: PREV_TURN_ID } }),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
],
|
||||
/contextRef inconsistent/,
|
||||
);
|
||||
|
|
@ -566,7 +564,7 @@ describe("context references", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created({ context: { previousTurnId: PREV_TURN_ID } }),
|
||||
requested(0, [user("hello")], {
|
||||
requested(0, [{ kind: "input" }], {
|
||||
contextRef: { previousTurnId: "some-other-turn" },
|
||||
}),
|
||||
],
|
||||
|
|
@ -578,7 +576,7 @@ describe("context references", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")], {
|
||||
requested(0, [{ kind: "input" }], {
|
||||
contextRef: { previousTurnId: PREV_TURN_ID },
|
||||
}),
|
||||
],
|
||||
|
|
@ -586,7 +584,7 @@ describe("context references", () => {
|
|||
);
|
||||
});
|
||||
|
||||
it("ignores tool messages inside an inline context prefix when checking order", () => {
|
||||
it("requires a {context} ref before {input} when inline context is nonempty", () => {
|
||||
const context = [
|
||||
user("earlier"),
|
||||
assistantCalls(toolCallPart("old1", "echo")),
|
||||
|
|
@ -594,11 +592,38 @@ describe("context references", () => {
|
|||
];
|
||||
const state = reduceTurn([
|
||||
created({ context }),
|
||||
requested(0, [...context, user("hello")]),
|
||||
requested(0, [{ kind: "context" }, { kind: "input" }]),
|
||||
completed(0, assistantText("done")),
|
||||
turnCompletedEv(),
|
||||
]);
|
||||
expect(deriveTurnStatus(state)).toBe("completed");
|
||||
expectCorruption(
|
||||
[created({ context }), requested(0, [{ kind: "input" }])],
|
||||
/references do not match/,
|
||||
);
|
||||
});
|
||||
|
||||
it("rejects a contextRef on a non-initial model call", () => {
|
||||
expectCorruption(
|
||||
[
|
||||
created({ context: { previousTurnId: PREV_TURN_ID } }),
|
||||
requested(0, [{ kind: "input" }], {
|
||||
contextRef: { previousTurnId: PREV_TURN_ID },
|
||||
}),
|
||||
completed(0, assistantCalls(toolCallPart("tc1", "echo"))),
|
||||
invocation("tc1"),
|
||||
result("tc1", "echo"),
|
||||
requested(
|
||||
1,
|
||||
[
|
||||
{ kind: "assistant", modelCallIndex: 0 },
|
||||
{ kind: "toolResult", toolCallId: "tc1" },
|
||||
],
|
||||
{ contextRef: { previousTurnId: PREV_TURN_ID } },
|
||||
),
|
||||
],
|
||||
/contextRef on a non-initial model call/,
|
||||
);
|
||||
});
|
||||
});
|
||||
|
||||
|
|
@ -610,7 +635,7 @@ describe("suspension", () => {
|
|||
);
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
invocation("a1", fetchTool),
|
||||
|
|
@ -628,7 +653,7 @@ describe("suspension", () => {
|
|||
);
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
invocation("a1", fetchTool),
|
||||
|
|
@ -648,7 +673,7 @@ describe("suspension", () => {
|
|||
);
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("a1", fetchTool),
|
||||
invocation("a2", fetchTool),
|
||||
|
|
@ -657,10 +682,9 @@ describe("suspension", () => {
|
|||
suspendedEv([], [{ id: "a1", tool: fetchTool }]),
|
||||
result("a1", "fetch", "async", "first"),
|
||||
requested(1, [
|
||||
user("hello"),
|
||||
call0,
|
||||
toolMsg("a1", "fetch", "first"),
|
||||
toolMsg("a2", "fetch", "second"),
|
||||
{ kind: "assistant", modelCallIndex: 0 },
|
||||
{ kind: "toolResult", toolCallId: "a1" },
|
||||
{ kind: "toolResult", toolCallId: "a2" },
|
||||
]),
|
||||
completed(1, assistantText("done")),
|
||||
turnCompletedEv(),
|
||||
|
|
@ -673,7 +697,7 @@ describe("suspension", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
permResolved("tc1", "allow"),
|
||||
|
|
@ -693,7 +717,7 @@ describe("suspension", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
invocation("a1", fetchTool),
|
||||
|
|
@ -707,7 +731,7 @@ describe("suspension", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
suspendedEv([{ id: "tc1", name: "echo" }], []),
|
||||
],
|
||||
/suspension while a model call is unsettled/,
|
||||
|
|
@ -719,9 +743,9 @@ describe("recovery-shaped histories", () => {
|
|||
it("accepts an interrupted model call closed and re-issued (§23 fix)", () => {
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
callFailed(0, "interrupted by process restart"),
|
||||
requested(1, [user("hello")]),
|
||||
requested(1, []),
|
||||
completed(1, assistantText("done")),
|
||||
turnCompletedEv(),
|
||||
]);
|
||||
|
|
@ -736,9 +760,9 @@ describe("recovery-shaped histories", () => {
|
|||
created({
|
||||
config: { autoPermission: false, humanAvailable: true, maxModelCalls: 1 },
|
||||
}),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
callFailed(0, "interrupted"),
|
||||
requested(1, [user("hello")]),
|
||||
requested(1, []),
|
||||
],
|
||||
/exceeds maxModelCalls/,
|
||||
);
|
||||
|
|
@ -748,7 +772,7 @@ describe("recovery-shaped histories", () => {
|
|||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
result(
|
||||
|
|
@ -758,7 +782,7 @@ describe("recovery-shaped histories", () => {
|
|||
"Tool execution was interrupted; its outcome is unknown and it was not retried.",
|
||||
true,
|
||||
),
|
||||
requested(1, [user("hello"), call0, toolMsg("tc1", "echo", "…")]),
|
||||
requested(1, [{ kind: "assistant", modelCallIndex: 0 }, { kind: "toolResult", toolCallId: "tc1" }]),
|
||||
completed(1, assistantText("done")),
|
||||
turnCompletedEv(),
|
||||
]);
|
||||
|
|
@ -770,7 +794,7 @@ describe("recovery-shaped histories", () => {
|
|||
const call0 = assistantCalls(toolCallPart("a1", "fetch"));
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("a1", fetchTool),
|
||||
suspendedEv([], [{ id: "a1", tool: fetchTool }]),
|
||||
|
|
@ -784,7 +808,7 @@ describe("recovery-shaped histories", () => {
|
|||
it("accepts a live model failure closing the turn", () => {
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
callFailed(0),
|
||||
turnFailedEv("provider exploded"),
|
||||
]);
|
||||
|
|
@ -797,7 +821,7 @@ describe("recovery-shaped histories", () => {
|
|||
created({
|
||||
config: { autoPermission: false, humanAvailable: true, maxModelCalls: 1 },
|
||||
}),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
result("tc1", "echo"),
|
||||
|
|
@ -816,7 +840,7 @@ describe("invariants", () => {
|
|||
});
|
||||
|
||||
it("rejects a log not starting with turn_created", () => {
|
||||
expectCorruption([requested(0, [user("hello")])], /first event must be turn_created/);
|
||||
expectCorruption([requested(0, [{ kind: "input" }])], /first event must be turn_created/);
|
||||
});
|
||||
|
||||
it("rejects duplicate turn_created", () => {
|
||||
|
|
@ -825,7 +849,7 @@ describe("invariants", () => {
|
|||
|
||||
it("rejects mismatched turn ids", () => {
|
||||
expectCorruption(
|
||||
[created(), { ...requested(0, [user("hello")]), turnId: "other" }],
|
||||
[created(), { ...requested(0, [{ kind: "input" }]), turnId: "other" }],
|
||||
/does not match turn/,
|
||||
);
|
||||
});
|
||||
|
|
@ -849,7 +873,7 @@ describe("invariants", () => {
|
|||
|
||||
it("rejects out-of-order model call indices", () => {
|
||||
expectCorruption(
|
||||
[created(), requested(1, [user("hello")])],
|
||||
[created(), requested(1, [])],
|
||||
/out of order/,
|
||||
);
|
||||
});
|
||||
|
|
@ -858,9 +882,9 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, assistantText("a")),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
],
|
||||
/out of order/,
|
||||
);
|
||||
|
|
@ -868,7 +892,7 @@ describe("invariants", () => {
|
|||
|
||||
it("rejects concurrent unresolved model requests", () => {
|
||||
expectCorruption(
|
||||
[created(), requested(0, [user("hello")]), requested(1, [user("hello")])],
|
||||
[created(), requested(0, [{ kind: "input" }]), requested(1, [])],
|
||||
/concurrent unresolved model call requests/,
|
||||
);
|
||||
});
|
||||
|
|
@ -888,7 +912,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, assistantText("a")),
|
||||
completed(0, assistantText("b")),
|
||||
],
|
||||
|
|
@ -900,7 +924,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
callFailed(0),
|
||||
completed(0, assistantText("a")),
|
||||
],
|
||||
|
|
@ -913,9 +937,9 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
requested(1, [user("hello"), call0]),
|
||||
requested(1, [{ kind: "assistant", modelCallIndex: 0 }]),
|
||||
],
|
||||
/while tool calls are unresolved/,
|
||||
);
|
||||
|
|
@ -928,11 +952,11 @@ describe("invariants", () => {
|
|||
created({
|
||||
config: { autoPermission: false, humanAvailable: true, maxModelCalls: 1 },
|
||||
}),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
result("tc1", "echo"),
|
||||
requested(1, [user("hello"), call0, toolMsg("tc1", "echo")]),
|
||||
requested(1, [{ kind: "assistant", modelCallIndex: 0 }, { kind: "toolResult", toolCallId: "tc1" }]),
|
||||
],
|
||||
/exceeds maxModelCalls/,
|
||||
);
|
||||
|
|
@ -942,7 +966,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(
|
||||
0,
|
||||
assistantCalls(toolCallPart("tc1", "echo"), toolCallPart("tc1", "echo")),
|
||||
|
|
@ -957,18 +981,18 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
result("tc1", "echo"),
|
||||
requested(1, [user("hello"), call0, toolMsg("tc1", "echo")]),
|
||||
requested(1, [{ kind: "assistant", modelCallIndex: 0 }, { kind: "toolResult", toolCallId: "tc1" }]),
|
||||
completed(1, assistantCalls(toolCallPart("tc1", "echo"))),
|
||||
],
|
||||
/duplicate tool call id/,
|
||||
);
|
||||
});
|
||||
|
||||
it("rejects tool-result ordering that differs from original call order", () => {
|
||||
it("rejects request references whose tool-result order differs from the batch", () => {
|
||||
const call0 = assistantCalls(
|
||||
toolCallPart("tc1", "echo"),
|
||||
toolCallPart("tc2", "echo"),
|
||||
|
|
@ -976,26 +1000,25 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
result("tc1", "echo"),
|
||||
invocation("tc2"),
|
||||
result("tc2", "echo"),
|
||||
requested(1, [
|
||||
user("hello"),
|
||||
call0,
|
||||
toolMsg("tc2", "echo"),
|
||||
toolMsg("tc1", "echo"),
|
||||
{ kind: "assistant", modelCallIndex: 0 },
|
||||
{ kind: "toolResult", toolCallId: "tc2" },
|
||||
{ kind: "toolResult", toolCallId: "tc1" },
|
||||
]),
|
||||
],
|
||||
/tool-result order differs/,
|
||||
/references do not match/,
|
||||
);
|
||||
});
|
||||
|
||||
it("rejects permission records targeting unknown tool calls", () => {
|
||||
expectCorruption(
|
||||
[created(), requested(0, [user("hello")]), completed(0, assistantText("a")), permRequired("ghost", "echo")],
|
||||
[created(), requested(0, [{ kind: "input" }]), completed(0, assistantText("a")), permRequired("ghost", "echo")],
|
||||
/unknown tool call/,
|
||||
);
|
||||
});
|
||||
|
|
@ -1005,7 +1028,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
permRequired("tc1", "echo"),
|
||||
|
|
@ -1019,7 +1042,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permClassified("tc1", "allow"),
|
||||
],
|
||||
|
|
@ -1032,7 +1055,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permResolved("tc1", "allow"),
|
||||
],
|
||||
|
|
@ -1045,7 +1068,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
permResolved("tc1", "allow"),
|
||||
|
|
@ -1060,7 +1083,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
invocation("tc1"),
|
||||
|
|
@ -1074,7 +1097,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
permResolved("tc1", "deny"),
|
||||
|
|
@ -1089,7 +1112,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
invocation("tc1"),
|
||||
|
|
@ -1103,7 +1126,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1", echoTool, { execution: "async" }),
|
||||
],
|
||||
|
|
@ -1114,7 +1137,7 @@ describe("invariants", () => {
|
|||
it("rejects progress without invocation", () => {
|
||||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
expectCorruption(
|
||||
[created(), requested(0, [user("hello")]), completed(0, call0), progress("tc1")],
|
||||
[created(), requested(0, [{ kind: "input" }]), completed(0, call0), progress("tc1")],
|
||||
/progress without invocation/,
|
||||
);
|
||||
});
|
||||
|
|
@ -1124,7 +1147,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
result("tc1", "echo"),
|
||||
|
|
@ -1139,7 +1162,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
result("tc1", "echo"),
|
||||
|
|
@ -1152,7 +1175,7 @@ describe("invariants", () => {
|
|||
it("rejects sync-sourced results without invocation", () => {
|
||||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
expectCorruption(
|
||||
[created(), requested(0, [user("hello")]), completed(0, call0), result("tc1", "echo", "sync")],
|
||||
[created(), requested(0, [{ kind: "input" }]), completed(0, call0), result("tc1", "echo", "sync")],
|
||||
/sync tool result without invocation/,
|
||||
);
|
||||
});
|
||||
|
|
@ -1162,7 +1185,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
result("tc1", "echo", "async"),
|
||||
|
|
@ -1176,7 +1199,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, assistantText("a")),
|
||||
stepEvent(0),
|
||||
],
|
||||
|
|
@ -1187,7 +1210,7 @@ describe("invariants", () => {
|
|||
it("rejects completion while tool calls remain unresolved", () => {
|
||||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
expectCorruption(
|
||||
[created(), requested(0, [user("hello")]), completed(0, call0), turnCompletedEv()],
|
||||
[created(), requested(0, [{ kind: "input" }]), completed(0, call0), turnCompletedEv()],
|
||||
/completion while tool calls lack terminal results/,
|
||||
);
|
||||
});
|
||||
|
|
@ -1197,7 +1220,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
result("tc1", "echo"),
|
||||
|
|
@ -1214,14 +1237,14 @@ describe("invariants", () => {
|
|||
it("rejects terminal failure while tool calls remain unresolved", () => {
|
||||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
expectCorruption(
|
||||
[created(), requested(0, [user("hello")]), completed(0, call0), turnFailedEv()],
|
||||
[created(), requested(0, [{ kind: "input" }]), completed(0, call0), turnFailedEv()],
|
||||
/failure while tool calls lack terminal results/,
|
||||
);
|
||||
});
|
||||
|
||||
it("rejects terminal events while a model call is unsettled", () => {
|
||||
expectCorruption(
|
||||
[created(), requested(0, [user("hello")]), turnCancelledEv()],
|
||||
[created(), requested(0, [{ kind: "input" }]), turnCancelledEv()],
|
||||
/cancellation while a model call is unsettled/,
|
||||
);
|
||||
});
|
||||
|
|
@ -1229,7 +1252,7 @@ describe("invariants", () => {
|
|||
it("rejects any event after a terminal event", () => {
|
||||
const base = [
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, assistantText("a")),
|
||||
];
|
||||
for (const terminal of [turnCompletedEv(), turnFailedEv(), turnCancelledEv()]) {
|
||||
|
|
@ -1244,7 +1267,7 @@ describe("invariants", () => {
|
|||
expectCorruption(
|
||||
[
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, assistantText("a")),
|
||||
turnCompletedEv(),
|
||||
turnFailedEv(),
|
||||
|
|
@ -1259,14 +1282,14 @@ describe("derivations", () => {
|
|||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
const pending = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
permRequired("tc1", "echo"),
|
||||
suspendedEv([{ id: "tc1", name: "echo" }], []),
|
||||
]);
|
||||
expect(deriveTurnStatus(pending)).toBe("suspended");
|
||||
|
||||
const idle = reduceTurn([created(), requested(0, [user("hello")]), completed(0, call0)]);
|
||||
const idle = reduceTurn([created(), requested(0, [{ kind: "input" }]), completed(0, call0)]);
|
||||
expect(deriveTurnStatus(idle)).toBe("idle");
|
||||
});
|
||||
|
||||
|
|
@ -1277,17 +1300,16 @@ describe("derivations", () => {
|
|||
);
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
invocation("tc1"),
|
||||
invocation("tc2"),
|
||||
result("tc2", "echo", "sync", { n: 2 }),
|
||||
result("tc1", "echo", "sync", "plain text"),
|
||||
requested(1, [
|
||||
user("hello"),
|
||||
call0,
|
||||
toolMsg("tc1", "echo", "plain text"),
|
||||
toolMsg("tc2", "echo", '{"n":2}'),
|
||||
{ kind: "assistant", modelCallIndex: 0 },
|
||||
{ kind: "toolResult", toolCallId: "tc1" },
|
||||
{ kind: "toolResult", toolCallId: "tc2" },
|
||||
]),
|
||||
completed(1, assistantText("done")),
|
||||
turnCompletedEv(),
|
||||
|
|
@ -1304,9 +1326,9 @@ describe("derivations", () => {
|
|||
it("omits failed model calls from the transcript", () => {
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
callFailed(0, "interrupted"),
|
||||
requested(1, [user("hello")]),
|
||||
requested(1, []),
|
||||
completed(1, assistantText("done")),
|
||||
turnCompletedEv(),
|
||||
]);
|
||||
|
|
@ -1316,7 +1338,7 @@ describe("derivations", () => {
|
|||
it("transcript excludes the context prefix", () => {
|
||||
const state = reduceTurn([
|
||||
created({ context: { previousTurnId: PREV_TURN_ID } }),
|
||||
requested(0, [user("hello")], {
|
||||
requested(0, [{ kind: "input" }], {
|
||||
contextRef: { previousTurnId: PREV_TURN_ID },
|
||||
}),
|
||||
completed(0, assistantText("done")),
|
||||
|
|
@ -1330,7 +1352,7 @@ describe("derivations", () => {
|
|||
const call0 = assistantCalls(toolCallPart("tc1", "echo"));
|
||||
const state = reduceTurn([
|
||||
created(),
|
||||
requested(0, [user("hello")]),
|
||||
requested(0, [{ kind: "input" }]),
|
||||
completed(0, call0),
|
||||
]);
|
||||
expect(() => turnTranscript(state)).toThrowError(/unresolved/);
|
||||
|
|
|
|||
|
|
@ -119,16 +119,30 @@ export const TurnCreated = z.object({
|
|||
}),
|
||||
});
|
||||
|
||||
// `messages` contains current-turn messages only; when the turn's context is
|
||||
// a reference, `contextRef` repeats it and the resolved prefix is implicit
|
||||
// (deterministically materializable because referenced turns are immutable).
|
||||
// When context is inline, `contextRef` is absent and `messages` begins with
|
||||
// the inline context.
|
||||
// A model request is a list of REFERENCES into the turn's own events; every
|
||||
// referenced byte exists exactly once in the file. The request records only
|
||||
// what is NEW since the previous model call:
|
||||
// call 0: [{context}?, {input}] (context only when inline and nonempty;
|
||||
// cross-turn prefixes ride contextRef)
|
||||
// call N: [{assistant: N-1}, ...that batch's toolResults in source order]
|
||||
// re-issue after an interrupted call: []
|
||||
// The system prompt and tool set are NOT repeated here — they are byte-
|
||||
// identical to turn_created.agent.resolved by construction. The exact
|
||||
// provider payload is rebuilt deterministically by the request composer
|
||||
// (core), which is the same code path the loop sends through.
|
||||
export const ModelRequestMessageRef = z.discriminatedUnion("kind", [
|
||||
z.object({ kind: z.literal("context") }),
|
||||
z.object({ kind: z.literal("input") }),
|
||||
z.object({
|
||||
kind: z.literal("assistant"),
|
||||
modelCallIndex: z.number().int().nonnegative(),
|
||||
}),
|
||||
z.object({ kind: z.literal("toolResult"), toolCallId: z.string() }),
|
||||
]);
|
||||
|
||||
export const ModelRequest = z.object({
|
||||
systemPrompt: z.string(),
|
||||
contextRef: TurnContextRef.optional(),
|
||||
messages: z.array(ConversationMessage),
|
||||
tools: z.array(ToolDescriptor),
|
||||
messages: z.array(ModelRequestMessageRef),
|
||||
parameters: z.record(z.string(), z.json()),
|
||||
});
|
||||
|
||||
|
|
@ -491,32 +505,44 @@ function applyModelCallRequested(
|
|||
}
|
||||
|
||||
const context = state.definition.context;
|
||||
if (isContextRef(context)) {
|
||||
if (event.request.contextRef?.previousTurnId !== context.previousTurnId) {
|
||||
fail("model request contextRef inconsistent with turn context");
|
||||
let expectedRefs: Array<z.infer<typeof ModelRequestMessageRef>>;
|
||||
if (event.modelCallIndex === 0) {
|
||||
if (isContextRef(context)) {
|
||||
if (event.request.contextRef?.previousTurnId !== context.previousTurnId) {
|
||||
fail("model request contextRef inconsistent with turn context");
|
||||
}
|
||||
expectedRefs = [{ kind: "input" }];
|
||||
} else {
|
||||
if (event.request.contextRef !== undefined) {
|
||||
fail("model request has contextRef but turn context is inline");
|
||||
}
|
||||
expectedRefs =
|
||||
context.length > 0
|
||||
? [{ kind: "context" }, { kind: "input" }]
|
||||
: [{ kind: "input" }];
|
||||
}
|
||||
} else if (event.request.contextRef !== undefined) {
|
||||
fail("model request has contextRef but turn context is inline");
|
||||
} else {
|
||||
if (event.request.contextRef !== undefined) {
|
||||
fail("model request has contextRef on a non-initial model call");
|
||||
}
|
||||
const previous = state.modelCalls[event.modelCallIndex - 1];
|
||||
expectedRefs =
|
||||
previous.response !== undefined
|
||||
? [
|
||||
{ kind: "assistant", modelCallIndex: previous.index },
|
||||
...batchToolCalls(state, previous.index).map((tc) => ({
|
||||
kind: "toolResult" as const,
|
||||
toolCallId: tc.toolCallId,
|
||||
})),
|
||||
]
|
||||
: []; // re-issue after an interrupted call adds nothing new
|
||||
}
|
||||
|
||||
// Model-facing tool-result order must match original call order: the tool
|
||||
// messages in the request (past any inline context prefix) must be the
|
||||
// concatenated batches of prior completed calls, in source order.
|
||||
const prefixLength = isContextRef(context) ? 0 : context.length;
|
||||
const toolMessageIds = event.request.messages
|
||||
.slice(prefixLength)
|
||||
.filter((m) => m.role === "tool")
|
||||
.map((m) => m.toolCallId);
|
||||
const expectedIds = state.modelCalls
|
||||
.filter((call) => call.response !== undefined)
|
||||
.flatMap((call) =>
|
||||
batchToolCalls(state, call.index).map((tc) => tc.toolCallId),
|
||||
if (JSON.stringify(event.request.messages) !== JSON.stringify(expectedRefs)) {
|
||||
fail(
|
||||
`model request references do not match the transcript: expected ${JSON.stringify(
|
||||
expectedRefs,
|
||||
)}, got ${JSON.stringify(event.request.messages)}`,
|
||||
);
|
||||
if (
|
||||
toolMessageIds.length !== expectedIds.length ||
|
||||
toolMessageIds.some((id, i) => id !== expectedIds[i])
|
||||
) {
|
||||
fail("model request tool-result order differs from original call order");
|
||||
}
|
||||
|
||||
state.modelCalls.push({
|
||||
|
|
@ -932,6 +958,69 @@ function toolResultContent(result: z.infer<typeof ToolResult>): string {
|
|||
return typeof output === "string" ? output : JSON.stringify(output);
|
||||
}
|
||||
|
||||
// The canonical model-facing tool message for a resolved tool call.
|
||||
export function toolResultMessage(
|
||||
toolCall: ToolCallState,
|
||||
): z.infer<typeof ConversationMessage> {
|
||||
if (!toolCall.result) {
|
||||
throw new Error(
|
||||
`tool call ${toolCall.toolCallId} has no terminal result`,
|
||||
);
|
||||
}
|
||||
return {
|
||||
role: "tool",
|
||||
content: toolResultContent(toolCall.result),
|
||||
toolCallId: toolCall.toolCallId,
|
||||
toolName: toolCall.toolName,
|
||||
};
|
||||
}
|
||||
|
||||
// Resolves one model call's request references to structural messages (the
|
||||
// NEW portion that call added relative to the previous one). Concatenating
|
||||
// calls 0..N yields the full current-turn conversation for call N; the
|
||||
// request composer (core) prepends the resolved cross-turn prefix and
|
||||
// encodes to the provider wire form.
|
||||
export function requestMessagesFor(
|
||||
state: TurnState,
|
||||
modelCallIndex: number,
|
||||
): Array<z.infer<typeof ConversationMessage>> {
|
||||
const call = state.modelCalls[modelCallIndex];
|
||||
if (!call) {
|
||||
throw new Error(`no model call at index ${modelCallIndex}`);
|
||||
}
|
||||
return call.request.messages.flatMap((ref) => {
|
||||
switch (ref.kind) {
|
||||
case "context": {
|
||||
const context = state.definition.context;
|
||||
if (!Array.isArray(context)) {
|
||||
throw new Error("context ref on a turn without inline context");
|
||||
}
|
||||
return context;
|
||||
}
|
||||
case "input":
|
||||
return [state.definition.input];
|
||||
case "assistant": {
|
||||
const response = state.modelCalls[ref.modelCallIndex]?.response;
|
||||
if (response === undefined) {
|
||||
throw new Error(
|
||||
`assistant ref to unsettled model call ${ref.modelCallIndex}`,
|
||||
);
|
||||
}
|
||||
return [response];
|
||||
}
|
||||
case "toolResult": {
|
||||
const toolCall = state.toolCalls.find(
|
||||
(tc) => tc.toolCallId === ref.toolCallId,
|
||||
);
|
||||
if (!toolCall) {
|
||||
throw new Error(`toolResult ref to unknown call ${ref.toolCallId}`);
|
||||
}
|
||||
return [toolResultMessage(toolCall)];
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
// The messages this turn contributed to the conversation: its input plus, per
|
||||
// completed model call, the assistant response and its tool results in source
|
||||
// order. Failed model calls contribute nothing. The turn's context prefix is
|
||||
|
|
@ -955,12 +1044,7 @@ export function turnTranscript(
|
|||
`turnTranscript requires terminal tool results; ${toolCall.toolCallId} is unresolved`,
|
||||
);
|
||||
}
|
||||
messages.push({
|
||||
role: "tool",
|
||||
content: toolResultContent(toolCall.result),
|
||||
toolCallId: toolCall.toolCallId,
|
||||
toolName: toolCall.toolName,
|
||||
});
|
||||
messages.push(toolResultMessage(toolCall));
|
||||
}
|
||||
}
|
||||
return messages;
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue