refactor(core): unify v2 tool architecture (#31168)

packages/core/src/session/runner/index.ts

@@ -7,6 +7,7 @@ import type { ContextSnapshotDecodeError, MessageDecodeError } from "../error"
 import { SessionRunnerModel } from "./model"
 import type { SystemContext } from "../../system-context/index"
 import type { SessionContextEpoch } from "../context-epoch"
+import type { ToolOutputStore } from "../../tool-output-store"
 
 export class StepLimitExceededError extends Schema.TaggedErrorClass<StepLimitExceededError>()(
   "SessionRunner.StepLimitExceededError",
@@ -24,6 +25,7 @@ export type RunError =
   | StepLimitExceededError
   | SystemContext.InitializationBlocked
   | SessionContextEpoch.AgentReplacementBlocked
+  | ToolOutputStore.Error
 
 /** Runs one local continuation from already-recorded Session history. */
 export interface Interface {

packages/core/src/session/runner/llm.ts

@@ -20,6 +20,7 @@ import { SystemContext } from "../../system-context/index"
 import { SystemContextRegistry } from "../../system-context/registry"
 import { SkillGuidance } from "../../skill/guidance"
 import { ToolRegistry } from "../../tool/registry"
+import { ToolOutputStore } from "../../tool-output-store"
 import { SessionContextEpoch } from "../context-epoch"
 import { SessionCompaction } from "../compaction"
 import { SessionEvent } from "../event"
@@ -63,7 +64,7 @@ import { toLLMMessages } from "./to-llm-message"
  *   - [x] Authorize and execute recorded local calls through a core-owned registry hook.
  *   - [x] Persist typed success, failure, and provider-executed tool outcomes.
  *   - [x] Start each recorded local call eagerly and await all settlements before continuation.
- *   - [ ] Add scoped runtime context, progress updates, output truncation, attachment normalization,
+ *   - [ ] Add scoped runtime context, progress updates, attachment normalization,
  *     plugins, and cancellation settlement.
  *   - [x] Reload projected history and start the next explicit provider turn after local tool results.
  *   - [x] Continue for durable user steering accepted during an active provider turn.
@@ -131,7 +132,7 @@ export const layer = Layer.effect(
       }
     })
 
-    const awaitToolFibers = (fibers: FiberSet.FiberSet<void, never>) =>
+    const awaitToolFibers = (fibers: FiberSet.FiberSet<void, ToolOutputStore.Error>) =>
       Effect.raceFirst(FiberSet.join(fibers), FiberSet.awaitEmpty(fibers))
 
     // Match V1: dismissing a question halts the loop instead of becoming model-facing tool output.
@@ -185,7 +186,7 @@ export const layer = Layer.effect(
         session.location,
         agent.id,
       ).pipe(retryAgentMismatch(promotion))
-      const toolFibers = yield* FiberSet.make<void, never>()
+      const toolFibers = yield* FiberSet.make<void, ToolOutputStore.Error>()
       let needsContinuation = false
       if (promotion) {
         const cutoff = yield* SessionInput.latestSeq(db, session.id)
@@ -211,6 +212,7 @@ export const layer = Layer.effect(
       const model = yield* models.resolve(session)
       const entries = yield* SessionHistory.entriesForRunner(db, session.id, system.baselineSeq)
       const context = entries.map((entry) => entry.message)
+      const toolMaterialization = yield* tools.materialize(agent.info?.permissions)
       const promptCacheKey = /^ses_[0-9a-f]{64}$/.test(session.id) ? session.id.slice(4) : session.id
       const request = LLM.request({
         model,
@@ -219,7 +221,7 @@ export const layer = Layer.effect(
           .filter((part): part is string => part !== undefined && part.length > 0)
           .map(SystemPart.make),
         messages: toLLMMessages(context, model),
-        tools: yield* tools.definitions(agent.info?.permissions),
+        tools: toolMaterialization.definitions,
       })
       if (yield* compaction.compactIfNeeded({ sessionID: session.id, entries, model, request }))
         return yield* Effect.die(rebuildPreparedTurn())
@@ -251,16 +253,16 @@ export const layer = Layer.effect(
             yield* publish(event)
             if (event.type !== "tool-call" || event.providerExecuted) return
             needsContinuation = true
+            const assistantMessageID = yield* publisher.assistantMessageID(event.id)
             yield* Effect.uninterruptibleMask((restore) =>
-              restore(tools.settle({ sessionID: session.id, agent: agent.id, call: event })).pipe(
-                Effect.catchCause((cause) => {
-                  if (isQuestionRejected(cause) || Cause.hasInterrupts(cause)) return Effect.failCause(cause)
-                  return Effect.succeed({
-                    result: { type: "error" as const, value: String(Cause.squash(cause)) },
-                    output: undefined,
-                    outputPaths: [],
-                  })
+              restore(
+                toolMaterialization.settle({
+                  sessionID: session.id,
+                  agent: agent.id,
+                  assistantMessageID,
+                  call: event,
                 }),
+              ).pipe(
                 Effect.flatMap((settlement) =>
                   publish(
                     LLMEvent.toolResult({
@@ -322,8 +324,8 @@ export const layer = Layer.effect(
             yield* withPublication(publisher.failUnsettledTools("Tool execution interrupted"))
           if (stream._tag === "Success" && !publisher.hasProviderError())
             yield* withPublication(publisher.failUnsettledTools("Provider did not return a tool result", true))
-          const attempt = stream._tag === "Failure" ? stream : settled
-          if (attempt._tag === "Failure") return yield* Effect.failCause(attempt.cause)
+          if (stream._tag === "Failure") return yield* Effect.failCause(stream.cause)
+          if (settled._tag === "Failure") return yield* Effect.failCause(settled.cause)
           return !publisher.hasProviderError() && needsContinuation
         }),
       )

packages/core/src/session/runner/publish-llm-event.ts

@@ -218,6 +218,11 @@ export const createLLMEventPublisher = (events: EventV2.Interface, input: Input)
     }
   })
 
+  const assistantMessageIDForTool = (callID: string) => {
+    const tool = tools.get(callID)
+    return tool ? Effect.succeed(tool.assistantMessageID) : Effect.die(`Unknown tool call: ${callID}`)
+  }
+
   const publish = Effect.fn("SessionRunner.publishLLMEvent")(function* (
     event: LLMEvent,
     outputPaths: ReadonlyArray<string> = [],
@@ -408,5 +413,6 @@ export const createLLMEventPublisher = (events: EventV2.Interface, input: Input)
     hasAssistantStarted: () => assistantMessageID !== undefined,
     hasProviderError: () => providerFailed,
     startAssistant,
+    assistantMessageID: assistantMessageIDForTool,
   }
 }